Ergonomics is a word that gets thrown around a lot when we talk about peripherals like a keyboard or mouse – but what does this actually mean? The layout and shape, for starters, should adjust to the body's natural position, and these devices should be engineered to reduce any strains or pains that might arise from prolonged use.

Logitech is no stranger to incorporating ergonomic design into its long history of products, but the Signature Comfort Plus Combo – a Bluetooth-powered keyboard and mouse set – takes this a step further by putting your comfort front and center. Designed for prolonged use, these chic and sophisticated peripherals introduce additional padding and softer buttons. Functionally, Logitech is also aiming to increase productivity with customizable buttons. But how do these peripherals cope with the daily workplace or home office grind?

Logitech Signature Comfort Plus Combo: Unboxing and setup

Open up the rather large rectangular packaging, and you'll find the keyboard and mouse both tightly and securely fitted side by side, wrapped in a soft white protective fabric. You'll find some documentation – but that's it. Both devices are pre-installed with batteries (AAA and AA for the K880 keyboard and M850 L mouse, respectively), and both are designed to be used wirelessly via Bluetooth only – so there is no additional cabling. This is quite surprising, given the additional flexibility USB-C ports might offer if, say, you prefer a wired connection, or there's something amiss with your Bluetooth.

Nevertheless, we're happy that setup is thankfully painless. First, switch both devices on using the physical switch, located on the rear right of the keyboard and the base of the mouse. Getting the mouse connected is straightforward – press the button to choose your input, and you're good to go. Above the navigation keys (Insert, Home, Page Up) on the keyboard, you'll find physical buttons to switch between inputs. All three have separate LEDs that indicate the status of your connection. Hold one of these buttons down and follow the instructions in the Bluetooth settings menu.

Logitech Signature Comfort Plus Combo: Design

Logitech has adopted a standard charcoal-esque Graphite colour for both the keyboard and mouse, although a brighter off-white finish is also available. Both are strong in isolation (with Off White winning out marginally for looking a little more interesting), but it's important to have the optionality to aesthetically complement the devices you already own. Each has complementary design features; for example, the grooves on the padded palm rest mirror the ridges on the thumb rest on the side of the mouse, and there's a similar darker tone for the padded regions on both devices.

The keyboard itself is really large – with plenty of room for a full array of keys you'd expect from a conventional full-sized device, including a Function row and a number pad. Each key is slightly indented to cup your finger, with the 3D quadrilateral shape meaning the base of each key is gapless while there's plenty of room between each surface.

The shape also flows like a wave, rising then falling from left to right, which is great for angling either hand slightly outwardly as you're typing with it. Beneath the keys, this flowing motion is reflected in the curved palm rest with rows of circular contours running in the opposite direction. It's a triumphant design scheme overall. There's also room for the pairing buttons, a customizable function key, and little extras like video conferencing shortcuts. Beneath, there are two kickstands (four degrees and eight degrees) that let you alter the angle depending on what works best for you.

Where we felt the K880 lacked was in the cheapness of the sticker effect of the key labeling, as opposed to a seamless translucent effect that would disguise this. Secondly, there's no backlighting or underlighting, sadly, so using the peripherals in a darkly lit room may not be such a great idea unless you can touch-type with near-perfect precision.

The M850 L mouse is a small, compact, and simple device with one unusual feature that sets it apart from the crowd: a palm rest. Although the keyboard palm rest makes sense, we weren't sure there was much of a point in the inclusion of one here, given the plastic feels fairly soft to begin with – and there's a certain firmness to the palm rest, and you'll only feel its efficacy if you push down in an unnatural way. In addition to the two main left-click and right-click buttons are a mouse wheel, a function key beneath it, and Forward and Back buttons above the thumb rest.

Incidentally, Logitech says both devices are made from 49% recycled plastics as it strives to incorporate sustainability into its engineering. Despite that, the build quality of both is excellent, and there's no sense of frailty while you're using either.

Logitech Signature Comfort Plus Combo: K880 keyboard

Typing with the K880 keyboard is superb, and we could feel just the right amount of feedback in each keystroke – with enough mid-pitched noise to satisfy those who need audible feedback as well as that sensory punch. Keystrokes are deep, too, for a membrane keyboard, and while it won't give you the depth of a true mechanical machine, it's a brilliantly simple setup that means you can touch type for extended periods of time without discomfort or disquiet. The only oddity is that the padded palm rest can warm up, retaining your body heat during usage – but this isn't excessive and doesn't spoil the overall typing experience.

There's one minor grumble, however, which is that the feedback in each key is not consistent; most keys behave the same way when pushed, but a small handful, including the column featuring the Backspace and Enter keys in particular, have muted and understated reactions. This is, perhaps, one casualty of the elevated contour design, but it affects only a small handful of keys in that column arrangement, with the Enter key being the least satisfying to tap.

Using the keyboard itself on a day-to-day basis is made simple and effective thanks to the specific function keys that Logitech has baked into it. The brightness functions, for example, are a nod to the fact you'll likely be using a laptop for the most part, while the videoconferencing keys – including the mic and camera activation toggles – are useful shortcuts for remote and hybrid workers. We also love the inclusion of functions to control audio volume and playback – important for those who want ultimate control as well as those who love listening to music while they work.

The most useful features, however, are the ways you can customize certain keys. The AI key, for example, can be assigned to your preferred agent, application, or chatbot. The circular action key above the number pad, meanwhile, is a great tool that you can configure to your whim with the Logi Options Plus app. Smart Actions let you perform actions with a single key press, letting you chain different apps, perform specific actions (like AI replying to emails or changing your interface into a focused work mode), or even create new functions yourself. This freedom and flexibility – especially the power to chain together several functions and applications – was a great one-tap way to save ourselves time and get into a groove with a simple automation.

Logitech Signature Comfort Plus Combo: M850 L mouse

There's an elegance in the simplicity of the M850L mouse, but there are still enough interesting details in the design and functionality of this peripheral to feel far from boring and plain. The sharp ergonomic curve on the left, where you place your thumb, for instance, adds to the sense of comfort you get from using this combo, and the device itself is suitable for larger hands despite being relatively compact. The strong grip also makes the mouse feel as if it's an extension of you as you're working on a day-to-day basis – giving you precision in every movement.

There's a little strangeness in that each type of physical key has a completely different physical sensation when you're using it. The left-click and right-click buttons are muted so heavily, with a quiet click the only indication you've used either, beyond the effect on-screen. There's little force feedback either, which makes for an understated and fuss-free experience – but you may want a little more bite. The Forward and Back keys offer a bit more of a punch, but again, this is on the softer side. You get the same sensation clicking the button below the mouse wheel (which controls cursor speed), and the mouse wheel itself is more akin to a conventional one – with hefty bite and satisfying rubberized grip as you scroll up or down. You can scroll horizontally by tapping either the Forward or Back buttons, then scrolling with the mouse wheel as normal.

Logitech Signature Comfort Plus Combo: Is it worth it?

Logitech has set out to design a keyboard and mouse set steps beyond simple ergonomics to give you maximum comfort, letting you work for hours on end without pain or strain thanks to a softer and rounder design language and plenty of padding. For the most part, the Logitech Signature Comfort Plus Combo is a triumph – and you'll find yourself not only giving your home office setup, or workplace desk space, a visual boost but a functional one too.

Both devices are expertly crafted and complement each other well from a looks perspective, but the thought that's gone into functionality and day-to-day usage also pays off tremendously. We are, however, concerned about one or two bugbears that prevent this from being an outstanding proposition – one being inconsistent keystroke sensations down one portion specifically, the lack of backlighting, and the absence of a biting feedback in left- and right-clicking. These minor grumbles aren't enough to discount what's otherwise an excellent pair of peripherals.

Available in the UK for £99, it's on the pricer side when you consider Logitech's other basic combo sets – the MK250 Compact Bluetooth Wireless Combo is available for £19.99, for example – but given it's a full-sized keyboard with plenty of productivity-centric customization, as well as the additional comfort and ergonomic elements, we reckon it's worth every penny

Plenty of appealing mini PCs have hit the market in recent years, with a broad range of miniature machines available for those who want to tap into powerful and energy-efficient computing, whether that's at home or in the workplace.

Mobile working is changing, and many office spaces are already kitted out with monitors, meaning employees may not need laptops as they did during the peak of hybrid working. Desktop PCs, meanwhile, certainly have their place – but the prospect of a smaller, leaner and more affordable substitute carries a certain appeal. Some machines can match the power of a conventional desktop computer, while massively saving on energy costs, given they tend to be far more power-efficient.

But choosing the best mini PC depends on a huge number of factors, including budget, expectations, and how it'll be used. You can pick up a basic device for under $300 or push the boat out and bag a powerful machine for multimedia work or graphic design, or even consider a supercharged Nvidia DGX Spark system (or similar mini PCs) if you plan on running local AI models. Below, we've rounded up some of the most compelling mini PCs you can deploy either in your home office or as a fleet among the workforce.

What to look for in a mini PC

The best mini PCs massively save on desk space, and the smallest ones can even fit into your pocket. While they're all similar in terms of their dimensions, give or take, there may be massive variations in weight. Which is something you should keep in mind if you're not planning on keeping these devices in one place.

They're also relatively cost-effective, despite there being a handful of pricier devices out there, too. However, the very cheapest options tend to include hardware that's at least two generations behind – so you need to be careful that affordability doesn't contradict the idea of value for money, and you end up picking up machines that aren't suited to the sorts of workloads you plan on running – despite the purported cost savings. Most entry-level processors can hold their own on simple, day-to-day tasks like web browsing or spreadsheets – but you should look out for multi-core performance, which we've highlighted in our testing methodology.

If you're banking on a faster machine that can handle multiple workloads and more complex software, like multimedia applications, then you also need to pair this with plenty of RAM, with a minimum of 8GB (but ideally 16GB or above), as well as hard drive space (and speeds) if you're going to be housing massive quantities of data like images or videos, or even the essential data to run AI models on more advanced systems.

In that vein, it's hard to find mini PCs with dedicated graphics cards without blowing the budget, and you should be wary of limiting factors like throttling or airflow if you're looking on the higher performance end of the spectrum.

Most mini PCs carry a similar aesthetic, but the build quality and design elements can vary wildly. Thankfully, they're quite small and innocuous, but there's every chance you'll pick up a device that stands out like a sore thumb rather than blending into a desk setup as it should. More importantly, functionality is king – and you should decide which peripherals and connectivity options you'll need to use on a daily basis. Most business-focused mini PCs take care of the basics, but the more options you have at your disposal (including less common ports like SD Card reader), the better. The same goes for wireless standards – and you should make sure the mini PC is future-proofed against future standard upgrades.

Apple's own take on the mini PC excels in almost every quarter, with this tiny cube of a computer providing ample power for almost all users. The base M4 model is plenty powerful to run sophisticated multimedia workloads for most users, and the cost is wholly reasonable, but you can also upgrade to a supercharged M4 Pro unit for more intensive daily tasks, especially if you're a software engineer or want to experiment with small-scale AI.

Its performance and overall minimalist design – a classic trait in Apple devices – may suit most users, but so to does its connectivity options, with five USB-C ports including two at the front. Some users may bemoan the lack of USB-A ports or an SD card reader, with users needing to pick up a dongle or use the ports in their monitor. With prices starting at a very reasonable $599 (£599 in the UK), it's a highly competitive machine that also outpaces most of the Windows mini PC market. In summary, it's how all technology should be – useful, elegant, and fantastic value.

Read our full Apple Mac Mini (M4) review here

The Geekom A9 Max packs a massive punch despite its small frame. There are plenty of rivals to contend with, especially the Intel-powered MSI NUC AI+ 2MG, but this alternative, built with an AMD Ryzen AI 300 Series chip, as well as a powerful graphics card and a 50 TOPS NPU, comfortably meets that challenge. The hardware configuration is premium, especially given you can tap into 32GB RAM and a stonking 2TB SSD, but so too is its £999 price tag – making it one of the most expensive devices on the market.

However, it performed excellently in our testing, and its graphics power, in particular, is comfortably ahead of its key Windows rivals. With so many ports, you can also use it as a brilliant little AV or data hub. The machine isn't perfect, however, with niggles including a 2230 SSD slot instead of another 2280 one, and the lack of a second DisplayPort slot. That said, if you don't plan on upgrading what's already an excellent piece of machinery, then you're likely not going to encounter these issues. Overall, it's one of the best mini PCs that money can buy and the best miniature Windows computer out there right now.

Read our full Geekom A9 Max review here

Acemagic has made waves with a series of solid and reliable entry-level and mid-range mini PCs that are shedding market share from some of the bigger players. Much like the Acemagic Vista Mini V1, the Matrix Mini M1 is a highly compact and well-designed device that features a solid, albeit aging, set of components as well as a hefty amount of RAM and SSD storage space for under £400. This makes the device ideal for those who need something a bit more than entry-level without breaking the bank.

It's definitely a little underpowered if you compare it with some of the newest and most advanced mini PCs, but you know what you're getting into, and it still manages to come out with respectable performance scores in our benchmarking. We also really appreciate the connectivity options, as well as the standards used (including USB 4.0 support), despite the lack of an SD Card reader – although it's hardly a deal breaker. Overall, it's a compact, nice-looking, and capable machine that does a bit more than the bare minimum for a near entry-level price.

Read our full AceMagic Matrix Mini M1 review here

The pace of miniaturization is compelling, with AMD, Nvidia, and others engineering AI-focused mini PCs that far outpace most conventional desktops and laptops. The Acer Veriton GN100 is a classic Nvidia DGX Spark mini supercomputer that is engineered for advanced users and AI practitioners who want to experiment with and deploy projects within their organizations. Frankly, there's very little separating the different DGX Spark machines, given they are all made to a similar spec. But Acer's take on this mini PC has an interesting design and very attractive pricing.

Although you can't use Windows 11, the custom Linux distro is highly intuitive and designed to coax even those dabbling in AI for the first time into trying out different programs and projects. With the Grace Blackwell Superchip, there's enough power to handle mid-sized local large language models (LLMs) to great effect, while you can also connect it with another system to double up on power. For businesses striving to take the first steps in their AI journeys, machines like the Acer Veriton GN100 are a brilliant way to get started.

Read our full Acer Veriton GN100 review here

How we test mini PCs

Our rigorous testing methods at ITPro are designed to put each device through its paces and glean how it performs against the expectations we have from the outset, keeping the end user and their expected usage in mind at all times. Whether you're picking up a device to augment your home office workspace or making a huge purchasing decision over a fleet of devices, we know that it's important to get as much information about the device itself, but also what it's like to use it and whether it'll fit the bill for how you plan on using it. We also weigh up whether the device offers value for money – comparing it with counterparts in the field to measure how it stacks up.

For mini PCs, we begin by studying the general design and build quality – highlighting any specific quirks or areas of strength in our appraisal. For instance, we'll remark on the connectivity options, the standards of those options, and the placement of any ports, alongside any features we feel could be missing. Our review will also consider the size and shape of the machine, as well as its weight, comparing this with similarly categorized mini PCs you may also be considering.

Next, we'll use a set of benchmarks to determine the machine's performance levels and general capabilities. These industry-standard tools help us determine how well it can cope under day-to-day load and intensive strain, as well as what kind of user it may be suited to. Our assessments also measure the speed of the storage system (whether that's an SSD or otherwise). We'll use the machine actively as our primary work unit and push it with software like Photoshop or Lightroom too to get a feel for how it copes under load. For specialist machines like workstations or AI-centric hardware, we'll run local AI workloads and see how the machine copes under sustained duress.

Everpure could face stiff competition in its ambitious data management pivot, as the company plans a deep push into what is already a crowded market.

At Pure Accelerate 2026, the company announced the launch of a new Data Intelligence platform aimed at streamlining data management, allowing enterprises to simplify data visibility, discovery, and control.

The launch builds on Everpure’s gradual slide toward the data management space over the last three years. At Pure Accelerate 2025, the company unveiled the Enterprise Data Cloud (EDC), designed to unify both storage and data management capabilities for customers.

The pitch by Everpure – which until recently went by Pure Storage – rests on the idea that enterprise data architectures need a comprehensive rethink, largely due to the demands placed on infrastructure by generative and agentic AI.

Many enterprises have been locked in an ‘application-centric’ approach with their IT environments for decades, which saw individual pockets of data fed into centralized applications.

Given software was expected to eat the world until a few short years ago, these practices made sense. However, this data is often locked in disparate, fragmented locations, creating bottlenecks.

According to CEO Charlie Giancarlo, Everpure wants to flip this trend on its head, enabling enterprises to adopt a ‘data centric’ – or ‘data primacy’ – approach whereby these data sets are mapped and unified.

“Instead of the apps themselves connecting to all of your data, we need to construct a context map, we need to understand the data itself, we need to understand the context of each of the data systems, each of the data environments,” he told attendees.

“Think of this as a universal data intelligence universalizing across your enterprise, where you understand not only the context of individual data sets, but how these data sets relate to other data sets as well.”

Everpure argues that this approach will reduce the number of data integrations required by enterprises, breaking down silos, barriers, and creating a single source of truth.

This is vital with the advent of agentic AI, according to Everpure. Agents require deep integration with enterprise data sets to provide context, and the “context layer” that Everpure’s new platform aims to create will, the company says, help streamline access.

That means CRM data, ERP data, you name it; this context layer is unified and far more accessible.

Everpure on a collision course

With the data management pivot, Everpure is redefining itself as more than just a ‘pure’ storage provider. It’s a natural evolution for the company given that others within its immediate sphere of influence have done the same.

Crucially, it’s a reaction to the fact that a myriad of vendors and service providers now want your enterprise data in their own platforms. Salesforce wants your data for its Agentforce service, while SAP wants you doing all your work with it, which just creates further fragmentation.

Everpure’s core value proposition here rests on unifying the various data sets you’re storing on their hardware, and relying on its solutions to consolidate and manage them, regardless of the other services you’re using.

It’s a bold move by Everpure, but one that increases the risk of butting heads with some serious big hitters in the tech industry, according to Henry Baltazar, research director at 451 Research.

“If we look at all these vendors in different silos, the reality is that to grow their revenues they have to go to adjacent. That’s what everybody’s doing,” he said.

“So, if you look at the backup side, data protection side, where did they go? They went to security and they bought the DSPM (data security posture management) vendors, kind of like what [Everpure] just did.

“[Everpure] went from storage … into more data management and governance. So, you've got these vendors on the same collision course, essentially, because that's the path forward to getting more influence, getting more revenues, and being able to talk to different stakeholders.”

Much of the same?

Data centricity is by no means a new concept. Indeed, ETL (extract, transform, load) processes date - and have been talked about at length - for more than a decade, begging the question of what’s different this time around.

This exact question was posed to Giancarlo during a Q&A with assembled media, who argued that most efforts on this front haven’t materialized. He added that as data intelligence also enables enterprises to create a shared context layer for data, that marks a “major change”.

ETL processes involved creating dedicated copies of data, which were then distributed to certain applications. The result, he said, is more fragmentation and “shadow data” that provides no value except to an individual application.

“I would argue that IT architectures, for the most part, didn't go down that direction,” he said.

“They knew data was important, but what they did is they took that important data and they copied it for every new application that they wanted to run, including all the analytics.”

AI, Giancarlo noted, appears to have been the “breaking point” for many enterprises, which have found themselves copying and replicating data sets for the various SaaS vendors they work with. It’s spiralled out of control, and Everpure thinks it has the remedy.

“AI is that breaking point for two reasons. One is the method that enterprises have started going down the AI route replicates the analytics portion, which is I copy the data again for this AI purpose, or I copy the data again for that AI purpose, and then you also have all of the, in particular the SaaS players saying, well, for my AI agent to be more used to you, you've got to give me a copy of your other data,” he explained.

“How many times are you going to replicate your data in all these different places? So I think it's reached a really clear breaking point now.”

New targets

While Everpure might face tough competition in the future, the opportunities are immense, according to Greg Macantee, senior analyst at 451 Research, particularly in terms of engagement with new enterprise audiences and leadership roles.

Simply put, Everpure isn’t just catering to IT and infrastructure operations leaders with data intelligence, it’s also opening up to CIOs and CTOs.

“It helps [Everpure] get into different personas and new accounts,” he told ITPro. “That’s kind of the contemporary issue, who are going to be your champions outside of the storage architect? Is it going to be the CISO, is it going to be the application owners, maybe the data engineer [or] data scientist?”

Making the pitch

One key question with Everpure’s data management shift is whether it has the visibility on this front. While the rebrand earlier this year reflected this, the company’s bread and butter since its inception has been flash storage.

Based on Giancarlo’s comments in the press Q&A, what is clear is that the firm is preparing for a big sales and outreach drive – and it’s a source of huge excitement internally.

“I think it’s a fairly extensive addition to what we are able to provide in the go-to-market area,” Giancarlo told journalists. “We’re very much bulking up in this area.”

“We’re hiring extensively, we will have forward deployed engineers who know a lot about our product and how to deploy it, but more importantly, can help consult with customers to be able to deploy it and start to go down this path towards data primacy.”

Partners will also play a “really important part” in championing the new approach, Giancarlo said, with systems integrators working on the frontline driving change.

“That opens up a huge opportunity for the channel, because all of that requires a lot of expertise, a lot of help, a lot of services, and so we believe that it’s a great opportunity for the channel,” he told journalists.

Execution will be critical for Everpure if it’s to have impact with this data intelligence drive, but it’s leading by example. Giancarlo revealed the company is deep into its own re-architecturing process internally, and it’s a “multi-year journey”.

Convincing customers to embark on a lengthy change process will most certainly be a challenge given the stakes at play for those dabbling in agentic AI.

Everpure is in a peculiar position when it comes to discussions around data sovereignty.

As a hardware vendor, it isn’t subject to the same scrutiny as cloud hyperscalers, but as a US company, it does face scrutiny from EU customers worried about dependency on foreign tech providers.

Patrick Smith, Everpure’s EMEA CTO, told ITPro the company is keen to engage in proactive and pragmatic discussions on this topic. Indeed, Everpure’s position on the matter is that once their storage solutions are deployed in a customer data center, it’s a hands-off approach.

“We like to be very upfront about that,” Smith told ITPro at Pure Accelerate 2026.

“When customers buy our tech and deploy it in their data centers, it’s under their control. We’re obviously a US provider of technology, but it’s under their control.”

Smith admits that the situation becomes tricky when it comes to telemetry monitoring. Everpure’s Pure One software solution collects telemetry data from arrays sitting in that data center, tracking configuration and performance metrics, and it runs in the cloud.

Everpure’s view on this, according to Smith, is that because this doesn’t contain mission-critical data, it’s not actually a big customer talking point in terms of data sovereignty posture. They at least have a choice here, though, he added.

“Every customer needs to make their own decision on that,” he told ITPro. “Some customers may choose to run what we call ‘dark sites’, where they do their own monitoring and telemetry in their own on-prem systems.”

Shifting priorities

Engagement with Everpure customers certainly shows data sovereignty has become a big concern over the last two years, according to Smith. Rising geopolitical tensions and regulatory considerations have prompted a rethink for many European firms on how they handle their data - and crucially, who they choose to safeguard it.

In response to legislative changes, a host of hyperscalers have launched dedicated ‘sovereign cloud’ services for European customers, including Amazon Web Services (AWS), Google Cloud, and Microsoft.

Meanwhile, the cloud repatriation trend, which dominated 2022 and 2023, is making somewhat of a comeback as enterprises look to bring data back on-prem to bolster both resilience and data protection capabilities.

The majority (97%) of companies plan to ditch public cloud in favor of in-house infrastructure, particularly for mission-critical workloads and data, according to a survey from Node4 from June last year.

The majority (97%) of companies plan to ditch public cloud in favor of in-house infrastructure, particularly for mission-critical workloads and data, according to a survey from Node4 from June last year.

“Right now, a lot of organizations are revisiting their cloud strategy, so digital sovereignty and data sovereignty are one of the triggers for them saying we’ve almost sleepwalked into this position where we have such heavy dependency on the hyperscalers,” Smith said.

Smith revealed that a prospective customer, a big European bank, was actively reconsidering a long-running cloud strategy while other customers have been making moves on this front.

“It’s a case of what was once a ‘cloud-first’ strategy, or a ‘cloud-only’ strategy, is now more of a ‘cloud considered’, and the criteria that go into that consideration are criticality of applications, the sensitivity of data,” he said.

Everpure’s sovereignty pitch

Everpure’s pitch to customers on this front is that if customers want a pure play hardware provider for their data center infrastructure, they’re the go-to option here, particularly for those repatriating workloads back on-prem.

“People have now got used to that cloud consumption model, and so the ability for us to deliver a cloud consumption model that spans on-prem, colocation, and the public cloud gives them that,” he claimed.

If they’re looking for more of a cloud-esque approach, then the company’s growing array of data management tools, such as its recently launched Data Intelligence platform, also represents a huge opportunity for Everpure.

The new platform aims to simplify data management processes, enabling users to improve data visibility and break down long-standing infrastructure silos.

Fred Lherault, EMEA CTO, Emerging at Everpure, said data visibility improvements delivered by the platform could provide marked benefits in terms of shoring up sovereignty practices.

“For a lot of organizations, they can't even say what data they have where, so even before they start thinking about changing, they can't answer basic questions, so that's one of the things I'm really excited about with data intelligence, is being able to answer those types of questions,” he told ITPro.

“That is actually the same set of questions that the compliance team is asking, that the cyber team is asking, that the AI team is asking,” Lherault added. “So there is really an opportunity to go and help a lot of people in a given organization that are trying to come to grips with that.”

Dependency concerns

For EU customers specifically, dependency isn’t just a case of who your cloud provider is and whether they’re a US-based company, Smith noted. Running parallel to data sovereignty considerations, concerns are growing over energy dependency and the impact of outages.

Two high-profile cloud outages at AWS and Microsoft last year shone a light on the scale of dependency in Europe and the potential domino effect that a major incident could cause for European organisations.

“I think the outages last year at both AWS and Azure were interesting in the fallout of those outages impacting UK-based companies, that perhaps they didn’t realize that they were dependent on,” he said.

“I think it gets even more interesting when you look at the degree of separation, because suddenly you realize that SaaS companies that you’re dependent on, where they are running their technology.

“So you may think that you’re nicely insulated or isolated from those outages, and then you suddenly find that your SaaS provider is impacted. It expands the scope of sovereignty.”

Sovereignty washing

With a big push on data sovereignty across Europe, the debate over what exactly constitutes ‘sovereign’ in terms of services has reached a boiling point, with ‘sovereignty washing’ emerging as a new industry buzzword.

CISPE, for example, has questioned the legitimacy of some services, going so far as to launch a framework designed to verify a provider’s sovereignty credentials.

Smith said the debate over true sovereign services cannot be ignored and weighs heavily on the minds of many executives. During a panel discussion ahead of Pure Accelerate this week, Smith noted that concerns over this topic will likely continue.

“I think that sovereignty washing is really important,” he said. “It’s going to come down to lawyers discussing it, because it’s all about the small print. Technologists aren’t going to make that decision on what good looks like; it’s going to be lawyers.”

“That comes with the hyperscalers bringing out regions that they call ‘sovereign regions’, and it’s a label, but what does it actually mean? It deals with the locality of the data; it may deal with the country and citizenship of the people who operate it, but does it tick all the boxes necessary to eliminate the risk?”

Smith compared sovereign services to insurance policies, noting that enterprises need to be “absolutely certain” that they’re getting what they pay for.

“You don’t want to go down a sovereign path and then suddenly find when things don’t go well that it hasn’t addressed all of the risks.”

FOLLOW US ON SOCIAL MEDIA

Rising hardware costs have placed significant pressure on Everpure in the last six months, according to CEO Charlie Giancarlo – but the company is keen to “share the pain” and ride out the storm.

Speaking at the company’s annual Pure Accelerate conference in Las Vegas this week, Giancarlo hailed customer resilience in the wake of rising costs, noting the company is frightfully aware of the difficulties many have encountered.

“Thank you for sticking with us,” he told attendees. “I apologize for the industry as a whole. We’ll eventually get through this. Unfortunately, it’s probably going to take a little bit of time.”

Giancarlo noted that Everpure is working closely with customers, but also partners, on orders to alleviate strain and pain moving forward.

Rising hardware costs have rocked the global tech industry in recent months, with enterprises and consumers alike feeling the pinch. According to Giancarlo, NAND prices have risen between 600-800% since October last year.

Naturally, this has resulted in price increases for customers, with the cost to produce the company’s products surging around 300%. While price increases have impacted customers, Giancarlo noted that the company is still operating at the “low end of our gross margin” rate.

“The lowest end of our gross margin rate that we’ve ever operated,” Giancarlo emphasized. “We want to be able to share the pain with you.”

If Everpure is indeed operating at the lower end of gross margin rates, then the company’s recent earnings report highlights a booming period for the data storage and management firm.

Revenue topped $1.1 billion in the first quarter of its 2027 fiscal year - its first billion-dollar quarter and a “major milestone” Giancarlo noted during the opening keynote.

“We raised later”

Notably, Giancarlo told attendees that Everpure “raised later” with regard to price increases compared to competitors.

He later doubled down on the company’s stance in response to a question during a press Q&A session, noting that increases haven’t reflected the broader price surges.

“The price that we have to pay for semiconductors, which, as I said, has gone up somewhere in the six to eight range. It’s breath taking. Our prices have gone up by far less than that, and we’re operating at a lower gross margin.”

“We’ve been following competitors’ price raises,” Giancarlo added. “They’ve been sooner than ours, and they’ve been larger than ours.”

Competitors have indeed raised prices in response to the memory crisis. In February this year, Dell Technologies hiked prices and tightened discounts in response to the crisis, per reports from Seeking Alpha.

Speaking during a fourth-quarter earnings call at the time, Dell COO Jeff Clarke revealed the firm was “compressing discounting”.

HPE, meanwhile, also adopted an “agile pricing posture” in response to rising DRAM and NAND costs.

FOLLOW US ON SOCIAL MEDIA

Nvidia first announced its DGX Spark mini supercomputer last year – and we've since seen a flurry of machines made by OEM partners hit the market. We've previously reviewed the outstanding Asus Ascent GX10, with a handful of alternatives like Lenovo's ThinkStation PGX and Acer's Veriton GN100. We've currently got our hands on the Dell Pro Max with GB10 – a mini PC designed to seamlessly fit into your existing IT setup.

This machine is designed specifically for what you would ordinarily describe as "power users". These include data scientists, researchers, as well as budding and current AI practitioners who want the freedom and flexibility to experiment with AI deployments within the enterprise, among a handful of other cohorts.

As such, it's powered by a custom Linux distro and hardware specifically – the Grace Blackwell Superchip – configured to let you run meaningful AI workloads on your device, rather than relying on a cloud service or remote communications. That means it's certainly not a device for everyone – and should only be considered for those who plan on making the most of the on-device AI capabilities these machines allow.

While DGX Spark-based systems are all very similar, they look very different and are also priced differently. The Dell Pro Max with GB10 comes in at a hefty price approaching £6,000 in the UK, which is close to £2,000 more than some of devices sold by its competitors. Can it make up for this asking price in other areas?

Dell Pro Max with GB10 review: Build and design

The device itself isn't much bigger than conventional mini PCs and fits snugly on your desk thanks to its standard dimensions of 150 x 150 x 51mm – which is consistent across all DGX Spark systems from partner OEMs. They're all roughly the same weight, too, with the Dell Pro Max with GB10 weighing 1.3kg – versus the Acer's 1.2kg and the Asus' 1.4kg.

Also identical between these machines are the ports you'll find on the rear. These include three USB-C 3.2 Gen 2x2 (20Gbps) ports with DisplayPort 2.1 compatibility, one USB-C Gen 2x2 with PD in, an HDMI 2.1 port, an Ethernet port, and Nvidia's ConnectX-7 SmartNIC port used to run two of these machines in parallel. There is a marked lack of USB-A ports, so you'll need additional dongles or docks to use peripherals, outside of connecting them to a monitor and accessing them through a USB-C DisplayPort setup.

With so little separating the different models, OEMs must make the most of their physical appearance, and Dell has adopted a more understated and neater design outlook than some of the alternatives. The machine's front is covered with a hexagonal honeycomb lattice grill, with Dell's logo centered perfectly. One neat addition is a tiny power LED indicator on the left of the front, alongside an LED built into the USB-C power cable. The rear adopts the same lattice design on the upper half, with the ports laid out in the standard format on the lower half – all of which are flush against the surface.

Overall, it exudes more of an overtly corporate feel – underlined by a relatively uninspiring grey finish. It won't look out of place in any office environment, but it does lack the flair that you might get with the Acer Veriton GN100, for example.

Dell Pro Max with GB10 review: Specs and performance

The GB10 Superchip is the beating heart of the Dell Pro Max, with GB10, and lets users tap into a Grace CPU and Blackwell GPU combined to deliver an efficient setup for AI-centric work. The 20-core Arm processor combines 10 Cortex-X925 performance cores and 10 Cortex-A725 efficiency cores.

Combined with the 5nm GB20B GPU, you'll have the local potency to reach a performance capacity of 1,000 teraFLOPS of AI performance with the learner and more efficient FP4 precision, and a maximum of 170 teraflops with the industry standard FP16 precision. Both are formats for training and inference, with the 4-bit floating point format allowing for lower-precision but accelerated AI performance.

The machine can also support models of up to 200 billion parameters, which is a little more than the size of GPT-3 – but this increases to a maximum of 405 billion parameters if you connect two of these together. Each one is also capped at a maximum power draw of 240W – with a 280W power brick – but you won't encounter it unless pushing the device to its absolute limits.

Comparing DGX Spark systems to conventional machines is hard – because they're a different class of device with specific purposes, which vary from general-purpose PCs designed for almost everyone. However, when testing with Geekbench 6, we registered a massive single-core score of 3,123 and a multi-threaded performance of 19,708. This is in line with similar systems and notably higher than the comparable Framework desktop.

Unscientific testing and experimentation also pointed to easy and quick setup on AI tasks, such as image generation with ComfyUI, with the machine running quietly throughout – while producing very little heat. While not as fast or as powerful as industrial-scale systems or remotely accessible infrastructure, we were impressed with how much you can push a small machine like this.

Dell Pro Max with GB10 review: Features

Setting up the DGX OS platform, which is based on Ubuntu 24.04 LTS (for ARM), is an absolute breeze. The overall look and feel of the system closely follows Nvidia's design aesthetic and colour palette, and is fairly intuitive to use. Helpfully, there are a few useful tools that you can use out of the box to get started on running AI tasks.

The best place to start would be Nvidia's playbooks, a web-based platform with different recipes that centre on specific AI-related tasks. For first-timers, this library of tutorials is the best place to experiment with the tools the DGX Spark-based system can let you tap into. This won't be accessible, or even necessary, for the average user – but anybody within an organisation with a willingness to engage in AI will find it an incredibly useful way of getting to grips with what this machine is capable of.

Each recipe includes the core purpose, the practical risks of engaging in the project, and a rough time estimate to get it up and running. Setting up DGX Dashboard, for example, takes roughly half an hour and provides a user interface from which you can manage your system, manage resources, and launch JupyterLab instances. ComfyUI, too, is a great place to start – taking about 45 minutes to build and run your own AI image generator.

Generally, these playbooks are easy to follow – but there's an active community, too, that you can check in with if you run into any trouble. Which you inevitably will if you're just getting started. But the triumph in this setup is the way it democratises AI development – breaking down rather complex and unwieldy processes into digestible and easy-to-follow chunks; that's much like the device itself, which gives a far broader cross-section of the workforce access to tools and capabilities once reserved for only the most technically minded.

Also useful for power users hoping to get more from their machine is the capacity to connect two DGX Spark systems together. Nvidia Sync, which is a new feature, lets you connect your local desktop or laptop with the machine, so you can operate it remotely within the same subnet. You'll be able to run workloads or launch applications remotely and even share the display between your machines. It's a perfect way to offload tasks onto this specialised piece of hardware if you prefer working primarily with your laptop.

Dell Pro Max with GB10 review: Is it worth it?

This new class of machine is the perfect way to get started on your AI experimentation and implementation journey within the enterprise. The systems are incredibly compact, benefit from fantastic specs courtesy of the GB10 superchip, and have a deep pool of 128GB unified memory to dip into, alongside plenty of SSD storage.

Whether the Dell Pro Max with GB10 is worth it depends on how it compares with its counterparts. Admittedly, it's expensive, but it adopts a relatively understated yet mature design, which contributes well to airflow. The power draw capacity is also slightly higher at 280W versus the standard 240W. For AI practitioners and those just getting started, it's a brilliant device – it's just worth weighing up whether the subtle tweaks are worth the additional cost. If you're all-in on the Dell ecosystem, it'll complement your existing setup perfectly, but the budget-conscious would understandably be drawn to the most cost-efficient DGX Spark system out there – which isn't this one based on current retail prices.

Dell Pro Max with GB10 specifications

Geekom is among a cohort of increasingly prominent Chinese manufacturers building a diverse collection of snug and supercharged miniature computers that threaten to replace conventional desktop machines. The manufacturer's AMD-based mini PCs have excelled under scrutiny, including the outstanding Geekom A7 Max, while ITPro also described the Geekom A5 2025 Edition as the best mini PC on the market at the time of our review.

Both these machines are aimed at users with higher demands, but the Geekom Air12 2026 Edition is more in line with the Beelink Mini S13 or Acemagic Vista Mini V1. These are relatively stripped-back mini PCs and aren't suitable for the intensive or graphics-based workloads their weightier cousins are; rather, they're largely for day-to-day performance. With a basic Intel Pentium chipset alongside 8GB RAM, and an attractive £299 price tag (if you can find the device on sale), can this compact and affordable mini PC turn heads for those seeking an entry-level option?

Geekom Air12 2026 Edition: Design

Fewer PCs out there are more compact than the Geekom Air12 2026 Edition, with the manufacturer claiming it's the thinnest mini PC ever. That's due to its incredible slenderness, with dimensions measuring 117 x 112 x 34.2mm. Whether or not this is accurate depends on your parameters, however. For instance, the Acemagic Vista Mini V1 boasts a depth of just 32mm – but it isn't currently available.

Its 460g weight is also slightly below average, but it's twice as heavy as the Beelink (275g) and the Acemagic (245g). Nonetheless, this pocket-sized device is delightfully compact and impressively robust, with the chassis able to withstand up to 200kg of pressure. The weight isn't a huge problem at all, given it's light enough to begin with.

The two-toned gray and black chassis gives the device some character, while the four rubber feet handily let it rest on any surface without causing marks or scratches, and subtly rounded corners take the edge off. The look, similar to that of the A5 2025 Edition, is certainly a lot more subtle than a handful of devices out there. The Lenovo ThinkCentre M70q, for instance, is much louder.

There's a decent array of connectivity options with this device, too. The wireless compatibility standards of Wi-Fi 6 and Bluetooth 5.2 – entirely standard for this power class – won't offer that much future-proofing. Otherwise, you'll benefit from a slew of options front and back. On the face of the device is one USB-C port and one USB-A 3.2 Gen 2 port, as well as a 3.5mm headphone jack and a power button (with a useful LED indicator). On the rear is a DC in, mini DP 1.4, Ethernet port, two USB-A 3.2 Gen 2 ports, HDMI 2.0 port, and a USB-C port with DP 1.4 Alt Mode. On the left-hand side is an SD card slot, and on the right a lock slot. It makes for a very strong collection of ports and connectivity options.

Geekom Air12 2026 Edition: Specs and performance

Fitted into the Geekom Air12 2026 Edition is a five-year-old Tiger Lake-U processor created initially for budget notebooks. The Intel Pentium Gold 7505 CPU is as entry-level as you can get, with the 10nm chip featuring two cores with a base frequency of 2GHz and a turbo capacity of 3.5GHz. Geekom has paired this with an Intel UHD Graphics chip, 8GB DDR4 SDRAM, and a 256GB PCIe Gen 3 SSD. By today's standards, you shouldn't ever consider a configuration weaker than this – and even this is hard to justify for anything beyond basic performance, unless you can't stretch your budget beyond £300.

The late-2020 CPU featured in this mini PC has been designed for efficiency and better performance in simple, single-core tasks rather than competence in multi-core processing. This is reflected in our testing with Geekbench 6, with the Geekom Air12 2026 Edition registering a single-core result of 1,460 and a multi-core result of 2,823. Performance is generally in line with Intel Processor N150-fitted mini PCs.

The Beelink Mini S13 Mini PC, for instance, registered 1,242 and 3,029 while the AceMagic Vista Mini V1 scored 1,278 and 3,169. Both these models are a tad cheaper at the time of writing, so evaluating performance depends on how you plan on using these machines. Given they're suited for workloads that utilize just one core, the Geekom Air12 may have the edge with its slightly better single-core performance.

An OpenCL score of 7,807 for GPU performance, meanwhile, is weak – especially when you compare it with the Acemagic Kron Mini K1's result of 9,630 with the same graphics chip. Finally, we should note the pretty poor SSD speeds of 557MB/s for reads and 480MB/s for writes. This is in line with what we saw from the Beelink's 510MB/s and 480MB/s, but considerably less than speeds of at least 3,000MB/s and 2,000MB/s for reads and writes in mid-range and higher-end mini PCs (that, admittedly, cost double the price).

Geekom Air12 2026 Edition: Is it worth it?

Price is king with entry-level mini PCs, and whether the Geekom Air12 2026 Edition is worth it depends on how good a deal you can get. We simply wouldn't consider anything higher than £299 – which is even slightly pricier than the Beelink Mini S13 Mini PC and the Acemagic Vista Mini V1. But its form factor is delightfully miniature and robust, its single-core performance is marginally better thanks to the Intel Pentium Gold CPU, and you can take advantage of a useful selection of ports for all contexts.

The simple fact is that your money will only get you so far in this power class. Increasing your budget by even just £100-£200 suddenly opens the door to modern hardware, better SSD speeds, and higher hardware standards. But if you're operating under a strict and limited budget, the Geekom Air12 2026 Edition is a solid entry-level mini PC, especially if you can bag a good bargain on one.

Geekom Air12 2026 Edition specifications

Windows All-in-One PCs are among the most difficult devices to get excited about. Designed and built to a price and aimed primarily at the most utilitarian end of the corporate computing market, they are the ultimate white goods of the PC world.

It thus came as something of a surprise when Medion's UK PR team sent over the press release for its latest Signium 27 A1 AIO. Here was a device that looked rather smart and stylish despite being aggressively priced at around £750, which undercuts the likes of the MSI Pro AP242P 14M and the Lenovo ThinkCentre M90a Gen 5.

Medion Signium 27 S1: Design

Windows' AIOs are usually boring black blocks of plastic with all the design interest of a suburban bus shelter. Credit then to Medion for cooking up something just a little different. The Signium 27 is still made of plastic, but the matte silver paint finish matches the elegant metal stand, giving the impression of something beyond call-centre utilitarianism. At first sight, you'd be forgiven for thinking it's made from aluminium rather than plastic.

It's impressively svelte, too. The screen bezels are just 15mm wide at the top and bottom, and a little over half that at the sides. The base of the stand is a compact and narrow affair that takes up a minimum of desk space, and the whole assembly only weighs 3.8Kg

Under the centre of the screen are three LED-lit squidgy rubberised buttons. The larger middle one is a simple on/off button, while the smaller one to the left opens a simple menu to switch inputs between the PC and the HDMI-in port and adjust the brightness when used in monitor mode. The right button disables the webcam, which seems like overkill given that there's also a manual privacy shutter.

Adjustment is very limited, with nothing available in terms of height or rotation. What you can do is tilt the screen forward and back between -20 degrees and +40 degrees. To further limit the positional possibilities, the two-piece metal stand doesn't feature anything in the way of a VESA attachment, so fixing it to a monitor arm is impossible without some sort of adapter.

Medion has equipped the Signium with a decent range of ports. On the right-hand side of the unit, for easy access, there are two 5Gbps USB ports: one Type-A and one Type-C.

Around the back, there are another three more USB-A ports, two 480Mbps 2.0 and one 10Gbps, and a second Type-C port, this time supporting 10Gbps data as well as DisplayPort video and PowerDelivery charging. Accompanying those data ports is a gigabit RJ-45 LAN port, two HDMI 1.4 ports, one in and one out, a 3.5mm combination audio jack, a Kensington lock slot, and a DC-in jack for the 120W power supply.

Wireless communications come courtesy of an Intel AX211 card, which supports 6GHz Wi-Fi 6E and Bluetooth 5.3. Neither of those protocols is the latest thing, but we wouldn't have been that shocked if the Medion had only supported 2.4/5Ghz Wi-Fi, so it's still a win in our book.

The bundled mouse and keyboard look rather more desirable than they are, thanks to a smart black and silver livery.

Both work well enough, though the keyboard lacks a backlight, has a rather shallow key action, and there's nothing in the way of angle adjustment. The mouse is too flat for our tastes.

On the plus side, the layout of the keyboard is good, there's a full-size numeric keypad, a wide range of status lights, and full-sized left/right arrow keys.

Both the keyboard and mouse feature a MicroUSB port for recharging duties. Thankfully, Medion includes a short charging cable. Rather than depending on Bluetooth, the mouse and keyboard connect via a 2.4Ghz wireless dongle.

Undo two Philips screws on the rear, and you can slide off the panel that covers the internal components. Not only is there space for a second M.2 2280 SSD and easy access to both SODIMM slots, but there is also space for a good old-fashioned 2.5in SATA3 HDD. This level of upgradability was wholly unexpected but most welcome.

Medion Signium 27 S1: Display, audio & webcam

Windows AIOs tend to come with very basic displays, but the Medion is just a little better. Granted, it's a standard IPS affair, and with a 1,920 x 1,080 matrix spread over a 27-inch diagonal, it's not the sharpest display around with a pixel density of just 82dpi.

Peak brightness is also nothing to write home about: 240cd/m2, though that is adequate for indoor use. There's rather more color on offer than with the likes of the MSI Pro AP242P and the Lenovo ThinkCentre M90a, with color gamut volumes of 126.5% sRGB, 89.6% DCI-P3, and 87.2% Adobe RGB.

For comparison, the MSI display only covers 92.7% sRGB, 65.6% DCI-P3, and 63.8% AdobeRGB, respectively. The Lenovo is better at 104.7% sRGB, 74.1% DCI-P3, and 71.2% Adobe RGB, but still can't match the Medion.

Combine that with a contrast ratio of 978:1, a Delta E color variance of 2.64 against the sRGB profile – just below 3, at which point the more acute observer will start to notice color deviations – and you get a screen that looks nicely colorful and natural to the naked eye.

On a more mundane level, the screen has an effective matte anti-reflective finish, so you don't need to worry about reflections from bright lights. Unsurprisingly, there's no support for HDR content, and the refresh rate is effectively limited to 60Hz at 1,920 x 1,080.

We say effectively because you can actually push the refresh rate to 75Hz, but when you do, the resolution drops to 1,080 x 720. Increase the resolution back to FullHD, and the refresh rate defaults back to 60Hz.

Buried in the bottom left and right corners of the Medion's cabinet are two loudspeakers of unstated output and dubious origin. They are better than nothing, but not by a wide margin.

Peak volume is a lowly 69dBA as measured against a pink noise source at a distance of 1m, and the sound they produce is constrained, muddy, brittle, devoid of bass, and generally tiresome. For a casual Zoom call, they'll do, but for anything else, you'll need headphones or remote speakers.

The 1080p webcam also belongs in the category of better than nothing, but only just. The images it produces are drab and extremely fuzzy, so fuzzy in fact that we at first assumed the lens was smeared in something and broke out the cotton wool buds to clean it. Alas, that was to no avail; the webcam is just plain bad.

For the record, the camera doesn't support Windows Hello facial recognition, and it also lacks the Windows Studio suite of filters and augmentations, though that's more down to the absence of an NPU on the chipset to support Windows local "AI" functions than the camera hardware.

Medion Signium 27 S1: Specs and Performance

Given the cost, the internal specification of the Signium 27 is pretty solid with a 10-core Intel Core 7 240H CPU, 32GB of dual-channel DDR4 RAM and a 1TB SSD. There's no way of accessing the internal parts of the system, so there's no way to add more storage or RAM.

The 240H processor inside the Signium is a Raptor Lake Refresh laptop chip with a 5.2GHz turbo boost speed, 6 performance cores, and Intel's basic integrated UHD Graphics GPU.

In our bespoke 4K Handbrake-based multi-media benchmark, the Signium scored 221 points. That's quite a bit behind the MSI Pro AP242P's 329, though, given that it runs a 20-core Core i7-14700, it is hardly a surprise.

The Medion's performance is more in line with what you'd expect from the likes of a 2023-era M2 MacBook Air (220) and not too far adrift of more modern Windows laptops like the Geekom X14 Pro (267) or the Lenovo ThinkPad X1 Carbon (245).

In simple CPU performance, the Signium performed very similarly to the Acer TravelMate P6 we tested recently, with its Intel Core Ultra 7 258V chipset. The Medion scored 2,605 and 10,341 in the GeekBench 6 single and multi-core tests compared to 2,483 and 10,021 for the Acer laptop.

Turning to graphics performance, the Medion did rather better than expected Intel UHD Graphics 770 iGPU, which boasts 64 execution units, double that found in the MSI Pro AP242P's UHD Graphics 770 GPU chip.

The SPECviewperf 3dsmax 3D modelling benchmark scored 15fps on the Medion. That compares to 12.5fps on the MSI. Naturally, you'll get much better results from the latest generation of Arc integrated GPUs (the Intel Arc B390 can run the same test at 35fps, the Arc 140V at 25fps), but the Medion has nothing to be ashamed of.

The Signium, then, is no powerhouse, but performance across the board is more than adequate for the sort of general computing tasks you'd buy a machine like this for.

When placed under maximum stress, the single cooling fan was easily able to prevent any thermal throttling and allowed both the CPU and GPU to run at 100% utilisation for prolonged periods. The fan was also impressively quiet, even when running at full speed.

The PCIe 4 M.2 SSD (and the RAM sticks) in our review sample are made by a company called Yclongsto. That's a new name to us, but the drive performed well enough, returning consecutive read and write speeds of 3,855MB/s and 3,027MB/s which is more than adequate for basic productivity tasks.

The Signium ships with Windows 11 Home rather than Pro, but if you fancy ditching Windows altogether, then you're in luck because Ubuntu 24.04 runs perfectly with everything, other than the pointless 75Hz/720p overdrive, working as it should right out of the box.

Medion Signium 27 S1: Is it worth it?

For the asking price, the Signium 27 S1 is a very attractive package. You certainly get a lot for your money, and once set up, you won't need to buy any accessories, even if the webcam and speakers leave rather a lot to be desired in terms of quality.

Performance levels are more than adequate, the display is rather more colorful than those you'll get with some of the more expensive competition, and you can use it as a monitor thanks to that HDMI input.

The fact that it comes with a 1TB SSD and 32GB of RAM as standard is another bonus, as is the ease with which you can add even more RAM and storage.

The Signium also looks rather less utilitarian than much of the competition, which could be a valuable feature if you want your office to look just a little less workaday.

Medion Signium 27 S1 Specifications

Acemagic's emergence as a solid mini PC manufacturer is underlined by a handful of solid machines in recent months, including the decent budget-friendly Vista V1 Mini PC and the higher-end Matrix Mini M1. Although they're both at opposite ends of the pricing and performance spectrum, the Chinese-made mini PCs share a few things in common, including good value for money as well as a sturdy, portable, and attractive chassis.

The Kron Mini K1 is a new model that is nearly identical to the Matrix Mini M1 but fitted with the four-year-old, mid-range Intel Core i5-12600H processor (or the three-year-old AMD Ryzen 5-7430U). However, at a steep retail price of £559 for the base model with 16GB RAM – or $559 in the US, reduced to $359 – this cosy desktop companion will need to do a heck of a lot of work to justify its worth considering over alternatives with newer hardware.

Acemagic Kron Mini K1 review: Design

We love the overall design aesthetic of Acemagic mini PCs; this is more or less consistent across all its models. The rounded corners contribute to a simple yet refreshing design, almost resembling the stunning Mac Mini. It's identical in every way to the Matrix Mini M1 – apart from a much more appealing silver finish versus the oppressively corporate gray.

Certainly, it'll fit into your home office or desk setup without standing out. Plenty of mini PC makers opt for bolder and louder design choices, like aggressive grills, hardened edges, and metallic flourishes – but the Kron Mini K1's stripped-back design choices neatly reflect its simplicity as a plug-and-play machine. This may not necessarily be the sort of mini PC to suit the personality of a power user, but it doesn't look too out of place on a tidy desk. You can also mount it onto a compatible monitor with brackets provided in the box.

Its similarities to the Matrix M1 Mini include its 128 x 128 x 41mm dimensions and a weight that's just 5g lighter (515g). Although Geekom's units are a tad deeper (49mm), they tend to be a little smaller, measuring 117 x 112mm wide and deep. They're also roughly 130g heavier at 652g. While there's no beating the Vista Mini M1 for portability – with this entry-level machine measuring 100 x 100 x 32g and weighing more than half (245g) of the Kron Mini K1 – all these mini PCs are plenty compact and light enough for daily transport.

On the front, you'll find the sole USB-C 3.2 Gen 2 port alongside two USB-A 3.2 Gen 2 ports for company, as well as a 3.5mm headphone jack. On the rear, however, you'll benefit from a plethora of additional ports, including four USB-A 3.2 Gen 1 ports, an HDMI 2.0 port, a DisplayPort 1.4b, an Ethernet port, and a DC input.

While we question the need for six USB-A ports over one USB-C port, they're arranged smartly overall. We would also bemoan the lack of an SD card reader, meaning you'll likely need a dock or converter if you need to transfer media files. There's only support available for Wi-Fi 6, rather than Wi-Fi 6E (which is found in most new mini PCs), as well as Bluetooth 5.2 – a couple of generations below the up-to-date 5.4 standard

Acemagic Kron Mini K1 review: Specs and performance

Despite a pretty lofty price tag for a mini PC, the Kron Mini K1 doesn't include the very latest hardware. Alongside 16GB DDR4 RAM, integrated Intel UHD Graphics, and a 512GB SSD, it features the 12-core Intel Core i5-12600H processor. This is an Alder Lake processor first released in January 2022, and was originally optimized for high-performance graphics at a reasonable budget. By today's standards, however, it's well behind the curve.

In our performance testing with Geekbench 6, it registered a single-threaded score of 2,233. This, admittedly, is not a bad result. It's slightly below average when compared with modern CPUs and enough to handle simple, day-to-day tasks. The multi-core performance, however, was abysmal compared with the 2026 norm, registering 5,508. By way of comparison, the Acemagic Matrix Mini M1 features the Intel Core i9-13900HK (14-core) and registered 2,806 and 8,912 for each result, respectively – and it retails at the same price.

Its GPU was also underwhelming, registering 9,630 in our OpenCL test. This isn't as bad a result, considering it beats the Lenovo ThinkCentre M70q's result of 7,995 (also featuring integrated Intel UHD Graphics). The Matrix Mini M1, however, scored a much healthier 13,514 with its Intel Iris Xe Graphics chip. Overall, it's a middling result – but perhaps more in line with the price you pay for this device. The SSD speeds, meanwhile, are slightly below average for mini PCs, with reads of 3,368MB/s and writes of 2,538MB/s.

Acemagic Kron Mini K1 review: Is it worth it?

We simply don't believe the Kron Mini K1 represents value for money, considering its surprisingly high retail price point. At £559, the device is approximately £150 more expensive than the Acemagic Matrix Mini M1, which is better in every respect, apart from its less appealing gray finish.

The device itself looks nice and is incredibly portable. Beyond that, its connectivity options are broadly useful – if you could forgive Wi-Fi 6 instead of 6E and Bluetooth 5.2 versus 5.4. But these qualities are sadly not enough to make up for the out-of-date hardware fitted within the body. If you're looking for something that can handle multiple workloads at once, it won't meet the standard you'd expect from modern devices based on our benchmarking tests. That said, it's more than enough for a mini PC that you could deploy for simple, day-to-day tasks – but we can't justify recommending this at the current retail price tag. If you find the Kron Mini K1 available at a discount, then it might be worth considering for an entry-level device. But that discount would have to be significant.

Acemagic Kron Mini K1 review specifications

Many conventional notebooks and ultraportables are being branded as AI PCs, despite their powers being limited by an integrated NPU that limits them to built-in Windows 11 features and OEM bloatware. The Nvidia DGX Spark is a different beast entirely. This specialized miniature workstation, powered by the GB10 Grace Blackwell Superchip, can tap into supercomputing-level performance.

Unlike conventional desktop PCs and mini PCs, this Linux-operated machine is a new class of machine designed for power users, AI developers, and software engineers, packing in a host of premium Nvidia components to deliver enterprise-grade performance to users who want to build and run their own models and workloads locally.

With Nvidia partnering with a host of OEMs for manufacturing and release, there are plenty of variants that you can choose from, including the excellent Asus Ascent GX10, Lenovo's ThinkStation PGX, and the Dell Pro Max with GB10, among others. These machines all adhere to the same dimensions and specs – but there will be slight differences in the build quality, overall design, thermal management, and pricing. Presently, we've got our hands on the Acer Veriton GN100 – a potent-looking machine that's priced at £3,999.99 in the UK.

Acer Veriton GN100 review: Build and design

With clearly defined dimensions (150 x 150 x 51mm), OEMs will chiefly lean on the design of these machines to help their own stand out. Acer's design is incredibly sharp, with a set of vertical metallic grills cut with a horizontal bar, and beneath it, a protruding accent line running from left to right and dipping in the middle to give way for the logo.

The machine itself looks great on your desk and evokes an air of sophistication. For what it's worth, it's also one of the lightest DGX Spark iterations, weighing 1.2kg versus the Asus' 1.4kg. It's certainly a far cry from the machine this was based on – the 2016 DGX1, which cost $130,000 and weighed 61kg.

On the rear, you'll find the standard set of ports fitted into this class of mini workstation, arranged in a standard configuration beneath a stack of ventilation grills flush against the device's metallic surface. The ports include three USB-C 3.2 Gen 2x2 (20Gbps) ports with DisplayPort 2.1 compatibility, one USB-C Gen 2x2 with PD in, an HDMI 2.1 port, an Ethernet port, and a specialist Nvidia ConnectX-7 SmartNIC port that's used to link two of these mini workstations together for double the potency. You'll also benefit from wireless connectivity in the form of Wi-Fi 7 and Bluetooth 5.4 courtesy of a MediaTek card. We do lament the lack of additional ports like USB-A or an SD Card reader, but you can certainly connect peripherals to this machine with a dock or through a monitor.

Acer Veriton GN100 review: Specs and performance

Powering the Veriton GN100 is Nvidia's monstrous GB10 Superchip, which combines a 20-core Arm processor (10 Cortex-X925 performance cores and 10 Cortex-A725 efficiency cores) with a Blackwell GB20B GPU. This is a 5nm graphics card designed for specialized or low-power AI applications. Also included is 128GB LPDDR5x unified memory and a 4TB PCIe Gen5 NVMe SSD that offers the memory capabilities to support large language models and generative AI workloads. This is the same set of components you'll get in all Spark DGX machines, regardless of the OEM Nvidia has partnered with.

The machine is capable of reaching a petaFLOP of AI performance with FP4 precision (the 4-bit floating point numerical format for accelerated training and inference). This drops to 170 teraFLOPS with FP16 precision – the conventional standard for AI training and inference. Incidentally, these kinds of workloads typically require 37GB of memory bandwidth, which would more than blow out consumer-grade GPUs and render local AI practically impossible on standard machines. Indeed, this device presents a far higher ceiling for AI workloads than any conventional desktop PC or enterprise-grade workstation.

The machine also draws a maximum of 240W of power, but you're more likely to draw far less than that most of the time, unless you really push it to its upper performance limits.

Standard PC benchmarks are fairly irrelevant, offering a dim picture of what the machine is capable of. But, for what it's worth, testing via Geekbench 6 generated a staggeringly significant result of 3,106 for single-core performance and 19,192 for multi-core performance – roughly in line with the 3,104 and 20,048 we registered with the Asus Ascent GX10.

Anecdotal testing showed how you can run generative AI workloads, including image generation via ComfyUI and others, within minutes. Throughout usage, the machine barely made a sound and ran at a comfortable temperature – incredibly impressive for a device that's as capable and as compact as this is.

Acer Veriton GN100 review: Features

As you would expect with all variations of Nvidia's mini AI workstations, the Acer Veriton GN100 runs on the DGX OS platform, an Ubuntu 24.04 LTS (for ARM) Linux-based system. Compared with the vanilla version of Ubuntu, Nvidia has customized the overall look and feel to suit its unique design, color, and font scheme – and has packed the machine with plenty of out-of-the-box apps and services to get you started in running AI tasks.

From the desktop, you can easily seek out resources and links to Nvidia's library of AI 'playbooks', which are step-by-step guides that walk users through the purpose, setup, and best practices of running their own AI workloads. While a certain degree of technical knowledge is essential, these are fairly straightforward to follow – even if you've never used a Linux machine before – and make AI far more friendly and accessible to workers who may have never interfaced with custom-built generative AI before.

Ultimately, however, a machine like this is designed for power users who are keen to experiment with AI trials and deployments within their organizations, run local generative AI-based tasks without external connections, and create AI agents (and so many other uses). There are plenty of applications, from image generation and media manipulation to running VS Code or even building multi-agent chatbots. The playbooks are essential to the democratization of AI within the workplace, and really do help bring all workers up to speed.

The new Nvidia Sync feature also helps you connect your local desktop or laptop with the DGX Spark so you can operate these machines remotely within the same subnet – or via an IP address. You can launch applications remotely and share the display, while also accessing monitoring tools and other applications.

Acer Veriton GN100 review: Is it worth it?

This new class of AI-powered mini workstations is a game-changer for AI deployment within the workplace. In that spirit, the Nvidia-powered Acer Veriton GN100 is an exemplary machine that shows just how quick and easy it is to get started on running and experimenting with AI workloads.

We should note that this is not a machine for the conventional user, with Nvidia stressing that there's really no point in picking one up unless you plan on tapping into enterprise-grade performance levels courtesy of the GB10 Superchip. In many ways, this mini workstation will feel inaccessible or frustrating to those who predominantly work with conventional applications in Windows environments or have no need to develop and deploy AI tools.

For approximately £4,000, it's slightly cheaper than speccing out larger desktops like the Dell Pro Max Tower T2. That said, heads may be turned with the AMD Ryzen AI Max+ 395-based systems (codename Strix Halo), which run Windows 11 (a plus point for non-power users keen to learn how to use AI) and are available for a little less. However, the Acer Veriton GN100 is still a great system that serves as a fantastic entry point into the world of localized AI deployment.

Acer Veriton GN100 specifications

The Dell Tower Plus is a desktop PC that will be whatever you want it to be. That's because there are nine distinct configurations you can opt for – each appealing to a different kind of end user, from day-to-day office bees to power users. As with most Dell products, you can also fully customize your own version of this unit.

Our review unit was configured to represent the most powerful version, with this Dell Tower Plus featuring among the best components out there right now, including an Arrow Lake Intel Core Ultra 9 200 series processor and a 5000 series Nvidia graphics card. Although the simplest configuration starts at £999, our specific combination is available at the sizeable retail price of £3,369.

This model is a direct replacement for the Dell XPS Desktop, which the company last manufactured in 2024. In truth, it's largely the same machine with a lick of paint and a hardware refresh. Although the XPS branding is returning, it's only for a new line of business-oriented laptops – with the Dell Tower Plus still the go-to desktop PC for users looking for a reliable desktop PC with the best specs.

Dell Tower Plus EBT2250: Design

The Dell Tower Plus is pretty much identical to the 2024 Dell XPS Desktop in its dimensions and weight, with a 372.9 x 173 x 426.9 mm chassis equating with an approximate volume of 27.5 litres. The weight also ranges between 8.62kg and 12.94kg, according to Dell, based on your configuration and manufacturing variability.

The one difference between the Dell Tower Plus and the Dell XPS Desktop is a slightly lower maximum weight, with the latter reaching up to 13.34kg. Overall, it's relatively convenient – and much smaller than, say, the Dell Pro Max Tower T2 – which occupies a 32L chassis size and boasts a huge 23kg weight.

Aesthetically, the Tower Plus desktop PC is very simple in nature, which may be favourable to most users. A ventilation grill occupies the lower half of the front of the chassis, with a clean space above and ports lining the right-hand side. You could have easily fit a disc drive in the space here, but few Dell desktop PCs now include one anyway – and you'll likely need a separate plug-in expansion unit.

The only finish this machine is available in is graphite, which is a little bland and uninspiring. This is fairly conventional, however, and won't stand out in a busy office environment – if blending in the ideal scenario. We'd have liked the option of a far fresher silver chassis coating, as was available with the old XPS desktop, but it's far from a dealbreaker.

Dell Tower Plus EBT2250: Connectivity and Features

Unlike the Pro Max Tower T2, there's very little room for customizing the ports available to you. But why would you need that? There's almost everything you could ask for already included with the Dell Tower Plus.

On the front, you'll find three USB-A 3.2 Gen 1 ports (5Gbps) and a USB-C 3.2 Gen 2 port (10Gbps), alongside an SD card slot and a 3.5mm audio jack, all lined up in a neat row an inch or two beneath the power button. These 'essentials' are useful for quick access.

The party is really happening at the rear, however, with an additional four USB-A 2.0 ports (480Mbps) – mostly useful for peripherals – as well as two additional USB-A 3.2 Gen 1 (5Gbps) ports for even more connectivity. You'll find this alongside a USB-C port with Thunderbolt 4 (40Gbps) and an HDMI port. You also have access to an RJ45 Ethernet port (2.5GbE) as well as A/V options in the form of an S/PDIF Optical port to transfer digital audio signals to a soundbar or receiver and 2x Sterio Audio Line in/Lind Out for stereo sound. There's also a Kensington Lock Slot for additional peace of mind.

In terms of wireless connectivity, the Intel Wi-Fi 7 BE200 (2x2) 802.11be chipset grants access to the latest Wi-Fi standards, with support for Bluetooth 5.4 also available.

Our configuration ditched one of the expansion bays in favour of an additional HDMI port and three DisplayPort slots. In total, there are several expansion slots available. There's space for two M.2 SSDs, two 3.5in HDD bays, and three PCIe expansion slots of x4, x4, and x16. You can also expand your RAM with two slots supporting up to 64GB DDR5 memory.

Dell Tower Plus EBT2250: Performance

The Arrow Lake-based Intel Core Ultra 9 285K is the beating heart of the highest-end Dell Tower Plus configuration – with 24 cores and a 3.7GHz clockspeed with turbo up to 5.7GHz. The other options available include Intel Core Ultra 5 225, Intel Core Ultra 7 265, Intel Core Ultra 7 265K, and Intel Core Ultra 285 processors.

Our graphics card was the high-end Nvidia GeForce RTX 5080, with 16GB GDDR7 memory – although you can also pick up this machine with the Nvidia GeForce RTX 5050, Nvidia GeForce RTX 5060, Nvidia GeForce RTX 5060 Ti, Nvidia GeForce RTX 5070, Nvidia GeForce RTX 5070 Ti GPUs, or integrated Intel UHD Graphics.

These components were included alongside 32GB DDR5 RAM and a 2TB SSD. There is as little as a 512GB SSD and up to 4TB SSD storage available in the most expensive configuration, and between 16GB DDR5 and 64GB DDR5 memory. The price of the machine, of course, varies wildly depending on which CPU, GPU, and memory options you select, with your choice of graphics card offering the biggest swings.

The performance in our configuration was exceptional, as determined through testing with Geekbench 6, with a single-core score of 2,914 and a multi-core score of 19,697. However, it's worth mentioning that it's a little less than we registered with the same CPU in the Dell Pro Max Tower T2 – with scores of 3,202 and 22,583. That's likely due to the higher power and cooling proficiency in this desktop. That said, it's still a result that trounces many other modern CPUs – and you won't find similar performance in a laptop. You are, however, able to find a handful of AMD processors that eclipse its performance, especially in the AMD Ryzen Threadripper PRO series.

The Nvidia GeForce RTX 5080 GPU is everything you could ask for, with our OpenCL benchmarking producing a brilliant result of 272,793 – putting it near the top for current desktop-grade GPUs and about 1.5 times more powerful than the same GPU you'll find in laptops. For performances much better than this, you're effectively looking at finding a workstation or desktop PC with workstation-grade performance.

There's nothing especially standout about its SSD performance, with great reads and writes of 7,044MB/s and 6,130MB/s respectively. They are, however, a lot slower than the extraordinary performance of the Dell Pro Max Tower T2's 12,008MB/s and 12,674MB/s.

Dell Tower Plus EBT2250: Is it worth it?

There's something for almost everybody in the Dell Tower Plus. But we should first highlight the elements that are consistent across all versions, including its excellent 27.5L size and under-13kg weight, alongside a great build and clean, minimalist design. Apart from the lack of a CD/DVD drive, there are plenty of ports that you can take advantage of, too.

Whether it represents value for money, however, comes down to the variations and which end of the spectrum you're coming from. For entry-level users needing a basic desktop PC that can handle the lightest, day-to-day tasks, it might be worth considering the Dell Tower Desktop – available from £599. For power users hoping for workstation-grade performance, it's worth considering an upgrade to a machine like the Dell Pro Max Tower T2 desktop PC, which offers much more juice. You can also choose from a far broader range of components to tailor your machine to the specific workloads you want to run, rather than stock options.

Our configuration was a little expensive for what it was, but not overpriced to the point we wouldn't recommend it – you're just more likely to get the best value if you're a mid-range user. In isolation, however, it's a brilliant desktop PC with a clean design and a small, lightweight frame that will please almost all users.

Dell Tower Plus EBT2250 specifications

The Geekom A5 Pro is, in simple terms, a cheaper and pared-down version of the A8. By buying the A5 Pro, you will save around £150, depending on the offers Geekom has running on the two machines at the time.

At the time of writing, you can pick the A5 Pro up from Amazon for £424 and the A8 for £568, a saving of £144. You can also pick it up directly from Geekom in the UK and the USA, where the code ITPROA5PRO should get you a 7% discount. The Amazon USA sales page is here.

Naturally, that saving means you are going to lose out on some of the A8's features, so you have to make do with a 6-core 4.5Ghz AMD Ryzen 5 7530U CPU and basic Radeon graphics rather than an 8-core 4.9GHz Ryzen 7 8745HS CPU and Radeon 780M graphics.

You'll also miss out on support for 6GHz Wi-Fi; the A5 Pro's MediaTek RTL8852 wireless card only supports 2.4/5GHz signals. And the main SSD interface is PCIe 3.0 rather than 4.0, as in the A8. Other than that, the two machines are so difficult to tell apart that you'll be forced to flip them over and see what's written on the underside.

Geekom A5 Pro: Design

Externally, the A5 Pro is a dead ringer for the A8. They are the same size, same weight, give or take a few grammes, made from the same materials, and have the same physical array of ports. That means you are getting a smart, solid, and compact little metal box that's easy to hide away or hook behind a monitor using the bundled VESA bracket. Other than said bracket, the only other accessory in the box is an HDMI cable.

Mirroring the A8, the front panel of the A5 Pro is home to the on/off button, two 10Gbps USB-A ports, and a 3.5mm audio jack. On the right side, you'll find an SD card slot, while at the rear, there are two more USB-A ports, one 10Gbps, one 480Mbps, two 10Gbps Type-C ports, two HDMI 2.0 video outputs, and a 2.5GbE RJ45 jack.

While neither Type-C port is USB4-Compliant, both can output DisplayPort video. Neither can be used to power the system, though. For that, you must rely on the 65W rat-and-tail DC power supply. There's nothing in the way of a Kensington lock, which is a bit of an oversight on a device that costs more than £400, is not much larger than a packer of cigarettes, and only weighs 375g.

Getting inside the A5 Pro involves prising the four rubber feet off the base, removing the four Philips screws beneath, removing the plastic base, and then removing the four more Philips screws and the metal shield plate.

There is a single antenna cable connecting the wireless card to the plastic base, but it's long enough to give full access without having to disconnect it, unlike the A9 Max, A8 and A7 Max. Having to reconnect the wireless antennas in all three of those machines is a major pain in the backside, as we can testify firsthand.

Once you've removed the metal shield, you'll find two SODIMM mounts and two SSD mounts, one occupied 2280 PCIe 3.0 (no surprise given that this is the best the chipset supports) and one unoccupied 2242 SATA3.

It's worth recalling here that the A8 doesn't have a second SSD slot, so when it comes to storing shedloads of data, the cheaper A5 Pro has a distinct advantage. The 1TB Wodposit SSD in our review unit delivered decent sequential read and write speeds of 3,131MB/s and 2,983MB/s, respectively. For a general-purpose box like the A5 Pro, that's good enough, but the A8 can do significantly better here and in wireless data speeds thanks to its support for 6GHz Wi-Fi 6E.

Geekom A5 Pro: Specs and Performance

There's only one version of the A5 Pro available, and it's built around the 6-core, 12-thread AMD Ryzen 5 7530U CPU, a low-power processor launched back in early 2023. Graphics are provided by a Radeon (originally called the Radeon RX Vega 7) integrated GPU, a component with performance levels broadly similar to Intel's basic mobile Integrated Graphics solution.

The A5 Pro ships with two 8GB RAM sticks, a preferred option compared to the single 16GB stick you get with the A7 Max. Granted, it means you end up with a spare stick if you want to add more, but in our view, that's outweighed by the benefit of getting the performance boost from dual-channel RAM in the first place, especially with after-market RAM costing what it does at the moment.

Even a casual glance at the specification sheet shows that the AMD chipset in the A5 Pro is no stump-pulling powerhouse, and testing bore this out. In the GeekBench 6 CPU test, it scored 1,977 and 6,859 in the single and multi-core test, while in ITPro's bespoke Handbrake-based 4K multi-media test, it scored 178.

We've got a 2020 vintage M1 MacBook Air lying about the office that returns very similar scores. For a more relevant comparison, the A8 scored 2,568 and 12,842 in the GeekBench test and 363 in the 4K multimedia. Clearly, there is a world of difference in CPU performance between the two Geekom boxes, even when allowance is made for the fact that the A8 had 32GB of RAM in it.

The same is true when it comes to graphics. The new A5 Pro scored 15,059 in the GeekBench 6 OpenCL test to the A8's 30,455, and in a more practical test, the A5 Pro ran the SPECviewperf 3dsmax 3D modelling benchmark at 15fps, to the A8's 40fps.

It's fair to say that in all meaningful ways, the A8 is twice as powerful as the A5 Pro. If all you need is a box for light productivity and media serving, the A5 Pro is a good alternative, but if you want something for demanding work, intense graphics jobs, and light gaming, the A8 is most certainly worth the extra money.

Neither the A8 nor the A5 Pro has anything in the way of an NPU, the former because the chipset was originally a China-only component and they have rules about such things, the latter because its chipset predates the AI fad. This means there's no support for Windows' CoPilot AI features like Recall or generative fill in Paint.

If you want those features, you'll need to look at something like the Geekom A9 Max or the MSI Cubi NUC AI+ 2MG, both of which are built around more modern chipsets with 50 TOPS (well, 47 in the case of the Intel-powered MSI) NPUs.

The low-power chipset doesn't cause any problems for the A5 Pro's single fan. Even under maximum stress, both the CPU and GPU run continuously at 100% utilization, with the fan making little more than a low whisper.

The A5 Pro is only available with Windows 11 Pro preloaded. Still, Geekom advises that it will work with Ubuntu and Fedora, and it proved true: both Ubuntu 24.04 LTS and Fedora 43 ran perfectly out of the box.

Geekom A5 Pro: Is it worth it?

At the risk of stating the blindingly obvious, whether you should choose the A5 Pro or A8 AI depends entirely on what you plan on using your mini PC for.

The A5 Pro's natural habitat is basic productivity and media serving, jobs that won't tax the rather gutless chipset but will take full advantage of the facility to add a second, albeit SATA3, SSD.

For anything more demanding, especially on the graphics front, the A8 is a better choice. If you want a box to run programmes like Adobe Photoshop or other GPU-intensive apps, the Radeon 780M is by far the better option.

Geekom A5 Pro Specifications

Intel is about as ubiquitous as you can get in the enterprise technology landscape. Chances are, everybody on this planet who's interfaced with a computer has used the company's chips. That's a degree of universality enjoyed by only a handful of organizations throughout history. Yet, until very recently, the narrative around a once irreproachable institution has been dripping in corporate angst, with anxiety rising around a perceived decline.

Over the last few years, Intel's rivals have made up ground. We needn't dwell too much on Nvidia, which enjoys a de facto stranglehold on the US economy. Qualcomm is in a period of ascendency, pushing its chips into new kinds of devices like laptops. AMD is also enjoying something of a renaissance period, thanks in part to its data center dominance. Samsung has also been an ever-present in the chipmaking game.

Intel, however, has been bidding to "turn things around" for the best part of a decade. Plans have largely centered on manufacturing, including being a producer for "fabless" companies and building additional factories. Recent efforts include job cuts and reorganizations. Under fresh leadership in the form of Lip-Bu Tan, these efforts are now beginning to pay dividends, especially over the last six months. With graphics processing units (GPUs) now the hottest commodity in the AI build-out, can a pivot into this space be Intel's solution to finally exorcise its demons?

Intel's rise, fall and signs of resurgence – how did we get here?

Despite a positive history, events since the turn of the millennium have contributed to a perception that Intel has struggled to adapt to the modern world. To some extent, these views are grounded in reality – with material consequences to the company's bottom line. A sense of complacency and poor business decisions that have compounded over time, the narrative goes, has opened the door wide open to its much hungrier rivals, while executives have unfortunately been caught resting on their laurels.

"Turbulent is very appropriate," Alvin Nguyen, senior analyst at Forrester, tells ITPro, when asked how he would characterize Intel's performance as a company in the last few years. "I would say that until the past year, Intel had been struggling in terms of perception, thought leadership, and reputation."

It's easy to see why this is the dominant narrative, and the collapse in the company's stock price between 2021 and 2025 – a nearly four-times drop – illustrates how the market has perceived the company's performance. But how did we get here? One of the earliest missteps could perhaps be from the early 2000s, in a story now mythologized, that former CEO Paul Otellini turned down a deal with Apple to make iPhone processors.

You may also point to another former CEO Brian Krzanich, who in the mid-2010s held back on deploying next-generation equipment used in the extreme ultraviolet lithography (EUV) chipmaking technique in favor of multi-patterning technology. Commentators say this set the company back by half a decade.

Despite efforts to turn things around in the years that followed, including a $20 billion plan in 2021 to develop a foundry service, things did not improve, and the company continued its slide. However, following the retirement of Pat Gelsinger and appointment of its new CEO in March 2025, the company has undergone a significant internal transformation.

The firm made key executive hires in June last year, before axing 24,000 roles and cancelling factory plans the following month. Now, reports suggest that Intel will aim to buy into the success that some of its key rivals have lorded over it for years by investing in GPU production for AI data centers.

Can Intel tap into the GPU goldmine?

There's no overstating how important GPUs are in the AI buildout. The parallel processing capabilities of these chips, once used primarily in PC gaming, are primed to perform the sorts of operations and calculations essential to AI training and inference. Enterprise-grade GPUs are also highly sought after, with the likes of AMD and Nvidia dominating this space. So does the market need another player?

"Nvidia may be the leader, but they are unable to satisfy the market demand by themselves," says Nguyen. "As long as the demand for AI (and GPUs) keeps beyond capacity, Intel is not entering a "crowded" market. If there is a market correction or multiple market corrections, then this changes."

For what it's worth, Intel has been quietly improving the standard of its consumer-grade GPUs for years. The first generation of Intel Xe integrated graphics processors paved the way for a set of excellent entry-level GPUs in recent years in the form of the Alchemist and Battlemage Intel Arc Graphics GPUs you find in some of the best business notebooks.

The next generation of chips will be launched in the coming year or two under the Celestial banner, with Xe-3 integrated graphics serving entry-level machines. Nvidia, meanwhile, has been planning a scaleback in its consumer-oriented GPUs. Although the margins are less profitable than enterprise customers, is this something of a void that Intel can step into?

The next generation of chips will be launched in the coming year or two under the Celestial banner, with Xe-3 integrated graphics serving entry-level machines. Nvidia, meanwhile, has been planning a scaleback in its consumer-oriented GPUs. Although the margins are less profitable than enterprise customers, is this something of a void that Intel can step into?

"Intel is addressing both: AI-related enterprise demand provides an untapped revenue source, but the demands of AI are pulling in resources from other markets like consumer, that Intel also has an opportunity in the consumer space which they are targeting with consumer GPUs," says Nguyen.

The opportunity is certainly there. But, ultimately, it comes down to the execution. The early signs suggest that Tan is the man with the plan, with Nguyen telling ITPro that "Intel simply needs to keep doing what they are doing".

"They have focused on changing their corporate culture, became leaner, and became more customer focused," he explains. "They have had some clear wins in terms of their CPUs (specifically laptop CPUs) and announced big partnerships (US Government, Nvidia) that have changed perception about them [positively].

"Shipments and revenue were not as bad as the previous perception painted, so focusing on putting out products customers want, addressing issues instead of letting them stymie them for years, and making high-visibility partnerships that provide positive benefits will help grow positive perception."

Intel is on the right track, as far as Nguyen is concerned. Under Tan's leadership, the firm’s share price has surged from under $19 per share to highs of over $45. Although its previous turnaround efforts have been ineffective, seizing the opportunity to build industrial-grade GPUs might just be the boon that Intel needs to prove that its woes are firmly in the past and it can go toe-to-toe with the new kids on the block.

Microsoft reckons it's made a major breakthrough in glass data storage technology, touting a new technique it says can preserve information for 10,000 years.

Magnetic tape and hard drives degrade within decades, leading to research into glass - a permanent data storage material that is resistant to water, heat, and dust.

In the past, glass-based data storage has required expensive fused silica - a type of glass that is relatively difficult to manufacture and available from only a few sources.

However, the new Project Silica technology - detailed in an article in Nature - means it's possible to use the same borosilicate glass found in kitchen cookware and oven doors.

"This advance addresses key barriers to commercialization: cost and availability of storage media," said Microsoft partner research manager Richard Black.

"We have unlocked the science for parallel high-speed writing and developed a technique to permit accelerated aging tests on the written glass, suggesting that the data should remain intact for at least 10,000 years."

Microsoft’s new glass data storage technology

The new technique allows hundreds of layers of data to be stored in glass just 2mm thick, as with previous methods. However, the reader for the glass now needs only one camera, not three or four, thereby reducing cost and size.

The writing devices also require fewer parts, making them easier to manufacture and calibrate and enabling them to encode data more quickly.

Femtosecond lasers, which fire high-power laser pulses quadrillionths of a second long, are focused on a point within the glass to modify its optical properties and encode data as voxels, the 3D equivalent of pixels.

Microsoft said it's now able to cut the number of pulses used to form the voxel from many to only two. Crucially, the polarization of the first pulse is not important to the polarization of the voxel formed.

Notably, the tech giant has been able to build on this to enable pseudo-single-pulse writing. This involves a single pulse that can be split after its polarization is set to simultaneously form the first pulse for one voxel - where the polarization doesn’t matter - and the second pulse of another, where the set polarization is essential.

New storage methods could be coming soon

As part of the project, Microsoft revealed it has invented a new type of data storage in glass called “phase voxels”, in which the phase change of the glass is modified instead of its polarization, showing that only a single pulse is necessary to make a phase voxel.

"We demonstrated that these phase voxels can also be formed in borosilicate glass and devised a technique to read the phase information from phase voxels encoded in this material,” Black said.

“We showed that the much higher levels of three-dimensional inter-symbol interference in phase voxels can be mitigated with a machine learning classification model."

Meanwhile, by using a mathematical model of pre-heating and post-heating within the glass with a new multi-beam delivery system, the team found that many data voxels can be written in proximity in the glass at the same time, significantly increasing writing speed.

The team has also developed a new way to optimize symbol encodings using machine learning, along with a better way to understand the trade-off between error rates, error protection, and error recovery when evaluating new digital storage systems.

Crucially, a new non-destructive optical method identifies the aging of data storage voxels within the glass, finding it can support data lasting 10,000 years.

The company demonstrated these techniques with several proofs of concept - including storing Warner Bros' Superman” movie on quartz glass, and partnering with Global Music Vault to preserve music under ice for 10,000 years.

"The research phase is now complete, and we are continuing to consider learnings from Project Silica as we explore the ongoing need for sustainable, long-term preservation of digital information," said Black.

FOLLOW US ON SOCIAL MEDIA

Amid the rise of hybrid work, bring-your-own-device (BYOD) policies, as well as increasingly portable devices like mini PCs and ultraportables, the thought of integrating a tower-style desktop into the workplace (or your home office setup, for that matter) may feel redundant. But don't underestimate the mighty tower PC; when intensive tasks demand more power and running AI workloads at full pelt is increasingly common, there is no point in compromising.

Designed for professionals who crave reliable high-end computing and flexibility, Dell's Pro Max Tower T2 is designed for those who need a machine that can handle virtualization, visual design, content production, and simulation, among other use cases, thanks to an NVIDIA RTX PRO 6000 Blackwell Workstation Edition. It's also engineered to handle and optimize AI-driven workflows, with an Intel Core Ultra 9 285K processor powering it.

It's worth noting this configuration is priced north of $10,000, but you can pick up a base machine for much less – starting at closer to $1,000, with ideal configurations for most users priced at between $1,500 and $3,000 depending on the components and optional features.

Dell Pro Max Tower T2: Design

Dell adopts a familiar 32L chassis size for its Pro Max Tower T2 desktop (that's a casing with an internal volume of 32 liters), meaning it makes for a fair balance between larger desktop PCs that could top 40L sizes and more compact sub-20L desktops. Overall, this design doesn't overburden your desk space while giving you plenty of airflow and access to hardware within.

The tower comes shipped with a full-length grill that you can clip onto the front of the machine to aid with ventilation. This is especially crucial in managing heat in high-wattage configurations, especially if you have a machine with large-core CPUs or high-end graphics cards, like our review unit. Aesthetically, this massive grill also adds some character to an otherwise bland black box. It's also easy to get into the box, thankfully, with a side latch that means you won't need to tear down the machine to get inside – either to maintain the components or upgrade any over time.

Dell Pro Max Tower T2: Connectivity and Features

Atop the front-facing side of the Pro Max Tower T2, there's room for two optional features that could prove useful if handling legacy media: an optical disk drive and a 3.5-inch lockable storage bay. Beneath that, you'll find the power button followed by an LED status indicator, a 3.5mm headphone jack, two USB-A 3.2 Gen 1 ports (the lower with PowerShare), as well as USB-C 3.2 Gen 2 and USB-C 3.2 Gen 2x2 (with PowerShare) ports.

The rear side depends on which optional ports you choose – and the beauty of this machine is that you can select from a wide variety of options – but will include a power connector port at a minimum, two DisplayPorts, a USB-C 3.2 Gen 2x2, USB-C with Thunderbolt 4, and four additional USB-A ports (two USB 2.0 ports and two USB 3.2 Gen 2 ports). You'll also get an RJ-45 Ethernet port, which supports up to 1 Gbps internet speed.

Optionally, you may choose one from the following selection: HDMI 2.1 port, VGA port, two USB-A 3.2 Gen 2 ports, a USB-C port with DisplayPort, a USB-C 3.2 Gen 2 port with Thunderbolt 4, a 5GbE LAN NIC, and another DisplayPort 2.1 port. There's also no wireless card in the machine by default, so you'll have to pay extra for Wi-Fi 7 2x2 and Bluetooth 5.4 support courtesy of either an Intel or Qualcomm chip.

Dell Pro Max Tower T2: Performance

The processor at the heart of our review configuration was the 24-core Intel Core Ultra 9 285K processor, a power-efficient and versatile high-end CPU that's optimized for AI and creative workloads. With a frequency of 3.7 GHz (with a maximum of 5.7 GHz), 16 efficient cores, and eight performance cores, it's a formidable upgrade to the Intel Core i9-14900K, previously one of the top desktop CPUs.

Data from Geekbench 6 confirms really positive results, with an improvement over its predecessor; it scored 3,202 in single-core performance as well as 22,583 in multi-threaded performance – a decent gain on the i9 predecessor. While plenty of AMD processors can outpace it – including multiple chips in the AMD Ryzen Threadripper PRO series – the Intel chip represents fantastic value for money. Meanwhile, as far as memory goes, our review machine was fitted with 32 GB DDR5 RAM and two 1TB M.2 PCIE 35 Performance SSD modules. The SSD speeds are phenomenal, with reads of 12,008MB/s and writes of 12,674MB/s.

The GPU, on the other hand, is an unparalleled monster. Nvidia describes its RTX PRO 6000 Blackwell Workstation Edition as the most powerful desktop graphics card available – and benchmark data offers compelling evidence for why. That said, it'll cost you a pretty penny, and the vast majority of the cost of the machine is in this beast of a graphics card. Combined with the CPU and other specs, there's very little this can't handle.

Dell Pro Max Tower T2: Is it worth it?

The Dell Pro Max Tower T2 is about everything you'd expect from a modern desktop PC, with a plethora of connectivity options that can be customized to your liking as well as a vast choice of hardware to suit your specific needs – from integrated Intel Arc Graphics to the most powerful desktop GPU available.

The machine itself isn't much of a looker. The front-facing grill adds some much-needed flair to an otherwise bland design, but it does give really strong airflow and thermal management when called on. Given the classic 32L chassis design, as well as ease of access for maintenance or upgrades, it's a versatile box too.

While our review unit was one of the most expensive configurations out there, for most office users and professionals, you can easily customize a version of this device that will perform more than well enough for under $2,000.

Dell Pro Max Tower T2 specifications

Unless you've been living under a rock for the past 18 months, it's impossible not to be aware of how many of its eggs Nvidia has put in the AI basket.

To further stoke the AI fire, Nvidia released a dedicated desktop box aimed at AI developers called the DGX Spark, which combines Nvidia's latest generation of ARM CPU, GPU, and networking technologies in one user-friendly and affordable package.

The phrases "user-friendly" and "affordable" are doing a lot of lifting in that previous paragraph, but given the lack of alternatives if you want to develop your AI project locally, this is as good as it gets.

Nvidia has also partnered with several OEMs to bring variations on the Spark theme to market. All the OEM boxes share the same basic specification as the Nvidia Ur-Box, suggesting Nvidia is keeping everyone on a pretty tight leash.

Dell has released the Dell Pro Max with GB10, Gigabyte the AI Top Atom. Lenovo with the ThinkStation PGX, MSI with its EdgeXpert MS-C931, and finally Asus with the Ascent GX10.

It's the Asus Ascent GX10 we are looking at today. All the variations on the DGX Spark will set you back around £4,000, depending on the amount of storage it comes with; the Ascent we are poking with a stick can be picked up in the UK for £3,640.

Asus Ascent GX10: Design

Given the esoteric hardware that lurks inside the Ascent GX10 is impressively small, stylish, and unobtrusive. It also has one major advantage over Nvidia's own box and the other OEM competition in that the power switch is at the front rather than the back or, in Dell's case, the top.

That may not be a big deal if you just have one box sat next to your monitor, but if you have several arranged in a rack, and that is very much how this system is intended to be used, then you'll be left groping around the back to power your system up.

Physically, the Ascent GX10 is a 150 x 150 x 51mm metal box weighing 1.48Kg. The front, back, and side panels are painted in matte slider while the top has a cross-hatched mirror finish.

The wavey slats on the front presumably provide more surface area for cooling, but as essentially an aesthetic consideration. Compared to the rather baroque DGX Spark and the utilitarian offerings from the other OEMs, the Asus box is by far the most visually appealing.

To be fair to Nvidia, the external design of the DGX Spark is very much an homage to the 2016 DGX1, a $130,000 61Kg lump the size of a large suitcase with a paltry-by-modern-standards 170TFLOPS capability.

All the ports are grouped at the rear below the large air vent. Running left to right, you'll find a Kensington lock, four 20Gbps (Gen 2×2) USB-C ports, one of which serves as the connection for the 240W power brick, an HDMI 2.1 video output, 10GbE Ethernet, and two 200Gbps ConnectX-7 QSFP112 ports.

Incidentally, the three USB-C ports that are not powering the box also support DisplayPort 2.1 DP alt-mode video output. Another consideration is that the GB10 chip has a TDP of 140W, which leaves 100W for all the other bits and bobs, including the ConnectX-7 NIC and the three USB ports. Wireless communications are handled by a MediaTek card that supports Wi-Fi 7 and Bluetooth 5.4.

Internal access is minimal, but given that everything important is soldered in place, this is not the end of the world. Unscrew the base grille and then another metal plate, and you can easily get at the 2242 SSD, which means you can buy the cheapest version of the box and upgrade the storage should the need arise.

Dig deep into the Asus product sheets, and you'll find that only the 4TB configuration ships with a PCIe 5.0 x 5 SSD. The smaller SSDs are PCIe 4.0 x 4 specification.

Like the Nvidia DGX Spark and all the other OEM iterations thereof, the Ascent GX10 runs on NVIDIA's DGX OS, a variant of Ubuntu 24.04 LTS (for ARM) Linux.

The primary differences between DGX OS and vanilla Ubuntu are really just a matter of the apps Nvidia has added, such as the dashboard to manage your AI projects, Nvidia Sync utility that manages back-end SSH keys and tunnelling, and all the relevant drivers.

The OS dashboard also directly links to Nvidia's DGX Spark Playbooks repository on GitHub. These provide a great selection of programmes for getting a wide variety of AI models up and running, including ComfyUI, Pytorch, and vLLM, to name but three.

These playbooks are an important part of the GX10's appeal to those who are rather new to the world of AI development, and one reason why the GX10 or any of its Spark cousins are perhaps a better bet for AI developers than the likes of a Mac Studio or a do-it-yourself rig built around an Nvidia discrete GPU or two.

Asus Ascent GX10: Specs and Performance

Running the show inside the GX10 is an Nvidia GB10 ARM v9.2-A CPU, which consists of 10 Arm Cortex-X925 and 10 Arm Cortex-A725 cores, a Blackwell GB20B GPU. You also get 128GB of soldered LPDDR5X-8533 RAM, a brace of ConnectX-7 200Gbps networking ports, and a 4GB 2242-format SSD.

Versions with a 1GB or 2GB SSD are also available for rather less money, but the majority of the boxes on sale through retail channels are the 4GB variety, and given the intended use, that would surely be the model to go for.

Before we press on, it's worth making clear that one reason these Nvidia-system boxes are so expensive is because of that Nvidia ConnectX-7 interface, which easily accounts for upwards of one-third of the asking price.

It goes without saying that 200Gbps of bandwidth is nothing like what you get from a full-fledged GB200/GB300 server, but it does provide developers with access to more power and also a way to see how their models perform on a scaled setup.

Measuring performance is borderline superfluous on a machine that boasts 1 petaFLOP (that's 1,000 TFLOPS in old money) of performance, and our testing regime isn't suited to measuring the exact potential of GX10 to run large LLMs and VLMs at the same time or create local AI agents.

Asus's demonstrations undoubtedly confirmed Nvidia's performance claims, such as the impressive abilityfine-tuningning Pytorch LoRA llama3 8B at 53K peak tokens per second (TPS) and generating a 1K FP16 image in Flux1.dev 12b in double quick time thanks to having ready access to the 35GB of RAM that the job required.

In a rather unscientific test, we ran some basic prompts in LM Studio on both the GX10 and the Framework Studio. The Asus box averaged 4.67 tokens per second while the Framework came in at 13.4 TPS. Of course, our Framework box only has 32GB of RAM.

Running some 1K image creation tasks in ComfyUI gave the Asus box an even larger advantage, the GX10 taking 216 seconds to do a job that the Framework box took over 400 seconds.

At a more general level, the GB10 CPU scored 3,104 in the GeekBench 6 single-core benchmark and 20,048 in the multi-core test. For comparison, a 2024 Mac Mini running on the M4 Pro CPU scored 3,986 and 22,857, respectively, while the AMD Ryzen AI Max 385 CPU in the Framework Desktop scored 2,952 and 16,599.

Throughout all the test the GX10 impressed by how quiet it was, the dual fan system seldom making much more than a loud whisper. Under heavy stress, the external case got a little warm in places, but by no more than your average laptop when under duress.

Our takeaway is that if you are developing large AI frameworks and want to do so locally, then the GX10 and its relatives are the smallest, fastest, and most convenient way to do it.

Granted, you could build yourself a development rig based on a couple of Blackwell gaming GPUs and potentially get more performance, but the result would be bigger, louder, far more power hungry, and probably more expensive when you factor in the cost of RAM, a CPU, a box, and everything else, let alone a comparable network card.

Asus Ascent GX10: Is it worth it?

With an item as esoteric as the Ascent GX10, that's not an easy question to answer. For a substantially smaller pile of cash, you can buy an Apple Mac Studio or a system built around one of AMD's Ryzen AI Max chips, like the Framework Desktop modular PC we looked at recently or HP's new Z2 Mini G1a Workstation, both of which can be configured with 128GB of RAM.

Of course, none of those machines have two 200GbE OSFP112 networking ports, and that's the first area in which the GX10 excels: scalability. Buy two of these things and hook them together, and you will really be cooking with gas. Though you'll also be over £7,000 out of pocket.

The value of the various Nvidia AI playbooks shouldn't be overlooked either. Having some GB10-specific workbooks is something we think will prove useful to developers wanting to further hone and expand their AI capabilities.

The Ascent GX10, then, is very much a thoroughbred designed for a specific job. Many developers will be happy with one of the forthcoming AMD Ryzen Max+ boxes, which are not only cheaper but run Windows 11, making them more versatile, but if you are deadly serious about your AI development and actually need the performance that more than one GX10 box can deliver, then it's a pretty impressive device for the asking price.

Asus Ascent GX10 Specification

Since its launch over nine years ago, Qnap's mighty 16-bay NAS cube still holds the record as the largest desktop storage appliance on the planet. On review, we have the latest TVS-AIh1688ATX, which targets AI apps such as image recognition and analysis, along with a range of enterprise tasks, including data backup and virtualization.

Starting life as the TS-1635, chassis design has generally remained the same but has seen many subsequent hardware upgrades as it's gone from a humble Amazon Alpine to an Intel Xeon W CPU in the TVS-h1688X. The TVS-AIh1688ATX takes performance to the next level as the top-end model we test sports a modern 24-core Intel Core Ultra 9 285 CPU teamed up with 48GB of DDR5 memory.

With eight P (performance) cores, sixteen E (efficient) cores, an embedded Intel Graphics GPU, and Intel AI Boost NPU (neural processor unit) to play with, the appliance boasts a high 36 TOPS (trillion operations per second). Typical AI-based applications Qnap is aiming the appliance at include its QuMagie photo album categorisation, along with the QVR surveillance app's facial recognition and enhanced people counting.

Qnap TVS-AIh1688ATX-U9-48G review: Hardware changes

Along with a hugely upgraded CPU, the appliance introduces significant updates in the storage department. The twelve main drive bays still support LFF and SFF SATA HDDs and SSDs, but the bank of four SFF bays can now handle U.2 NVMe Gen4 SSDs as well as SATA models.

One feature we're sad to see go is the dual internal M.2 NVMe SSD slots. Qnap recommends running its QuTS hero OS on SSDs for the best performance, and previously we've always used the M.2 slots for these duties as they free up the main storage bays. On a brighter note, Qnap has sensibly provided the 48GB of base DDR5 memory on one stick, so you can upgrade to the maximum 192GB without having to replace it.

Networking sees changes too, as although the 2.5GbE multi-Gig port count has been halved, the dual-port 10GbE adapter card in the TVS-h1688X has been replaced with two embedded ports. This matters as it frees up the inner PCIe x16 expansion slot, which supports 260mm full-length full-height cards – the two outer half-height slots have length restrictions of 150mm due to the positioning of the dual CPU blower fans in front of them.

The TVS-AIh1688ATX is blessed with a wealth of USB ports, as you have two USB-C 4, two USB-C 3.2 Gen2 plus three USB-A 3.2 Gen2, and you can add dual Thunderbolt 5 with Qnap's optional QXP-T52P PCIe card. The HDMI port has also been upgraded from 1.4b to 2.1, allowing it to support 4K video at 60Hz.

Qnap TVS-AIh1688ATX-U9-48G review: Deployment and 10GbE performance

Qnap doesn't enforce the same draconian drive restriction policies as Synology, and we had no problems using two 1TB Western Digital SA500 SATA SSDs for system pool duties, four 22TB Western Digital Red Pro NAS HDDs for general data storage, and a quartet of 3.84TB SanDisk U.2 NVMe SSDs for performance testing. Deployment is as simple as Qnap's QFinder Pro desktop app spotted the NAS on our network and transported us to the browser-based setup routine, which automatically installed the latest QuTS hero h6 OS – more on that later.

Using the Storage Manager app, we set the SATA SSDs up as a RAID1 mirror while the HDDs and NVMe SSDs were configured as RAID5 arrays. With a NAS share on the NVMe array mapped to a Dell PowerEdge Windows Server 2022 host over 10GbE, we watched Iometer report a splendid 9.3Gbits/sec across the board for sequential and random read and write operations.

It was the same story for IP SANs with a 1TB iSCSI target delivering identical speeds across all tests. We ramped up the pressure with a dual 10GbE MPIO link to the target and saw sequential read and write speeds increase to 18.5Gbits/sec and 17.4Gbits/sec while random operations returned an equally impressive 18.5Gbits/sec and 16.7Gbits/sec. We also noted that the Intel CPU never went about 24% utilization during our test, so there's plenty left in the tank for more demanding tasks.

Qnap TVS-AIh1688ATX-U9-48G review: QuTS hero h6 new features

The latest QuTS hero h6 has a sharp focus on security and the TVS-AIh1688ATX is currently the only Qnap NAS that supports the new Secure Boot feature. Implemented at the hardware level, it verifies the appliance's firmware signature and prevents it from loading if it's been compromised.

There's support for FIDO2 passkeys for secure password-free logins and the new KMIP (key management interoperability protocol) app connects with remote KMIP servers for managing, storing and retrieving storage encryption keys. Other security features on the horizon are the Ransomware Guard which is accessed from the Malware Remover app, detects known malware and monitors processes for suspicious activity while the Secure IP Access feature will augment the QuFirewall app with granular IP-based access controls.

The High Availability Manager app allows two identical appliances to be placed in a fault tolerant active/passive cluster and has been updated with support for more Qnap NAS models. The Control Panel, iSCSI & Fibre Channel and Storage Manager apps see substantially redesigned interfaces making them even easier to use. Snapshot management gets its own dedicated app which adds extra ransomware protection with policies for creating immutable snapshots that cannot be tampered with or deleted until the retention period you set has expired.

Previously, Qnap's Qtier app was only available with its QTS OS but this has now been integrated into QuTS hero. This manages up to three storage tiers comprising NVMe SSDs, SATA SSDs and HDDs, monitors storage pool activity and applies custom schedules to control the migration of hot data across the tiers.

Qnap TVS-AIh1688ATX-U9-48G review: Is it worth it?

With Qnap's UK online store listing it for £5,660 including VAT, the TVS-AIh1688ATX-U9-48G represents a considerable investment but does deliver an impressive hardware package for the price. Qnap also offers the lower-end TVS-AIh1688ATX-U7-32G, which comes with a 20-core Intel Core Ultra 7 CPU plus 32GB of DDR5 and costs £4,464.

Its powerful CPU is clearly capable of handling a wide range of surveillance tasks, and we were impressed with Qnap's QVR Face Insight and QVR Human apps as they worked extremely well with fast reactions to activity in our predefined camera detection zones. The QuMagie app also delivered efficient categorisation, face recognition, and thumbnail generation for our big photo collections.

For surveillance operations, the appliance has almost no competition as Synology's current AI-powered NVR alternatives have dwindled to its entry-level 2-bay DVA1622. The high storage bay count and feature-rich QuTS hero h6 OS also make the TVS-AIh1688ATX appliances great all-rounders that look capable of handling a wide range of enterprise data protection, backup, and general storage duties.

Qnap TVS-AIh1688ATX-U9-48G specifications

Santa Clara, California headquartered, 100% indirect Pure Storage added 12% year-on-year to its $3.17 billion revenues for full-year 2025, and to its Q1 2026 revenues of $779 million. Product sales rose 5%; subscriptions 22%.

Founded in 2009 by John Colgrove and John Hayes as Os76 before branding as Pure Storage in 2011, Pure is 6000 employees strong, led by CEO Charles Giancarlo. The focus is on flash and data management, integrating storage hardware, embedded licensed software, and services. Partners like Kyndryl, Tech Mahindra, and NTT are joined by resellers like Bechtle, CD,W and ePlus.

Geoff Greenlaw, EMEA and LATAM channel sales vice president, says it's expanding flash into historical disk use cases, leveraging its flash software as well as quad-level cell (QLC) flash.

General manager, commercial, Amy Fowler, oversees regional channel heads, including Greenlaw, the Americas' Hope Galley, and Andrew Fisher for Asia-Pacific and Japan (APJ).

Key offerings

Within the Pure Storage platform and Pure's Enterprise Data Cloud concepts are several offerings expected to power partner revenues. And in June, Pure announced various updates targeting volatile, power-hungry workloads.

Within the Pure Storage platform and Pure's Enterprise Data Cloud concepts are several offerings expected to power partner revenues. And in June, Pure announced various updates targeting volatile, power-hungry workloads.

FlashBlade hardware - unified file and object storage for unstructured data. For instance, the FlashBlade/EXA storage platform is being marketed towards AI deployments seeking to make better use of organizational data.

As Greenlaw tells ITPro: "{FlashBlade's] very focused on very high-end performance and throughput. It can become very big in healthcare where there's high volumes, very large volumes of data that need storage for long-term data retention."

FlashArrays C and X - "We're pushing on with our FlashArray tech, which is really about database[s]. In addition, we have major enhancements to our X and XL range, which is up to 30% faster than its predecessor. That's mission-critical in the world of database[s]," he says.

Portworx - a cloud-native container data management platform for Kubernetes.

"We're seeing companies increasingly competing with Broadcom VMware," Greenlaw adds. "VMware has drastically increased the price of their licensing, and customers are looking for alternative virtualized environments."

Partners are seen as drivers for seamless mobility, cost optimization, resiliency, sustainability, consistent storage services, and unified data environments, including tailored and certified offerings for AI and high-volume data. In addition, Pure Fusion v2 software caters to AI-focused data management, with Pure1 AIops included in active subscriptions and support contracts.

Pure Storage in the news

• Founder John Colgrove, in June, talked up an all-flash future, with Pure also doubling down on data management.
• Also in June, Pure//Accelerate showcased its high-performance flash hardware, broad software stack, STaaS, and Enterprise Data Cloud (EDC) for combined data storage and management.
• January 2025 saw Pure announce its hyperscaler 'design win' licensing deal as hastening the death of disk.

Pure Storage for partners

Greenlaw says February's partner program update will fuel profitability for partners and Pure as changes bed in, while tech alliances with vendors such as Rubrik and CrowdStrike will further enhance its portfolio with regard to gaining net new logos (adding new customers).

"We're really doubling down on fewer, bigger bets within our channel to make Pure far more relevant to them, rather than spreading our revenue across 1000s of partners."

Partners lead customers through data storage transformation, leveraging the Pure Storage Platform for solution-specific use cases from AI to cyber-resilience. Data access, preparation, and management often bar AI success, for instance.

Also, partners will be incentivized to replace disks with Pure flash, with new solution-oriented training and enablement.

Partners are crucial to Pure's sustainability push, Greenlaw says, not least because data centers suck 2-3% of generated electricity worldwide, with datacentre storage, cooling, and space comprising most of that.

Services remain "highly relevant" to partners too, including storage as a service (STaaS) EverGreen//One, //Flex, and the more traditional //Forever.

Meanwhile, Pure's Service Specialization Program means partners can offer their own branded professional and support services for Pure technology.

In the partner program announcement, channel chief Fowler said the "reimagined" program experience will position partners for rapid growth and customers for "a sustainable future".

"Pure Storage is deeply committed to driving shared success with its partners," she says.

Partner programs, tiers, and partner types

• Partners are divided into resellers, distributors, global systems integrators (GSIs), and managed services partners (MSPs). In addition, there are Services specializations, and technology alliance partners for jointly engineered solutions.
• Tiers are Elite or Preferred, based on annual dollar values booked and technical certifications, with Elite partners eligible for higher back-end rebates and more investment in training and certification.
• Pure Storage Global Enablement certifications last two years.
  • FlashArray Architect Associate
  • FlashBlade Architect Associate
  • Platform Solutions Associate
  • FlashArray Architect Professional
  • FlashBlade Architect Professional

Partners should first qualify for the Pure Platform Positioning Certificate on explaining and selling Pure, then choose from product-focused learning paths and assessments, with or without additional training as required. Exam ordering instructions are here.

• IT professional services certifications.
  • Associate (data storage)
  • Professional (FlashArray, FlashBlade, Portworx, data protection)
  • Specialist (FlashArray and FlashBlade implementation or support, cloud block store, migration)
  • Expert (platform architect)

Further benefits, rewards, and incentives

• Individual and entity rewards for Preferred and Elite partners with partner-sourced competitive take-out deals.
• Rebates for Elite partner subscription bookings and new or strategic solutions sales.

Beyond the above, other rewards or benefits won't apply to all regions or partners, Pure says.

• Deal registration.
• Pure Digital Master Services Agreement (PDMSA) - margin for self-service purchases.
• Pure Partner Intelligence - real-time insights and analytics dashboard.
• Pure Realize - support for use-case proposals, price quotes, and solution expertise.
• Configure, price, quote (CPQ) tooling.
• Pure Storage digital badges.
• Partner Library of resources and info - log in here.
• Training catalog via Pure's Partner Academy page - log in with Partner Account details.
• Further enablement via the Partner Central partner
• Collaborations with the Technical Services teams.
• Collaborations with Business-Critical Services teams.
• Pure360 demos, videos, and test-drives.
• Pure Storage Partner Finder listing.
• Monthly cost estimator for Evergreen//One partners.
• Cost calculators for cloud block storage, STaaS, and FlashBlade//S.
• Remote installation support for FlashArray and FlashBlade.
• Further online resources - online solution briefs, how-tos, webinars, blog posts, e-books, and whitepapers.

Sign up and get on board

To become a partner, explore program details and express interest by contacting Pure Storage through its website. No membership fee is required just to join; eligibility is determined by actual sales activity. Send an enquiry on this page.

For support and questions, try this web form, or contact sales in the US by phoning 800-976-6494 or emailing info@purestorage.com.

Pure Storage is also on LinkedIn, X, Instagram, YouTube and Facebook.

Welcome to ITPro's live coverage of HPE Discover 2025. We're on the ground at the Fira Grand Via in Barcelona this week to bring your rolling coverage of all the big news and announcements from Hewlett Packard Enterprise’s annual European conference.

While we're waiting for the day two keynote to begin, you can find all our coverage from the show so far here.

The show floor is already getting busy, both with people and tiny football playing robots.

Here's one of the bots in action – getting a bit overwhelmed by the attention at the end it seems. According to our tour guide they're not just on the floor to look cute and entertain – they demonstrate technology and connectivity that can be used in manufacturing.

The auditorium is starting to fill up as CEO Antonio Neri is expected on stage imminently

And here's the man himself. He's addressing the delegates, acknowledging that there's a mix of technologists and business leaders here. He's also – as expected – talking about the future of AI including 'physical AI', AKA robots. Anecdotally, this is a trend that I'm hearing more and more, which feels very "back to the future" – it was a topic of conversation pre-pandemic, as was digital twinning...

Networking, cloud, and AI are the watch words here – a bit of a shift from the edge to cloud to AI mantra of Discover Las Vegas 2024, but an understandable one. The Juniper Networks acquisition (which he's discussing now) has perhaps obviously made networking an important story for HPE. Realistically, it's also one of the main enablers of edge computing anyway.

Neri adds that he's excited to share some news about how Juniper Networks (now HPE Juniper Networking) will integrate with the company's existing Aruba Networking portfolio.

Joining Neri on stage now is Rami Rahim, formerly CEO of Juniper Networks, now VP and GM of networking at HPE. He's talking up the importance of networking in the future of enterprise infrastructure, including AI infrastructure connecting thousands of GPUs together for massive processing power.

Rahim is also talking about the idea of "self driving" networks. Expect to hear more about this both here at Discover and from HPE in the future. It seems to be a key marketing point, USP, and concept for how AI networks will work in the future.

Rahim is now talking about the differences between Juniper Mist and HPE Aruba Networking Central, how Mist was built for the cloud while Aruba was more of an on-premises play.

"We have two industry leading network AI platforms, Aruba Central and Juniper Mist, Aruba Central and Juniper Mist, both delivering real and valuable results to our customer each," says Rahim. "Every day by adopting our platforms, our customers have seen up to a 90% reduction in travel tickets and on site service calls. They've also seen a drastic decrease in the amount of time they have to spend in hands on network operations."

"Each of these solutions is unique and compelling in different ways," he says.

The two platforms are being "cross pollinated", says Rahim. Some elements of Mist are moving to Aruba Central and vice versa. Eventually, says Rahim, there will be no difference between the two.

HPE is also launching a joint WiFi-7 access point that unifies Juniper and Aruba technology. This will be launching in Q3 next year, around the time of HPE Discover Las Vegas 2026.

You can read more on these announcements in my full write-up here.

We're now onto GreenLake, the ever shifting backbone of HPE offerings. For those who have been familiar with HPE for a while, you'll know that GreenLake was originally pitched as something akin to a cloud service, with on-demand and consumption-based pricing at its core. It's now morphed into something like a managed service.

"We make deploying and managing AI data centers simpler and much more efficient, but we don't stop there; with GreenLake only HPE can provide end to end data center solutions across compute, storage and networking, along with expertise and services," Rahim explains.

Some more product announcements: the QFX5250 switch with HPE tomahawk 6 technology and MX301 edge router. Once again you can read more here.

Rahim is also talking up the AMD Helios and Nvidia AI Factory announcements from earlier in the week.

"I could not be more proud of our partnership with Nvidia, and I believe these joint solutions will give customers the assurance they need to deploy our routing technology with Nvidia's cutting edge products with confidence," he says of the continuing Nvidia partnership.

On AMD, he adds: " Iam stoked to announce the HPE will be offering AMD Helios AI rack worldwide starting next year. The Helios rack scale architecture is an open platform for large scale AI workloads, putting industry standards at the forefront of the design. So recall I said earlier that scale up networking has been uncharted territory. Well, that's about to change. I'm very excited to announce an HPE Juniper network and scale up youth nets, which designed specifically for Helios AI rack. This is an industry first scale up solution using open standard Ethernet, in other words, no proprietary vendor login."

Wrapping up, Rahim says: "It's been, what, less than six months since we closed this acquisition, and I hope you agree with me that we have already made massive progress in our integration efforts, the fact that we are here today announcing new products and solutions and not just announcing the intent, but showing them the help on the Discover show floor should tell you that we are committed not only to our vision, but to truly bring it into life."

"Now as we look forward, one thing is clear, it's time for a new generation of networks – HPE networks – to be a new leader," he continues. "We have the tech. We have the track record of innovation. We have the worldwide go to market scale, including a fantastic partner community.

"We have a bold vision, and we have the team that can execute on that bold vision and bring it all to life. So a very big thank you for believing in us and for supporting us. We could not do it without you. Thank you for joining us on this journey."

Neri is now back on stage, continuing Rahim's thoughts saying: "Personally could not be prouder of what a new combined network HPE has already accomplished and what they will do next. Rami and I had the same vision from the very beginning, which was to create a new network together, and I truly believe we are on the path to do so."

We're now moving onto hybrid cloud and, inevitably, GreenLake.

"Hybrid Cloud gives you resilience, flexibility and speed with more control and lower cost, and that's why customers are choosing GreenLake," claims Neri. "When we first mentioned GreenLake, our goal was not to build another cloud. It was to deliver a better cloud experience, one that ... [offered] a consistent cloud operating model across your entire IT estate."

A few connectivity issues have meant I've not been able to post some updates on GreenLake, Alletra Storage, and security. There will be some standalone articles on these announcements from me later!

A few bullet points on what was announced in the meantime

• New capabilities for HPE Morpheus Software, including software-defined networking for VMs hosted on HVM hypervisor in VM Essentials and Morpheus Enterprise Software
• Apstra Data Center Director from Juniper, which will be integrated into Morpheus
• Stretched cluster capabilities for Morpheus
• Zerto cybersecurity software is being integrated into Morpheus
• HPE VM Essentials backup on Veeam Data Platform 1.3
• HPE Alletra Storage MP X10000 Data Intelligence Nodes with Nvidia AI Data Platform reference design

Neri is now going over some of the supercomputing announcements that were recently made just before and at SC conference in October. You can read more about the new Cray GX5000 and K3000 systems here, and the DoE supercomputer announcements here.

Welcome to Day Two of HPE Discover Barcelona 2025. No fainting robots or other foibles on the way to the main theatre this morning, sadly, but we're expecting the networking general session (aka keynote) to kick off any moment.

Networking has been one of the big topics so far this week, with former Juniper Networks CEO Rami Rahim taking up almost the same amount of time on stage during Antonio Neri's keynote yesterday as Neri himself.

Roll VT, it's time to start the general session.

Rami Rahim is on stage now. "It has been quite a year for us at networking inside of HPE," he says, which feels like a bit of an understatement.

The networking side of HPE's business is now 20,000 strong, says Rahim. "It's absolutely remarkable how quickly Aruba and Juniper have come together and started working side by side on a common vision, with a common objective, to create a new industry leader in network," he adds.

"Networking has never been more important in our industry, and networking has never been more important for HPE," says Rahim. This has been the theme of the whole event – HPE is a server, storage, AND networking company. Aruba has, of course, been in place for 10 years but this emphasis on networking is stronger than ever, in my experience.

We have a customer on stage now for a chat with Rahim from Lanes Group, a water treatment provider, to explain how important networking is for their business. What this means, in practice, is connecting CCTV cameras, sensors, and other operational data. They're also implementing AI for predictive maintenance, so anticipating a problem before it becomes an issue, as well as agents.

There's about an hour left of the general session and apparently we're going to be getting some live demos of the technology. Always a brave choice...

Another exec has taken to the stage to talk about uptime and the importance of the "self driving" network enabled by Marvis (a technology that came over from Juniper but is now with Aruba via the 'cross pollination' policy) and Aruba Central (still part of Aruba).

A network being "up" is not the same as being "good", he says, and really UX needs to be part of the networking conversation.

After a short VT based on two real life scenarios, we're now into what a self-driving network means. There are three elements: speed of rollout, fewest tickets and the perhaps more subjective idea of "better business outcomes".

The self driving network is dependent on AI agents and microservices, he says, highlighting the existing reputation of HPE Aruba Central and Juniper Mist.

HPE are leaning hard on the example of autonomous vehicles to explain this – how we've gone from cars with at best an oil warning light to parking assistants, blindspot cameras etc

We're having a demo of Marvis, which is a natural language trouble-shooting bot within Mist, to be perhaps a little reductive. It's a previously created demo, which is probably for the best for all involved.

The announcements are coming thick and fast. An inevitable focus for a European conference is digital sovereignty and we've just had the announcement that microservices through Aruba Networking Central will be available 'on prem' – inverted commas because some of the data processing is still done in the cloud, but inferencing is done on premises.

Ok some more innovation announcements. First "The power of agentic AI" – Apstra Data Center Director can identify if one device "catches a cold ... who will start sneezing next". Marvis can then be used to identify the issues and help fix them via natural language interactions. Worth noting that while these are novel to HPE, Marvis and Apstra have been imported through the Juniper acquisition.

Next innovation: This is about the importance of open systems, which the presenter says is "important for cloud and neocloud providers". Consequently the QFX5240 and QFX5250 switches are available immediately with support for SONiC (Software for Open Networking in the Cloud).

Innovation three: Data center networking operations are now integrated with Opsramp.

This means a "seamless solution that locates your application and network issues", he says.

Innovation four appears to be... Rami Rahim??

Oh no, innovation four the QFX5250 data center switch. This is apparently the frist switch using Broadcom Tomahawk 6 chips and also features direct liquid cooling, which is a specialism for HPE on its server infrastructure. You can read a bit more about the QFX5250 here.

AE Natarajan is next on stage talking about networking for the post-internet, AI era

"When we first started the internet, it was built and architected for downloads," he says. This could lead to buffering issues for large files (e.g. from Netflix), which led to caching to overcome this issue.

AI doesn't work this way, in his words AI "never sleeps". Cached data is instantly obsolete in the world of AI, and upload/download is symmetrical. Constantly, the network architecture needs to be completely different.

Ok, we're onto network security. "While AI is making our life and we can speed up things. Guess what? AI is also doing the same things to the bad actors in the world. It's the same tools that is available that these guys become more sophisticated," he says.

David Hughes, SVP and GM of SASE and security, is now on stage talking about the overlap of networking and security.

"Traditionally, the networks team is tasked with ensuring that there's great performance and reliable connectivity for people, for devices, for workloads, he says. "[For] decades now, networking has at its foundation ... the ability to connect any endpoint to any other endpoint."

"Meanwhile, the security team is tasked with protecting users data and workloads, and historically, this is being done by restricting, often with solutions that are altered on or overlaid on top of the template," he explains, adding that users are often "caught in the middle", which causes a bad experience for them.

"obody wants to choose between performance and protection. We need both, and what we believe is that a network with physical security is genius," Hughes says, before rattling through some of Aruba's pedigree in security despite being a networking company.

The addition of Juniper to HPE has enhanced this, Hughes says, highlighting the SRX appliance as a particularly stand-out example.

The answer to some of these issues is zero trust everywhere – stopping lateral movement, isolating compromised endpoints such as laptops etc.

Additionally, HPE has some product ranges that can help sitting in the categories of unified SASE, full-stack AIOps, universal zero trust networking access (ZTNA), and hybrid mesh firewall.

"Delivering with integral security, there's no need to choose between performance and protection," Hughes concludes. "We are innovating with the goal of providing solutions that align the working and security teams behind a common zero trust mission."

Rahim is back on stage to wrap up for the day. "Networking has never been more important," he says, "and this is why it is for technology. We have the track record of innovation. And most importantly, as I think you have seen by now, we have the team to deliver on everything you are seeing here at discovers. HPE networking was created to be you leader with a whole vision and the ability to execute on that vision and to bring it to light."

With that he thanks the audience and it's lights up in the auditorium. This is the end of the live blog, but not the end of our coverage, so stay tuned for more throughout this week.

Amazon Web Services (AWS) has announced significant performance improvements for its Amazon S3 Vectors service in a bid to compensate for surging enterprise AI workload demands.

New updates for the cloud object storage service, first unveiled in July this year, mean S3 Vectors now scale up to two billion vectors per index, which the hyperscaler noted is a 40-times increase compared to preview capacity.

The service also supports up to 20 trillion vectors per bucket, AWS revealed, and will offer users 2-3x faster query performance.

The update aims to accommodate surging data volumes amidst continued enterprise AI adoption, according to AWS. The S3 service already stores more than 500 trillion objects and hundreds of exabytes of data, AWS noted.

To meet this growing demand, the hyperscaler plans to increase the maximum S3 object size to 50TB, marking a ten-times increase. This, the hyperscaler said, will enable customers to "store massive data files like high-resolution videos, seismic data, and AI training datasets as single objects in their original form".

Amazon S3 Vectors

When AWS announced the launch of Amazon S3 Vectors in July, it hailed the service as the first "purpose-built durable vector storage solution" aimed at improving affordability for enterprises with large-scale data storage requirements.

The hyperscaler claims the service can reduce the cost of uploading, storing, and querying vectors by around 90%. These cost savings are crucial given the intense enterprise focus on generative AI over the last three years.

Vector search is a technique used in generative AI applications to identify similarities between specific data points. According to AWS, vectors are a "numerical representation of unstructured data created from embedding models".

"You use embedding models to generate vector embeddings of your data and store them in S3 Vectors to perform semantic searches," the company explained in a July announcement statement.

In launching S3 Vectors, AWS aims to bring these capabilities to customer data, integrating closely with Amazon Bedrock Knowledge Bases and the Amazon OpenSearch service.

The goal here is to make it easier to build AI agents, RAG systems, and semantic search applications that understand both context and intent; crucial factors for enterprises developing customer-facing applications.

According to the hyperscaler, a host of customers, including BMW Group, Twilio, and Qlik are using S3 Vectors to "accelerate AI search and power recommendation systems at scale".

MORE FROM ITPRO

• AWS has a chance to show its mettle at re:Invent 2025

Acemagic is an emerging name in the mini PC space, and the Chinese brand has delivered a handful of diminutive computers to spruce up your home office setup. We enjoyed using the Acemagic Vista Mini V1 earlier this year, for example, despite a couple of niggles.

The Acemagic Matrix Mini M1 is the brand's newest and most powerful device, featuring an Intel Core i9-13900HK processor and plenty of RAM for a reasonable price. Inside the box, you'll get a power cable, an HDMI cable, as well as a VESA mounting bracket, as well as screws if you wish to fix it to your monitor.

Although it's available for $799 in the US (reduced at the time of writing to $459), it's priced at just £399 in the UK and comes with similar specs to the excellent Geekom Mini IT13 – which is considerably more expensive at an RRP of £849 (although it's likely you can nab all mini PCs for cheaper on various deals). One of the key differences is that you get double the RAM in the Acemagic, meaning it's a mouth-watering prospect if you're looking for a compact mini PC on a tighter budget.

Acemagic Matrix Mini M1: Design

The Acemagic Matrix Mini M1 design shares plenty in common with its more slender cousin, the Vista Mini V1. Unlike plenty of mini PC designs out there, its rounded corners and minimalist overall design feel like a nod to the gorgeous Mac Mini, which is one of the best-looking mini PCs ever made. This softens the look when you compare it to more rigid design choices, like in the Lenovo ThinkCentre M70q and even the Geekom Mini IT13 – which (to varying degrees) have much harder edges.

You can pick up the device in either a graphite or a silver finish, with both looking the part. If there's any downside, it's that its shell has a plasticy feel to it, which is also visibly noticeable whenever you glance towards it on your desk. While this isn't a major issue, it does cheapen the device somewhat – although any complaints immediately soften when you remember it's very reasonably priced.

This also shaves its weight to 520g – meaning it's much easier than many to carry around with you should you ever need to. It's nowhere near as light as the 245g Vista Mini V1, but it is easily much lighter than most – especially the 2.85kg Lenovo device. On the flipside, however, the device is chunkier than many at 128 x 128 x 41mm – a little more compact than the Mac Mini (127 x 127 x 50mm) but a little larger than many comparable devices. The aforementioned Geekom Mini IT13, for example, measures 117 x 112 x 49mm. These are slight margins, ultimately, but worth considering if your desk space is limited.

As for connectivity options, there are plenty. The machine supports Wi-Fi6E and Bluetooth 5.2 (many now support 5.3) in terms of its wireless options, although you are stacked with plenty of excellent I/Os – including four USB-C 4.0 ports, six USB-A ports (two USBA 3.2 Gen2 x 2 and four USB 3.2 Gen 1), as well as HDMI 2.0, DP1.4, and an 2.5GbE ethernet port. There's also a power cable input and a 3.5mm headphone jack. They're arranged excellently, with one USB-C port on the front and rear, alongside two USB-A ports on the front and four on the rear. The audio jack is also situated on the front, next to the power button. It's a smart layout that makes sure you can plug in from any angle. The one downside is a lack of an SD Card reader, but you could always invest in a converter if you find that you really need to use one.

Acemagic Matrix Mini M1: Specs and performance

Our review unit came fitted with the Intel Core i9-13900HK with 14 cores and 20 threads, although you could also pick up a version with the Intel Core i9-11900H processor or AMD's Ryzen 7 6800H chip. All our 45W processors, but the i9-13900HK has the highest core and thread count – and can reach a maximum turbo speed of up to 5.4GHz. This is complete with an Intel Iris Xe graphics card, 32GB DDR4 RAM, and a 1TB M.2 SSD.

The Intel processor is designed for professional use cases – including content creation (such as 4K video and photo editing), software development, as well as multitasking. Although it's nearing three years old, it holds its own in testing across the board versus the newer Skylake and Lunar Lake CPUs, only offering a little less power on average. But in the Acemagic, the CPU slightly underperformed against expectations.

On one hand, the processor scored 2,806 on single-threaded testing via Geekbench 6 – a highly respectable result. However, it scored only 8,912 on multi-core testing. This is less than the Geekom Mini IT13, which scored 10,286 – and a scan of public records suggests that the CPU can generally hit at least 10,000 in the test. This is disappointing – but not critical. It's worth saying, though, that the M4 in the Mac Mini trounces it, having scored 3,496 and 14,505, respectively.

The graphics card, on the other hand, pulled in a decent score of 13,514 in the OpenCL test, which is in line with the 13,854 that the outstanding Geekom A5 scored with its AMD Radeon Vega 7 chip in the same test. We should stress, however, that modern GPUs from the likes of Nvidia, AMD, and now even Intel can easily surpass 20,000. While this mini PC does pack a bit of a punch, it certainly doesn't come close to competing with the most premium hardware. That said, it's certainly a decent return for its price.

The SSD performed well enough, with reads and writes of 3,478MB/s and 3,370MB/s respectively – it's a little above average, so not a huge complaint, and does eclipse the speeds of the Geekom A5 as well as Geekom Mini IT3.

Acemagic Matrix Mini M1: Is it worth it?

For the current price of £399, the Acemagic Matrix Mini M1 is an excellent proposition. Although it's a little underpowered against what we expected, this is only marginally so, and you still get far more bang for your buck with it than you might with some of its key rivals.

The device is a little larger than most (marginally) but more than makes up for it with its minimalist and stylish design that won't look out of place on a neat and tidy desk – whether you pick it up in silver or graphite. Its connectivity options are also fantastic, with access to speeds of up to 40Gbps with USB 4.0 support in both its USB-C ports.

Acemagic Matrix Mini M1 specifications

A third of UK organizations are failing to back up all of their sensitive data, hindering their ability to restore operations and ensure service continuity in the event of a breach.

Almost half of those surveyed by security firm Cohesity reported not backing up all their workloads, including virtual machines, applications, and unstructured data, while 38% aren’t applying consistent data categorization, backup controls, and policies globally.

Crucially, only 45% consistently follow the ‘3-2-1’ backup rule - three copies of data, stored on two different media types, with one copy kept off-site. The same number are failing to keep tamper-proof copies by using immutability across all their organizational backup data to ensure resilience against cyber attacks.

Fraser Hutchison, VP for Northern Europe at Cohesity, warned that enterprises need to sharpen their focus on backup strategies as failure to do so could exacerbate recovery in the wake of a cybersecurity incident.

"Most organizations now recognize the need to identify phishing scams or social engineering tactics; however, we can’t lose sight of what to do when disaster does strike. While complete prevention is near impossible, assurance of rapid recovery is fully within organizational control," he said.

"Our research shows that UK organizations still aren’t taking adequate precautions when it comes to data backups. By storing data on immutable platforms, they can ensure business-critical information remains beyond the reach of adversaries and that operations stay up and running, even when systems are compromised."

Sharpen up on backups

Last year, research from Beaming revealed that more than 800,000 UK firms had lost company data in the previous five years, at a total cost of more than £1 billion a year.

Businesses with between 10 and 250 staff saw the worst effects, while the most expensive data loss incidents occurred in the manufacturing sector, where the median cost was more than £60,000.

Backup strategies are now front of mind for many IT professions, alternative research shows. A survey from Kaseya earlier this year found 30% are losing sleep over lackluster backup and recovery strategies, with some pushing for a stronger focus on this area.

Complacency was also identified as a recurring problem for many enterprises, according to Kaseya. Nearly two-thirds (60%) of respondents said they believed they could fully recover from a data loss incident in the space of a day.

The reality of the situation was different, however. Kaseya’s research indicated that only 35% actually could.

Can AI solve the problem?

The boom in the use of AI is a mixed blessing, according to Cohesity. While a third of UK firms believe that generative AI adoption is progressing much more quickly than risk tolerance, 54% agree that AI capabilities are highly effective for anomaly detection and analysis.

Half of businesses, meanwhile, said they find AI tools useful for accelerated threat hunting and investigation.

"UK organizations are understandably concerned about the risks of AI technologies, with GenAI adoption still outpacing risk tolerance," said Hutchison.

"However, our research reveals some silver linings: AI is already augmenting previously time-consuming, manual security processes. By the end of next year, AI will play an increasingly central role in data security operations, enabling businesses to identify and respond faster to threats."

MORE FROM ITPRO

• IT workers have become the cool kids on the block since the rise of AI
• The best backup software for your business
• How IT professionals can change careers to cybersecurity

Wasabi Technologies has cut the ribbon on the latest iteration of its EMEA partner program for 2026, as the cloud storage provider looks to drive further growth across the European channel.

The refresh aims to equip partners for the next wave of cloud adoption as AI workloads and data volumes continue to gather pace.

Following the move, partners now have access to a new Systems Integrator program for improved solution delivery, new AI-ready storage offerings, as well as a host of tools, certifications, and incentives.

In an announcement, Wasabi said it is also enhancing its channel-centric strategy in EMEA by actively recruiting new partners in the Nordic, Benelux, and Southern Europe regions.

“Europe is entering a new era of data growth as AI projects mature, and our partners are on the front line of that transformation,” explained Kevin Dunn, Wasabi’s vice president and general manager for EMEA.

“Our evolved partner programme puts simplicity, profitability, and scalability at the core — whether it’s helping customers dissolve data silos to successfully run AI models, manage operational resilience, or modernize backup.”

Wasabi eyes new partner benefits

The introduction of the new Systems Integrator program will help partners address complex and changing customer needs as AI transforms the security landscape, according to Wasabi.

This initiative equips partners with a range of technical tools, as well as bespoke marketing and enablement materials, to help deliver solutions across AI, physical security, and cyber resilience with “predictable” pricing options and high-performance cloud storage.

The vendor’s EMEA channel refresh also includes new AI-ready storage solutions capable of handling performance-intensive workloads, with partner margins that aim to drive profitability in competitive markets.

Meanwhile, sales and technical teams have access to hands-on enablement and Wasabi certifications, alongside co-marketing materials, dedicated account support, and incentives.

“We’re giving partners not just the technology, but the training, incentives, and margin to thrive in this fast-moving market,” Dunn added.

Wasabi’s cloud storage offering aims to provide organizations with an alternative to hyperscale cloud providers, offering a predictable and high-performance solution without the burden of hidden fees, complex tiers, or vendor lock-in.

Since the launch of its cloud storage platform in 2017, the firm has found success through its expanding ecosystem of independent cloud application partners, which includes GmbH, Waterstons, Autodata, and Dynamips.

MORE FROM CHANNELPRO

• Proofpoint targets further expansion with Cork investment, new AI innovation center
• Palo Alto Networks to acquire Chronosphere in $3.35bn deal
• HPE launches first phase of new-look Partner Ready Vantage program

Magnetic tape storage might seem like a relic of a bygone era, but it’s still going strong in 2025 – and new upgrades mean it’s set to get even better.

The LTO Program, a joint initiative between IBM, Quantum Corp, and HPE, has unveiled a new generation of LTO Ultrium data cartridges offering enterprises up to 40 TB of native capacity - marking a significant improvement on previous generations.

The launch of the new range comes in direct response to surging AI-related data storage demands, with the capacity boost underpinned by “Aramid”, a new base film material which offers longer tape lengths.

“Aramid permits the manufacture of significantly thinner and smoother media, enabling longer tape lengths in a standard LTO Ultrium cartridge form factor,” the organization noted in a statement.

“This material innovation provides an extra 10 TB of native capacity than the currently available 30 TB LTO-10 cartridge, which is manufactured using different materials.”

Stephen Bacon, VP for data protection solutions product management at HPE, said the new cartridges are aimed at enterprises spanning an array of industries dealing with high data volumes, from manufacturing to financial services.

“AI has turned archives into strategic assets,” Bacon commented.

“The new 40 TB LTO-10 cartridge will help enterprise-class organizations – across healthcare, financial services, media, research, manufacturing, the public sector and beyond – consolidate petabytes efficiently, strengthen cyber resiliency with true offline air-gapping, and keep long-term retention affordable and sustainable.”

The magnetic tape renaissance

Magnetic tape storage has been going through something of a renaissance in recent years, and the advent of generative AI in late 2022 has played a key factor in this revival.

Indeed, figures published by the LTO Program in early 2024 showed a whopping 152.9 exabytes of total (compressed) tape capacity was shipped in 2023. That record was broken in July this year, with 176.5 exabytes shipped.

This marked the fourth consecutive year of growth for tape storage, underlining its continued popularity.

Tape storage has a number of distinct advantages, including low cost, durability, and easy portability. According to previous analysis from the LTO Program, companies using tape recorded an 86% lower total cost of ownership (TCO) compared to disk storage.

TCO compared to cloud storage was also 66% lower across a 10 year period, figures showed.

Notably, the use of tape for unstructured data storage also adds to the appeal, with this now vital in the training process for large language models (LLMs).

Tape is only going to get better

Long-term, tape storage is only going to improve, at least if the LTO Program’s roadmap is to be believed.

Through generations 11 through to 14, enterprises can expect to see significant capacity gains, eventually peaking with a 913 TB cartridge.

“The new capacity points will prioritize reliability and compelling value in terms of cost per terabyte, and ensure everyday LTO Ultrium tape solutions will remain a practical and affordable storage choice for every type of business or workload,” the LTO Program noted.

“Meanwhile, the new maximum capacity of 913 TB for LTO-14 will reduce the time needed to store and recover data, while preserving exabyte-scale growth paths for tape libraries.”

MORE FROM ITPRO

• Our top picks for the best cloud file sharing services
• Looking for the best cloud storage for your business? Look no further
• Our top picks for the best cloud storage services in 2025

Qnap continues to make inroads into the enterprise storage market as the TS-h2477AXU-RP NAS appliance offers a powerful rack-mount package with an impressively large capacity. Stepping up as the new flagship model in Qnap's TS-hx77AXU-RP rackmount family, it also has a keen focus on value and is a lower-cost alternative to Qnap's 30-bay TS-h3087XU-RP hybrid appliance.

Powered by a speedy 3.5GHz 8-core AMD Ryzen 7 7700 CPU and up to 192GB of DDR5 memory, the 4U chassis sports 24 hot-plug LFF/SFF drive bays along with dual internal M.2 PCIe Gen5 slots for NVMe SSDs. True, you don't get the extra rear SFF bays offered by the TS-hx87XU-RP range, but the TS-h2477AXU-RP has big expansion plans as it can handle eight of Qnap's 24-bay 4U TL-R2400PES-RP JBOD shelves for a total of 216 drives and up to 5PB of raw storage.

Networking options are plentiful as the appliance presents dual embedded 2.5GbE and 10GbE RJ45 ports, which are all the multi-Gigabit variety. Three spare PCIe Gen4 expansion slots mean there's plenty of room for more, and Qnap offers a wide choice of its own certified 2.5GbE, 10GbE, and 25GbE network adapters.

Qnap TS-h2477AXU-RP review: Design and deployment

The appliance is clothed in a sturdy steel chassis with solid metal drive carriers, and the price includes dual 800W hot-plug PSUs. Popping the lid reveals the motherboard lurking in the base, and closer inspection shows it is the same as that used by Qnap's TS-h3077AFU all-Flash array, but with the two rear SSD cages removed and the dual M.2 NVMe SSD slots added.

The Ryzen CPU is mounted by a large passive heatsink, and cooling is handled by a bank of four cold-swap fans mounted behind the drive backplane. Four long DIMM slots sit to the right, with one occupied by the base 32GB of memory, leaving room to upgrade to 128GB without having to replace it.

The 'h' in the model name clearly shows the preferred operating system is Qnap's QuTS Hero – you can load the nimbler QTS software during the installation phase, but you'll be missing out on a wealth of data protection and storage reduction features. QuTS hero offers end-to-end checksums for transparent data corruption recovery, triple parity RAID and triple mirroring, WORM (write once read many) policies to protect NAS shares from tampering, and inline data deduplication, which requires a minimum of 16GB to function.

Qnap doesn't enforce the same drive restriction policies as Synology, and we had no problems using two Kingston 480GB NVMe SSDs for system pool duties and four 22TB Western Digital Red Pro NAS HDDs for general data storage. The web-based deployment wizard installs QuTS hero, and we then use the Storage & Snapshots app to create a mirrored array for the system pool and a big RAID5 array for our data.

Qnap TS-h2477AXU-RP review: Data protection

QuTS hero offers ZFS copy-on-write for fast, almost unlimited snapshots of NAS shares and iSCSI LUNs, which can be scheduled to run as often as every minute for near-continuous protection. NAS shares and iSCSI LUNs can be encrypted, and the SnapSync app provides fast replication to remote appliances using scheduled and real-time block-level snapshots.

When creating NAS shares, you can opt to apply one of two WORM policies to protect them from ransomware attacks, and once assigned, they cannot be modified or disabled. The Enterprise policy allows a protected folder to be deleted, but stops anything in it from being individually modified or removed, while the stronger Compliance policy blocks folder deletion and only allows it to be removed by deleting the entire storage pool.

There are plenty of free backup apps available with the Hybrid Backup Sync (HBS) 3, managing 3-2-1 protection strategies and securing local appliance data to remote NAS appliances or cloud providers, including Qnap's myQNAPcloud service. The Hyper Data Protector (HDP) app secures VMware and Hyper-V virtualized environments, presents a dedicated management web console, and can run block-level backups of Windows PCs and servers that have Qnap's free NetBak PC agent installed.

During testing, we updated the appliance with the latest QuTS hero 5.3, which introduces plenty of new features, including a high availability manager that allows two identical appliances to be placed in a fault-tolerant active/passive cluster. The Control Panel, iSCSI & Fibre Channel, and Storage Manager apps see completely redesigned interfaces, while snapshot management has been moved to a dedicated app, which offers protection policies for creating immutable snapshots.

Qnap TS-h2477AXU-RP review: 10GbE performance

For performance testing, we used the appliance's embedded 10GbE ports and hooked it up to the lab's Dell PowerEdge R760xs Xeon Scalable Windows Server 2022 host. With a NAS share mapped to the server, we recorded Iometer sequential read and write rates both of 9.2Gbits/sec, and swapping to random operations returned 9.2Gbits/sec and 9.1Gbits/sec.

IP SAN speeds are equally good, with a 1TB target delivering the same sequential and random rates as our NAS share. We increased the pressure with a dual 10GbE MPIO link to the target and watched sequential read and write speeds increase to 18.5Gbits/sec and 17.4Gbits/sec.

Swapping to random read and write operations delivered 18.3Gbit/sec and 16.5Gbits/sec, and we noted for the latter test, the Ryzen CPU never went beyond 17% utilization. These numbers also highlight the efficiency of the QuTS hero ZFS ARC (adaptive read cache) and ZIL (ZFS intent log) as they negate the need for SSD caches to improve random operations.

Qnap TS-h2477AXU-RP review: Is it worth it?

Yes, it is, with Broadbandbuyer confirming to us an end-user price for a diskless model of £6,635 excluding VAT. Qnap's 16-bay TS-h1677AXU-RP model comes in £5,378, so the TS-h2477AXU-RP offers a lower cost per drive bay.

The TS-h2477AXU-RP is a great choice for businesses with big capacity demands, as it offers a huge storage potential, and Qnap's lack of drive restrictions means you can shop around for your storage media. The appliance performed very well in our 10GbE lab test,s and the latest QuTS hero software is packed with plenty of business apps and smart data protection features.

Qnap TS-h2477AXU-RP specifications

Dropbox has announced a raft of new AI-powered features for users, including integration of its Dash AI assistant in the storage platform.

As part of the move, Dropbox said the Dash AI assistant will offer users “smarter search, intelligent organization, time-saving summaries, and contextual answers” for files hosted across the storage service.

The AI tool was first launched in 2023 as a standalone application, allowing users to search for files both within Dropbox and in third-party applications such as Slack, Google Workspace, or Jira.

So far, Dash has been restricted to selected customers. However, Dropbox said wider availability is “expected in the coming months”.

While an exact timeline on general availability has yet to be confirmed, the move marks a significant step toward full integration of AI capabilities in Dropbox.

So what can customers expect from the update?

Dropbox eyes intelligent storage search

Dropbox said Dash will improve ease of use and visibility of content and files stored across the service. Underpinning this is a natural language search function that lets users essentially describe the files and content they’re looking for.

Example queries touted by the company include insights on a company “content calendar for the Fall launch” or “what’s new in the latest version of the client proposal?”.

Thereafter, the AI assistant can provide document summaries, highlight specific data points within files related to the aforementioned prompts, or create a single overview of files stored across multiple locations.

“Because Dash in Dropbox learns from all the content you’ve already stored, you get powerful AI features within the familiar Dropbox experience you already know and trust,” the company said in a blog post.

“And the more you use Dropbox, the better Dash understands you and your team. The result: sharper search results, more relevant answers, and actually useful insights that help move work forward.”

Easy setup and multimodal features on the horizon

Integration of Dash within Dropbox will make the tool far more accessible for users, the company noted. As it was previously a standalone platform, customers were required to contact the company directly to request early access.

Users are now able to skip that step, however, and integration of the assistant takes just a few minutes.

“You don’t need to talk to our sales team to get started, and you can set everything up yourself in minutes,” Dropbox said. “Dash takes your scattered, siloed content and gives you a single source of truth, turning information overload into organized clarity.”

Alongside the announcement, Dropbox also confirmed plans to acquire AI startup Mobius Labs in a bid to ramp up multimodal capabilities in the storage platform.

Mobius Labs will help enhance in-app search features across a wider array of file formats, spanning audio, video, or images.

“The Mobius Labs team has been developing custom AI models optimized for large-scale multimedia processing,” the company said. “We'll soon be able to support even more complex workflows for search and answers in Dash.”

MORE FROM ITPRO

• The best secure cloud storage options for your business
• Our pick of the best cloud file sharing services
• Choosing the best cloud storage options for your business

OpenAI has announced plans to begin offering data residency for UK businesses following a landmark deal with the government.

Under the terms of the agreement, ChatGPT customers can now store their API Platform, ChatGPT Enterprise, and ChatGPT Edu content in the UK.

This data includes conversations – spanning text, image, and voice – code interpreter and data analysis artifacts.

Similarly, files such as uploaded images and documents, custom GPTs, and associated prompts/outputs will be covered by the data residency scheme.

"With UK data residency, new public sector partnerships, and our ongoing work through the MOU, we’re helping lay the foundation for trusted and secure AI adoption in the UK," OpenAI said in a statement confirming the move.

"We look forward to working with the government to help more British people, businesses, and institutions benefit from AI."

The government said the deal will not only enhance privacy and accountability but also reinforce national resilience in the face of growing global cyber threats, as well as unlocking further investment from businesses.

OpenAI move follows successful pilot scheme

The announcement came alongside the news that ChatGPT Enterprise and ChatGPT Edu are to be rolled out to 2,500 Ministry of Justice staff – taking advantage of UK data residency.

This follows a pilot that showed time was saved across a range of routine tasks including writing support, compliance and legal work, data and research processes, and document analysis.

“The number of people using our products in the UK has increased fourfold in the past year. It’s exciting to see them using AI to save time, increase productivity, and get more done," said OpenAI CEO Sam Altman.

"Civil servants are using ChatGPT to improve public services and established firms are reimagining operations. We're proud to continue supporting the UK and the Government's AI plan.”

ChatGPT is already powering some tools used by public sector workers across the UK, including Humphrey, an AI assistant designed to help civil servants speed up admin tasks in areas such as planning and social care, as well as in central government.

Meanwhile, the government's Consult tool, also powered by ChatGPT, automatically sorts public consultation responses, cutting the time taken from weeks to minutes.

"We’re making sure we have the infrastructure we need to power AI here in Britain so we can transform our public services, drive growth and unlock new opportunities for every community in the country. This partnership with OpenAI is a step further in delivering that," said technology Secretary Liz Kendall.

"By using AI across our public services we’re giving key workers like probation officers and NHS staff more time to focus on delivering better outcomes, reducing time spent on admin and speeding up vital services for people."

MORE FROM ITPRO

• OpenAI just revealed what people really use ChatGPT for
• Everything you need to know about OpenAI's new open weight AI models
• OpenAI has a bold plan to pay for its $1 trillion spending spree

Data storage specialist Seagate has announced a new strategic partnership with cybersecurity vendor Acronis to help MSPs meet modern AI-driven storage needs.

The alliance will focus on equipping MSPs and global enterprises with secure and scalable archival storage solutions as AI adoption continues to drive rapid data growth across industries.

The two companies will work together to deliver Acronis Archival Storage, integrated with the Acronis platform, which is designed for infrequently accessed data and offers long-term, secure, and S3-combatible storage under a predictable cost model.

This collaboration will see Acronis incorporate Seagate’s S3-compatible Lyve Cloud Object Storage platform into its offerings, bringing a host of enterprise-grade security features such as encryption of data in-transit and at-rest, role-based access control, identity and access management, and immutability options.

In an announcement, Seagate said this combination will enable MSPs to provide customers with “seamless, cost-effective, and compliant storage” capable of handling the demands of modern AI-driven environments.

“Acronis has been at the forefront of delivering innovative cybersecurity and backup solutions, including secure data management capabilities for hybrid cloud environments,” commented Melyssa Banda, SVP of Edge Storage and Services at Seagate.

“By joining forces, Seagate and Acronis can leverage our combined global reach to deliver the highly secure storage solutions that the new frontier of AI requires—empowering enterprises to better protect their data from cloud to edge to endpoint.”

The pair said the new combined solution focuses on the needs of MSPs and IT leaders in highly regulated sectors such as healthcare, finance, legal, utilities, and public administration.

MSPs can leverage the offering to address the needs of organizations that are required to retain large volumes of infrequently accessed data for years at a time, while ensuring and maintaining regulatory compliance.

Commenting on the partnership, Acronis president Gaidar Magdanurov said Seagate’s expertise in mass-capacity storage “perfectly aligns” with Acronis’ goal to offer MSPs solutions that prioritize security, compliance, and cost control.

"With over 60% of organizations managing over 1 PB of data, archival storage is a necessary solution to keep storage costs manageable,” he explained.

“Our partners will be able to address the rapidly growing demand for long-term storage of data, and address compliance, legal, and security requirements of their customers."

MORE FROM CHANNELPRO

• Schneider Electric names Karlton Gray as new UK&I channel director
• Check Point buys Lakera to secure the full enterprise AI lifecycle
• inforcer named as Microsoft partner for new AI-focused MSP initiative

In a world where many folk will baulk at even installing their operating system on a PC, asking them to also undertake some physical assembly may seem a foolhardy idea. The small amount of basic work you need to undertake with a screwdriver should, however, not deter you from buying the new Framework Desktop PC.

Framework is not a new player in the PC market, though its previous products have been laptops rather than desktops. While the new Framework Desktop is quite a different beast, you can still use the same slide-in modular expansion cards that feature on its current laptop models.

You can order your Framework PC with either Windows 10 Home or Pro, or as a barebones system without an operating system. Either way, you will need to install your OS, as the Windows option simply means you get an activation code from Framework.

We loaded our review machine with Ubuntu 20.04 and Windows 11 Pro in a dual-boot setup, even though Framework recommends Fedora and Bazzite as officially supported Linux distros.

It also recommends Arch, NixOS, and Bluefin as community-supported distros, but the fact that Ubuntu loaded and worked with no issues suggests most versions of Linux will run without any problem.

Framework Desktop: Design

Once you've unboxed the various components that make up the Framework PC package, you are left with a 4.5L metal PC tower, which already includes the motherboard, heat sync, power supply, and all the other essential gubbins, as well as the bits you need to install yourself and a Torx screwdriver.

These include a fan (from a choice of three), one or two M.2 2280 SSDs, assuming you've specified them, clip-on plastic side panels for the tower, and various decorative plastic tiles to clip into the magnetic front panel.

You also receive any modular expansion ports you've ordered, which slide into the two docking bays at the bottom of the tower. They are the same as the ones you'll find on Framework's laptops. In our case, this was a selection of USB-C and USB-A expansion cards.

The assembly part of the process involves installing the SSD and the fan, then the side panel, the two expansion ports, and the decorative tiles. Framework reckons this all takes 10 minutes, but we timed ourselves and got it done in under 8. It took longer to unbox all the parts and clear the mess away.

If you want a get a better taste of what is required in the assembly process, Framework provides a selection of foolproof guides to putting the tower together and installing your software.

While assembling the unit, it became clear that Framework hasn't skimped on the quality. Everything went together smoothly and securely, all the clips and screws felt well-made, and the result felt reassuringly stiff and sturdy.

Adding a second SSD takes even less time: all you need to do is slide out the plastic panel that covers the main board and screw it into place. The Framework can take SSDs up to 8TB in capacity, making it a great choice if you want to store very large quantities of data.

Framework offers a wide choice of expansion cards for the two slots at the bottom of the unit. You can choose from 10Gbps USB-C or USB-A I/O ports, MicroSD or SD card readers, 2.5Gbit Ethernet, and 3.5mm audio jacks and storage cards with either 256GB or 1TB capacity. The cards vary in price from £9 for the USB-C/A cards up to £39 for the Ethernet card and £45/£125 for the storage modules.

The two expansion slots on the front of the tower are in addition to a wide selection of fixed ports at the back, namely two USB-C 4 and two USB-A 3.2 Gen 1 ports, two DisplayPort 2.1 and one HDMI 2.1 video outputs, 5Gbit Ethernet, and 3.5mm audio jacks.

All together, that makes the Framework Desktop one of the best-connected small-form PC we've ever seen here at ITPro. Wireless communications are handled by a Realtek RZ717 card supporting the latest Wi-Fi 7 and Bluetooth 5.4 standards.

The slots for the expansion cards are at the very bottom of the case, and this is the only small problem with the design because it puts the ports only a few millimetres above the surface the Framework is resting on, which isn't ideal for ease of access, especially for bulky USB thumb drives.

If you have to pick the box up to use the expansion ports, you'll thank the optional screw-on carry handle that fixes to the top of the tower. It's not strictly necessary, given that the Framework PC only weighs 3.1Kg, but it does make grabbing it easier.

The magnetic front panel can be decorated with 21 plastic tiles. These are available in packs of 7 coloured tiles for £15, so you have to spend £45 just to fill all the spaces. For £5 each, you can buy individual tiles with various decorations such as the AMD, Tux, Framework, and Linux Mint logos. Of course, nothing is stopping you from making your own tiles if you have a 3D printer to hand.

Given that the unit looks half-finished without a full set of tiles to fill the front panel, and the £45 cost for something purely cosmetic, perhaps Framework could have considered also including a simple plastic cover for the front?

Framework Desktop: Specs and Performance

The Framework Desktop is the first PC we've encountered running on AMD's new Halo Strix Max CPUs. You can have your Framework Desktop with either an 8-core AMD Ryzen AI Max 385 and 32GB of RAM or the 16-core Max+ 395 CPU with 64GB or 128GB (the RAM is integrated, so there's no option to add more post-purchase), along with a 32-core Radeon 8050S or 40-core 8060S GPU. Our test unit came with the basic 385 CPU, but even that proved to be surprisingly powerful.

In ITPro's in-house 4K multi-media benchmark, the Framework scored a blistering 510 points, while the PCMark10 test score came in at 8,856. If you want a small-form PC with meaningfully higher levels of performance, you'll need to spend twice as much or more on something like a M4 Max Mac Studio.

To put the Framework's performance into context, the new Core Ultra 288V-powered MSI NUC AI+ 2MG scored 301 points in the 4K test, while the Core Ultra 9 285H-powered Geekom IT15 managed 361.

In the PCMark10 benchmark, the NUC AI+ 2MG scored 6,857 and the Geekom box 7,450 points. The result is clear: when it comes to demanding workloads, the new Framework Desktop is in a league of its own at this price point.

Graphics performance was, if anything, even more impressive, with the SPECviewperf 3dsmax 3D modelling test running at 118fps, which is close to five times as fast as the 25fps you'll get from one of Intel's latest Arc integrated GPUs.

The Framework Desktop is not being pitched as a gaming platform, but it seemed churlish not to test its gaming chops. The Cyberpunk 2077 benchtest ran at 73fps at FullHD in the Ray Tracing: Medium preset. That's very impressive.

In more general terms, the Ryzen AI Max 385 chipset delivers GPU performance that is roughly comparable to an Nvidia 4060/4070 GPU with an 80-100 watt TGP. That's an order of magnitude more powerful than any iGPU we've come across, including the likes of AMD's 890M and Intel's Arc 140T.

Perhaps even more impressive than the outright performance is the lack of fuss with which the Framework Desktop can handle even the most demanding tasks. During our standard stress test, when we ran the CPU and GPU flat out, the system was still running at 100% utilization after three hours, and the fan noise was very subdued.

The 1TB Western Digital Black SSD in our review unit proved to be a strong performer, recording average sequential read and write speeds of 5,020 MB/s and 3,240MB/s, respectively.

Framework Desktop: Is it worth it?

The system we were sent to test with a single Western Digital Black 1TB SSD, the basic fan, and two expansion cards, but no Windows licence will set you back £1,226. A Windows activation licence will cost a further £120 for Home or £220 for Pro if bought from Framework, but many of you reading this may already have access to a corporate or educational Windows licence.

With or without a Windows licence, the Framework Desktop certainly offers a serious amount of performance for the asking price. Indeed, it's hard to argue with the basic offer of a small form PC with power enough to run even the most demanding graphics jobs quickly and demanding 3D games for under £1,500.

The modular expansion ports are perhaps not as important in the Desktop as they are in Framework's laptops because even without them, the Desktop still has a good selection of ports, but they are still a useful feature.

We'll leave it to the reader to decide if the decorative tiles are essential parts of the Framework Desktop concept or more of an aesthetic (and expensive) indulgence, but given the price, performance, and quality of the overall package, we like that the Framework Desktop doesn't look like just another boring black boxy small-form PC.

Framework Desktop Specifications

The US government is reportedly weighing up plans to take a stake in Intel as the chip giant’s troubles continue.

According to reports from Bloomberg, the proposed deal aims to support the company's efforts to expand manufacturing operations across the US.

A key focus here would reportedly be ramping up development of its delayed Ohio manufacturing facility. While initially touted to open this year, the company has repeatedly pushed back the launch date.

While first delayed until late 2026, in February this year Intel admitted the site wouldn’t become operational until 2030-2031.

The reports from Bloomberg come in the wake of sharp criticism from the US administration. Last week, President Donald Trump suggested CEO Lip-Bu Tan resign amid claims over alleged ties to China.

Subsequent talks between Tan and the administration appear to have been productive, with President Trump later hailing the meeting in a post on Truth Social.

"His success and rise is an amazing story," he wrote.

ITPro has approached Intel for comment, but received no response by time of publication.

In a statement given to Bloomberg, a spokesperson for Intel said the company is “deeply committed to supporting President Trump’s efforts to strengthen US technology and manufacturing leadership”.

“We look forward to continuing our work with the Trump Administration to advance these shared priorities, but we are not going to comment on rumors or speculation,” the spokesperson added.

Intel’s long-running woes

Intel has experienced significant troubles in recent years, falling behind competitors such as AMD and, notably, Nvidia, since the advent of the generative AI race. Intel's market value, for example, has halved since 2020.

The company has gone through repeated rounds of layoffs and a restructuring effort. Last month, the company announced plans to cut its global workforce by 15% and cancelled plans to develop new chip factories in Poland and Germany.

Tan’s appointment as CEO earlier this year was viewed at the time as an effort to steady the ship following Pat Gelsinger’s retirement. In the wake of the appointment, reports from Reuters detailed plans for an overhaul of the company’s chip and AI strategies in a bid to breathe new life into the company and reverse sluggish revenue growth.

Any involvement from the Trump administration could represent a lifeline for the semiconductor giant, and the White House has made clear its plans to ramp up domestic manufacturing and support for fellow chip giants in recent months.

Earlier this week, Nvidia and AMD announced an agreement with the administration to pay 15% of revenues from Chinese chip sales.

MORE FROM ITPRO

• Intel makes high-level hires while factory workers are warned of layoffs
• Intel layoffs confirmed as CEO eyes 'sustainable growth'
• Intel needs to “get its story right” to turn things around and capitalize on the AI boom

It has taken a good six months for the first Core Ultra Series 2 mini PC boxes to arrive, with a new model just launched and sent to us for review by Geekom – the Geekmom IT15 AI. Have Arrow Lake mini PC's been worth the wait? Are they worth the extra cost over their roughly equivalent Meteor Lake brethren?

At the time of writing, the Geekmom IT15 AI is available via Amazon in two flavours, both packing 32GB of RAM. One features the Core Ultra 5 225H CPU with a 1TB SSD and costs £687 inc VAT (£572 ex VAT) or $679, while the other is based on the Core Ultra 9 285H chip and has 2TB of storage and sells for £845 inc VAT (£704 ex VAT) or $949. Geekmom has given us a promo code of ITPROIT15, which should get UK buyers a 10% discount until August 31st.

To put those prices into context, the Geekom Mini IT13 2025 Edition, which features a Core i9-13900HK CPU, 32GB of RAM and a 1TB SSD is currently available to buy for £616 inc VAT (£513 ex VAT) which makes it around £150 cheaper than the new IT15 when you factor in the cost of an extra 1TB of storage.

Geekom IT15 AI: Design

Given the pretty hefty price, you'd expect a high level of build quality, and Geekmon has certainly delivered here. The whole unit is made from stiff metal sheeting and feels very solid and robust. The matte-finish black paint job lends the unit an impressive degree of understated elegance.

Given that it's built around such a powerful processor, the IT15 is a very petite device measuring up at just 117 x 112 x 45.5mm and weighing a piffling 560g. This makes it an ideal candidate to be hidden away behind a monitor, and indeed Geekom bundles a VESA bracket and screws for just this use.

Given the Liliputian dimensions, the designers have done a good job of shoehorning in a wide range of I/O ports. On the front panel next to the power button, you'll find a 3.5mm audio jack and two 10Gbps USB-A ports.

On the rear panel, there are two 40Gbps USB-C ports, both supporting DisplayPort video, as well as a 2.5GbE RJ-45 jack, 10Gbps and 480Mbps USB-A ports, two HDMI 2.1 video outputs, and the DC-in jack.

One of the USB-C ports also supports Power Delivery, which is useful if you want to leave the unusually small and light 120W power brick at home and use a phone-friendly Type-C power supply. The 100W USB-C charger I had to hand powered the IT15 perfectly. Rounding things out, there's an SD 4.0 memory card slot on the left side and a Kensington lock on the right.

Our only comment on the ports is that, given the price and the target market, we'd have preferred to see a standard DisplayPort video output, even if that was in place of one of the HDMI ports.

As you'd expect, getting inside the Geekom IT15 AI is a very easy operation. Just slacken off the four Philips screws that sit inside the rubber feet at each corner, and the base plate comes away cleanly.

Once in, there is very easy access to the two SODIMM mounts and the two SSD emplacements, one for an M.2 PCIe 4 2280 drive, the other for a SATA 3 2242 drive. Welded to the inside of the base plate is a bracket for a 2.5in HDD, though Geekom doesn't bundle any of the cables to connect said to the motherboard or the HDD LED power supply.

The Wi-Fi card is buried deeper inside the assembly, but given that it's a bang-up-to-date Intel BE200 unit that supports Wi-Fi 7 and Bluetooth 5.4, it's an unlikely candidate for replacement.

Geekom IT15 AI: Specs and Performance

The CPU inside the Geekmon IT15 AI is an Intel Arrow Lake Core Ultra 9 285H with 16 cores, of which 6 are rated Performance, a maximum turbo frequency of 5.4GHz, and a 65W TDP. That combines with 32GB of 5600MHz DDR5 RAM to deliver a very useful level of performance.

Indeed, the chipset is considerably more potent than the Meteor Lake Core i9-13900HK-based competition used in the likes of the Beelink SEi13 and Geekom Mini IT13, as demonstrated by its impressive score of 361 in our 4K multimedia benchmark. For comparison, the Beelink SEi13 scored 272 and the Geekom Mini IT13, 262.

The new Geekmon box also performed well in the PCMark 10 benchmark, scoring 7,450 points, comfortably ahead of the 4,945 scored by its junior sibling, the Ryzen 5-powered Geekmon A5, or the 6,066 scored by the Lenovo ThinkCentre M70q Gen 4 Tiny, which runs on an Intel Core i7-13700T CPU.

When it comes to plain CPU performance, the difference is less marked but still clear: In the Cinebench R23 multi-core test, the IT15 scored 15,524 points, compared to the Beelink box's 14,199, Lenovo's 12,914, and the Geekom Mini IT13's 12,127.

Turning to graphics performance, the new Geekom box shows the Meteor Lake competition a very clean pair of heels, running the SPECviewperf 3dsmax 3D modelling benchmark at an impressive 29fps, twice as fast as the Irix Xe iGPU that comes as part of Intel's Meteor Lake chipsets.

If you want a mini PC for any sort of graphics work, but can't stretch to a box with a discrete GPU, the IT15 is certainly worth considering.

The one drawback with an Arrow Lake CPU is that the associated NPU is far less powerful than the one you get with a Lunar Lake chip, rated at only 13 TOPS rather than 48. This means that despite the 285H being a new CPU, you'll miss out on certain of Microsoft's Copilot+ features like Recall, which takes regular snapshots of your on-screen activity for later reference.

Of course, given the security and privacy concerns that accompanied the initial announcement of Recall and the subsequent changes made before its eventual rollout by Microsoft, the absence of this feature may not be regarded as an altogether bad thing by many commercial users.

The Geekom's single cooling fan goes about its business quietly and without drama, but isn't quite up to the job of keeping the CPU fully cooled when running hard for prolonged amounts of time. Under heavy stress, the core temperature quickly peaked at 90 degrees, which cut the CPU utilisation to just 40 per cent.

After around five minutes, the CPU and the cooling system reached a mutual understanding, and utilisation jumped back to around 60%, where it remained for the duration of the stress test. This is hardly a deal-breaking problem, but it does just take the edge off the CPU's theoretical maximum performance level.

Our review unit came fitted with a 2TB Crucial SSD, which proved to be an impressive performer, returning sequential read and write speeds of 5,211MB/s and 4,409MB/s, respectively.

The Geekom IT15 comes loaded with Windows 11 Pro, and there's no option to buy without an operating system. As is our want, we booted it up into Ubuntu 24.10 and found that everything worked as it should.

Geekom IT15 AI: Is it worth it?

The Geekom IT15 is one of the more expensive mini PCs that we've tested, be it from a niche Chinese OEM or a more established manufacturer like Asus or Lenovo. That said, it's also one of the most powerful and one of the smallest mini PCs we've tested. Indeed, it's a little bigger than the likes of the budget N150-powered Acemagic Vista V1.

Granted, Geekom's efforts to squeeze a full pint of PC into a half-pint container may have slightly compromised the cooling system under extremis, but there's no arguing with the impressive achievement of fitting such a potent chipset as well as a wide range of I/O ports, three internal storage bays and a brace of SODIMM modules into such a wee machine.

Given all that, plus the fact you get a brace of USB-C 4.0 ports, the very latest wireless communication standards, a 2TB SSD, and 32GB of RAM for your money, it's difficult not to conclude that the Geekom IT15 represents solid value for money when placed alongside the Meteor Lake alternatives.

Geekom IT15 AI Specs

The Chinese-made mini PC market is becoming increasingly crowded with similar machines, so manufacturers are seeking ever more subtle ways to differentiate their products from those of their competitors.

Beelink's pitch for its new SEi13 Pro mini PC is that it has built-in speakers and a microphone, something you seldom find in mini PCs. Needless to say, Beelink highlights the benefit this will have with interacting with AI models, though we suspect that Teams and Zoom calls will be a more common use.

Beyond the audio system, the SEi13 Pro is pretty standard fare, built around an Intel Core i9-13900HK with 32GB of RAM and a usefully wide range of ports. It's currently available for £669 (£582 ex VAT) in the UK, but $539 in the USA, which makes it very close in price to the Geekom Mini IT13, which uses the same CPU.

Beelink SEi13 Pro: Design

As you'd expect, given the current design trend for small-box PCs coming out of the Shenzhen region, the SEi13 Pro is a small metal cube with rounded corners. At 135 x 135 x 44.7mm and 780g, it's rather larger and heavier than the Geekom machine, the extra size needed to accommodate the speaker module, the extra weight taken up by the CPU and SSD heat sinks.

The SEi13 Pro doesn't come with any sort of VESA attachment or bracket, so there's no easy way of attaching it you the rear of a monitor. The PSU is external, but the 120W rat-and-tail affair is small enough, and the power cable long enough, for it to be easily tucked away.

The front panel of the SEi 13 Pro is home to an on/off switch, two 10Gbps USB ports, one Type-C (data only) and one Type-A, a 3.5mm audio jack, a CLR CMOS pinhole, and four small apertures that cover the microphone array.

At the rear, there are HDMI 2.1 and DisplayPort 1.4 ports, two 10Gbps USB-A ports, a second 10Gpbs Type-C port that also supports DisplayPort video, a USB-A 2.0 port, 2.5GbE Ethernet socket, and the barrel jack for the PSU.

The rear panel also houses a second 3.5mm audio combo jack. If both jacks are occupied, the one at the front takes priority, and the rear one is temporarily disabled.

Compared to the Geekmon Mini IT13, the new Beelink box takes with one hand, but gives with the other. It's rather disappointing not to have at least one 40Gbps Type-C port, but on the other hand, we were happy to see a full-size DisplayPort output.

Given that the SEi13 Pro is larger than the Geekom machine, the ports are not as closely spaced, which makes using them rather easier. It's a shame that Beelink couldn't find space for a memory card slot of some description.

To open up the SEi13 Pro you need to pop out four plastic plugs cover the screw holes and then undo said screws. Once the base is off, you need to undo two more screws and remove a metal mesh frame which has the stereo speaker module attached to it.

The ribbon cable that connects the speakers to the motherboard is just long enough to enable you to work without having to disconnect anything. That's just as well because to access the two M.2 2280 SSD bays, you also need to unscrew and remove a heatsink. Unlike some mini PCs, there is no room for a 2.5-inch HDD in the SEi13 Pro.

The internal design of the SEi13 means that the two speakers face downwards through the two grates that form the base of the unit. This hampers the sound quality of a system we suspect doesn't sound great even without said interference, and when not facing the floor.

Beelink doesn't give the speaker module a power rating, but judging by the limited volume – just 70.5dBA when measured against a pink noise source at a distance of 1m – it's not an awful lot.

The sound it makes is clear, detailed, and devoid of distortion even at maximum volume, but there's not much in the way of bass and even less stereo separation. For voice calls and Teams meetings, the sound system does just fine, but you won't want to use it to listen to music for an extended period of time. We've heard worse from laptop speakers, but we've also heard much better.

According to Beelink, the microphone array is backed by an AI chip to improve voice recognition and reduce background noise. Certainly, we found no issues with speech recognition, which was impressively accurate while office background hubbub was well suppressed.

Beelink SEi13 Pro: Specs and Performance

The Core i9-13900HK chip inside the Beelink SEi13 Pro is a 14-core, 20-thread mobile chip, with a 65W maximum TDP, and it's a strong performer with 6 Performance cores and a maximum turbo clock of 5.4GHz.

The memory is soldered in place, but given that you get 32GB of dual-channel LPDDR5 6000MHz RAM as standard rather than the Geekom's DDR4-3200, that's something we imagine most potential buyers will be satisfied with.

Wireless communications are handled by an Intel AX200 card. That's a component that's been knocking about since 2019, which means support for Bluetooth 5.2 and 2.4/5Ghz Wi-Fi 6 rather than 6Ghz Wi-Fi 6E.

The fast RAM gave the Beelink box a small performance advantage in our 4K multimedia benchmark, where it scored 272 points to the Geekom Mini IT13's 262. The same proved to be true in the Cinebench R23 multi-core test where the Beelink had the Geekom beaten by 14,199 to 12,127.

Those are not dramatic differences in performance, but the SEi13 Pro has an ace up its sleeve: The presence of two quite large heat sinks. These pay dividends because after an hour of stress testing, the CPU was still running at 90% and the fan was spinning very quietly. The Geekom Mini IT13 ran louder under the same conditions, and CPU utilisation dropped to 60%.

The Xe graphics integrated GPU is, as always, a weak spot when compared to the latest Arc iGPUs from Intel. To put that comment into context, the SPECviewperf 3dsmax 3D modelling benchmark ran at a pedestrian 16.5 fps, marginally faster the the Geekom boxes' 15fps but way off the sort of speeds you can expect from Intel's Arc 140V iGPU, which can hit 25fps or higher.

The Micron SSD in our review box performed well, returning sequential read and write speeds of 4,225MB/s and 4,103MB/s, respectively. In this metric, the new Beelink box has the Geekom beaten like a rented mule.

The Mini IT13 comes loaded with Windows 11 Pro, but we found that everything worked perfectly well under Ubuntu 24.10, making this a good platform for anyone wanting to switch to an open-source OS. As with the Geekom machine, there's no option to buy the box without a preinstalled operating system.

Beelink SEi13 Pro: Is it worth it?

If you want a mini PC with a built-in sound system, then there's not a whole lot of competition, so the Beelink has a definite selling point there. On the other hand, a cheap set of external speakers or a soundbar will sound much, much better and hardly break the bank

The sound system aside, we like the fact that Beelink has given the SEi13 Pro a full-sized DisplayPort video output, fast, albeit fixed, RAM, and that it can be run hard without any sign of thermal throttling or excessive noise. For the price, though, we'd have liked to see a 40Gbps data port, 6GHz Wi-Fi, and a memory card slot.

That makes it a photo finish with the similarly priced Geekom Mini IT13, which does have those faster data transfer speeds and a memory card slot, but struggles to maintain its performance under stress, lacks a sound system, and doesn't have a DisplayPort video output. Neither box is likely to disappoint; it's simply a matter of choosing the features that matter most.

Beelink SEi13 Pro Specifications

The HP OmniStudio X is HP's answer to the iMac, only bigger, brawnier, and aimed squarely at creative professionals who want more screen real estate than Apple's flagship all-in-one offers.

HP claims the OmniStudio X, as its name suggests, is a studio-grade machine that's designed not just to replace your PC, but your entire creative setup. The model we tested features a 32-inch 4K-resolution display with an ergonomic, height-adjustable stand, an Intel Core Ultra processor paired with Nvidia RTX graphics, and a minimalist aesthetic that clearly indicates HP's pursuit of the same design-conscious crown that Apple has long held.

Unsurprisingly, this all comes at a price. The Windows all-in-one starts at £1,699 in the UK, putting it firmly in premium territory, especially when compared to more budget-friendly desktop setups. So, is the HP OmniStudio X just another flashy all-in-one, or is it a genuine contender for the creative desktop accolade?

HP OmniStudio X: Design

HP's OmniStudio X 32-inch makes a strong first impression. The all-in-one oozes premium appeal, from its expansive 4K screen framed by slender bezels to the solid aluminium stand that gives it an almost sculptural-like aesthetic. Unlike many all-in-ones, there's no bulky base or clunky plastic in sight. Instead, HP has opted for clean lines, minimal branding, and an understated matte finish – a far cry from Apple's multicoloured iMac lineup.

Setting up the OmniStudio X was easy, too. Once out of the box, you simply screw the base to the stand, and attach the latter to the monitor. It's worth noting that the 31.5-inch display is quite large and heavy, so you may need two hands for the initial setup, although no additional tools are required. The sheer heft of the OmniStudio X also means you'll need a sizable desk to accommodate it; this isn't a machine designed for small boxroom offices.

You'll also find a matching keyboard and mouse inside the box, which automatically pair with the desktop once two AAA batteries have been inserted.

One of the best things about the design of the HP OmniStudio X is that it comes loaded with ports; three USB port – one Type C and two Type A – can be found on the stand, and the back of the monitor comes equipped with three more USB ports, a HDMI-in port for using the panel as an external display, a HDMI-out port, an audio-jack, a 2.5Gbps ethernet port, and a full-size power port.

Another net design trick is the pop-up 4K webcam that neatly retracts back into the monitor when not in use, ideal for the privacy-conscious and for keeping the machine looking sleek.

HP OmniStudio X: Display

As we've mentioned, the HP OmniStudio X's display is huge. We may have struggled with our small office setup, but for those looking for a statement screen for creative work or multi-window multitasking, the 4K panel offers a ton of space – and then some.

It delivers quality-wise, too. The 3840 x 2160 resolution panel is exceptionally bright, offering richly saturated colours and deep contrast – unsurprising, given it's primarily designed for creative professionals editing in Photoshop and juggling multiple windows at once. The IPS panel also offers excellent viewing angles with minimal colour shift, even when viewed from the side.

While it's one of the best all-in-one displays we've tested, it's perhaps not the best for gaming. It's limited to a 60Hz refresh rate, which means that while it can handle some light after-work gaming, it's not designed for top-tier titles or extended gaming sessions.

HP OmniStudio X: Keyboard, mouse and webcam

As mentioned, the HP OmniStudio X ships with a bundled keyboard and mouse, which match the all-in-one's overall aesthetic.

We were impressed with the keyboard, which has a premium, cool feel like the Magic Keyboard that ships with Apple's iMac. Though it can't be adjusted beyond its built-in tilt, the low-profile keys are surprisingly responsive for such a slim peripheral. We found typing quiet, feedback tactile, and the key spacing generous, which helps reduce typos. The keyboard also features a full-size numpad, ideal for productivity tasks, and a variety of handy shortcut keys, including search, paste, and screen capture.

There's also a built-in Copilot button that fires up Microsoft's AI assistant and lets you generate images and summarise text.

Unfortunately, the mouse bundled with the HP OmniStudio X sadly doesn't live up to the premium standard set by the all-in-one itself. While it shares the same minimalist design language, it feels noticeably flimsy in the hand. The lightweight plastic construction lacks the heft and solid build you'd expect, and the buttons have a slightly hollow click that detracts from the overall experience.

Tracking is accurate enough for basic tasks, and Bluetooth connectivity is stable; however, this mouse isn't suited for more intensive use, particularly for creatives or power users who rely on precision.

The top of the HP OmniVision X houses a 5MP 1440p webcam, which is a nifty bit of engineering. Although the camera's quality is somewhat average, it conveniently pops up from a hidden panel and can be pushed back down when not in use.

HP OmniStudio X: Performance

The HP OmniStudio X features an Intel Core i7-155H processor, paired with an Nvidia GeForce RTX 4050 GPU, 32GB of RAM, and a 1TB SSD.

The Core Ultra 7 chip delivers speedy responsiveness and smooth multitasking, while the RTX 4050 brings a welcome boost for GPU-accelerated tasks, from creative software to casual gaming.

However, it's worth noting that the RTX 4050 is a mid-tier Nvidia GPU, typically found in mid-tier laptops. While it's capable enough for creative workflows and productivity tasks, it's not designed for intensive 3D rendering or high-end gaming. Power users with more demanding graphics want to find something with a bit more grunt.

During testing, the machine remained impressively quiet under pressure, with fans rarely making a fuss. The onboard 32GB of RAM allows you to run multiple heavy applications without any issues, and the 1TB SSD provides fast load times and ample space for large media files or project libraries.

HP OmniStudio X: Is it worth it?

The HP OmniStudio X is an impressive all-in-one PC that clearly aims to give Apple's iMac a run for its money, and in many ways, it succeeds. Its huge 32-inch 4K display is excellent, delivering the screen space and colour accuracy that creative professionals demand. The minimalist aluminium design not only looks high-end but feels built to last, and the wealth of ports tucked around the back is a welcome bonus.

Performance is another strong suit. With Intel's Core Ultra 7 and Nvidia's RTX 4050 under the hood, this machine is more than capable of handling demanding creative workloads, from video editing to 3D modelling. However, the mid-tier GPU may leave the most power-hungry users wanting more, particularly those working with real-time 3D or high-end gaming.

At £1,699, this isn't an impulse buy, but for professionals seeking a powerful, design-led all-in-one for creative work, the HP OmniStudio X makes a compelling case.

HP OmniStudio X specifications

Storage is a “hard problem you can never solve unless you know the future,” says John ‘Coz’ Colgrove, but it hasn’t stopped him trying over the last 16 years.

Speaking to ITPro, the Pure Storage founder has been a vocal champion of all-flash storage for nearly two decades now, having become frustrated with the sluggish performance and diminishing returns on traditional disk-based storage.

This inspiration came during his time working at Veritas Software, where he noted there were aspects to the industry that were “silly or annoying”. These included the difficulties experienced by enterprises in migrating from one storage system to another, he says, or how “ridiculously complicated they made it”.

“I had observed hard drives through 25 plus years personally, and seen how much they improved,” he tells ITPro. “But I also saw that they stopped getting faster about the early 2000s, and they then began to slow down.”

This, Colgrove says, showed there was an opportunity to shake up the market, and flash storage would be the mast he would pin his colors to.

“That provided the opening to say ‘okay it's finally time for a new, better storage medium’, and we can use that as a catalyst to fix these other things around how simple and easy it is to use, the pain of upgrading and operating and all that sort of stuff.”

The benefits of flash storage, from performance to energy efficiency and cost, have been key to the company’s value proposition since it was founded in 2009.

Flash storage has been around for decades, but it wasn’t until the late 1980s and early 1990s that it really began to take off, with a consumer device boom playing a crucial role in driving this growth.

Again, this was something Colgrove noticed at the time and it was clear that the enterprise market had to take a leaf out of the consumer domain’s book.

“In the 70s, all the cool tech was the big enterprise IT systems and things,” he tells ITPro. “Then you start getting the consumer stuff as you went through the late 80s and 90s and, suddenly, consumers had this cool, easy to use tech.”

The enterprise, by contrast, was clunky and not very user friendly, with an attitude of “here’s your system, and here’s a 1,000 page manual you have to study to use it”, he adds.

“It just made no sense. If you rented a car when you came here, you got in it, it’s not the car model you drive, and yet you can drive it and you don’t need to read the 500 page manual that they do for all the nitty gritty controls.”

The core focus here is essentially simplicity, Colgrove says, both on the hardware and software fronts, and when the company launched it had three pillars around which its strategy depended. Naturally, the first of these was leading an industry transition to all-flash storage.

“I believed then that 100% of storage was going to be all-flash and tape,” he says, adding that they’re “not going after tape” and likely never will. “100% of disk should become flash,” Colgrove adds.

Colgrove admits that at the time people “thought I was nuts” for suggesting that a transition of this sort could happen.

“Now they don’t,” he says. “The hard drive guys still think I’m nuts, but they’re just ostriches burying their heads in the sand.”

Dreams of the disk drive’s demise haven’t quite come to fruition yet, but flash has grown in popularity and this is expected to continue in years to come. The all-flash array market was valued at $14.59 billion in 2023 and is expected to reach $74.62 billion by 2032.

Elsewhere, Colgrove says attention also turned to creating products that were simple and easy-to-use. Looking back to the 90s when buying a new phone, he says the process of transferring contacts manually was an indictment of the inherent difficulty that consumers faced.

In the enterprise market, the situation was similarly tedious and offputting - Pure Storage wanted to change that.

“Ease of use is just so important, and then the better customer service/business model,” he says, referring to Evergreen, Pure Storage’s subscription model. It allows the company to continually upgrade customer storage options without the need for disruptive migrations or costly upgrades.

Evergreen sits at the heart of the company’s strategy and, Colgrove says, this has enabled customers to “get rid of the migration problems” faced in years past with disk drives.

Colgrove notes that the company has “stuck to its principles” on these topics over the years, which has proven highly beneficial and enabled it to maintain a clear, long-term roadmap.

Focusing on the data, not the storage

Looking ahead, Colgrove says the company plans to continue its focus on the data management front, which was a key talking point at Pure//Accelerate – the company’s annual conference – particularly with the launch of Enterprise Data Cloud (EDC).

“Don’t manage the storage, manage the data” has been the warcry in Las Vegas,

“I think [trends are] continuing to shift to getting more intelligence about the data,” Colgrove says.

He notes there’s a sense that dealing with data storage is almost like dealing with a “black box” and a major hurdle to adoption of EDC is educating customers that there’s value in it.

“I think we’re going to see a thrust from us in that direction around opening up the black box,” he says. “That’s a bit of a delicate balancing act at times because there are some customers who [feel that] from the array you shouldn’t be able to see data at all.”

“There’s a balancing act there that we have to crack open, that black box is the next level.”

Ultimately, Colgrove believes that the launch of the EDC will prove to customers that there are benefits to be unlocked with regard to automation, visibility, and optimization.

“I am really excited about the Enterprise Data Cloud,” he says. “And I’m really excited about the roadmap and what we’re doing with hyperscalers to wipe out disks. It’s like I said, what we started out to do. Over the next few years, we’re going to make gigantic strides in both those areas.”

Welcome to ITPro's coverage of AMD Advancing AI 2025, coming to you live from the San Jose McEnery Convention Center. Hundreds of delegates are expected to fill the auditorium today to hear AMD CEO Lisa Su reveal the company's latest strategic plans for AI.

Su is expected to take to the stage at 9.30am PST and with her keynote slated to last for two hours, we can undoubtedly expect guests from both AMD and its partners to grace the stage.

With a little under two hours to go before Lisa Su's keynote starts, it's worth reflecting on what happened at the company's last Advancing AI conference, which only took place eight months ago in October 2024.

At that event, Su launched the enterprise-focused Ryzen AI Pro 300 Series CPU range, revealed the Pensando Selina data processing unit, and announced the arrival of Instinct MI325x accelerators. It's unlikely that the company has invited customers, partners, and press to hear updates on these products, but it could indicate the direction of travel.

At Computex in June 2024 Su announced the company was moving to an annual release cadence for its GPUs and with the launch of the Instinct MI325x accelerator series at the last Advancing AI we could see an update in this area.

The auditorium is starting to fill up at the San Jose McEnery Convention Center as we wait for Lisa Su to take to the stage for her keynote address this morning.

And we are underway with an opening video – apparently the key for the AI future is trust.

And here is the woman of the hour, Dr Lisa Su, CEO of AMD.

"I'm always incredibly proud to say that billions of people use AMD technology every day, whether you're talking about services like Microsoft Office 365 or Facebook or zoom or Netflix or Uber or Salesforce or SAP and many more you're running on AMD infrastructure," says Su. "In AI, the biggest cloud and AI companies are using instinct to power their latest models and new production workloads. And there's a ton of new innovation that's going on with the new AI startups."

As with many companies now – including partner Dell and rival Nvidia – AMD is talking up AI agents and its (positive) impact on hardware.

Su says that agentic AI represents "a new class of user".

"what we're actually seeing is we're adding the equivalent of billions of new virtual users to the global compute infrastructure. All of these agents are here to help us, and that requires lots of GPUs and lots of CPUs working together in an open ecosystem," she says.

AI isn't just a cloud or data center question now, it's also an endpoint question, and particularly the PC. "We expect to see AI deployed in every single device," she says.

Su hits on another theme that's becoming popular among many enterprise hardware providers: open source and openness.

"AMD is the only company committed to openness across hardware, software and solutions," Su claims. "The history of our industry shows us that time and time again, innovation truly takes off when things are open."

"Linux surpassed Unix as a data center operating system of choice when global collaboration was unlocked," she continues. "Android's open platform helped scale mobile computing to billions of users, and in each case, openness delivered more competition, faster innovation and eventually a better outcome for users. And that's why for us at AMD, and frankly, for us as an industry, openness shouldn't be just a buzzword. It's actually critical to how we accelerate the scale adoption and impact of AI over the coming years."

Su is now joined on stage by her first guest, Xia Sun of xAI, to talk about how the company behind Grok is working with AMD.

Sun focuses on how his small team is being bolstered by AMD hardware like MI300x and – of course – openness

Sun leaves the stage and now we have our first product announcement, the AMD Instinct MI350 Series.

"With the Mi 350 series, we're delivering the largest generational performance lead in the history of instinct, and we're already deep in development of Mi 400 for 2026," says Su.

"Today, I'm super excited to launch the Mi 350 series, our most advanced AI platform ever that delivers leadership performance across the most demanding models this series," she says, adding that while AMD will talk about the MI355 and the MI350 "they're actually the same silicon, but MI355 supports higher thermals and power envelopes so that we can even deliver more real world performance."

Su claims that the MI350 Series "delivers a massive 4x generation leap in AI compute to accelerate both training and inference". Some of the key claimed specs are:

• 288GB memory
• Running up to 20 billion paramaters on a single GPU
• Double the FP throughput compared to competitors
• 1.6x more memory than competitors

(These are AMD claims that haven't been independently verified by ITPro)

And now another guest: Yu Jin Song, Meta VP of engineering – Meta, of course, being the force behind Llama LLM.

"We're seeing critical advancements that started with [inferencing[ and now extending to all of your AI offering," says Yu Jin Song. "I think AMD and Meta have always been strongly aligned on ... neural net solutions."

"We try to shoot problems together, and we're able to deploy optimized systems at school. So our team really see AMD as a strategic and responsive partner and someone that we really like. Well, we love working with your engineering team. We know that we love the feedback, and you also hold the high standard. You are one of our earliest partners in AI," he adds.

Yu Jin Song explains that MI350 will become a key part of the company's AI infrastructure (it's already using MI300x accelerators).

"We're also quite excited about the capabilities of MI350x," he says. "It brings significantly more compute power and acceleration memory and support for FP4, FP6 all while maintaining the safety pull factor at mi 300 so we can deploy quickly."

These are some of the Instinct MI350 Series partners, which will have products coming to market with the chips in Q3 2025, including Oracle, Dell Technologies, HPE, Cisco, and Asus.

We're going from hardware to software, now, with Su saying this is "really [what] unlocks the full potential of AI". She leaves the stage and Vamsi Boppana, SVP of AI at AMD, takes the metaphorical mic to talk about ROCm.

"We have been relentlessly focused on what matters most, making it easy for developers to build better out of the box capabilities, easy setup, more collateral, stepping up community engagements," he says.

"We've been running hackathons, contests, meetups and mail, and our customers are deploying AI capabilities at unprecedented pace, and that's why we've significantly accelerated hour release cadence," he continues. "New features and optimizations are now shipping every two weeks. Leading models like llama and DeepSeek work on day zero."

He also points out AMD's partnerships with Hugging Face, Pytorch, OpenAI Triton, and Jax.

Now for his announcement: ROCm 7. This will, he says, "bringing exciting new capabilities to address ... emerging trends, and brings support for our MI350 series of GPUs".

"[It] introduces the latest algorithms, advanced features like distributed inference, support for large scale training and new capabilities that make it easy for enterprises to deploy AI effortlessly within ROCm 7," he says, adding that inference has been "the largest area of focus [and] we've innovated and invested at every layer of inference".

Boppana now invites his own guest to the stage, Eric Boyd CVP of AI platforms at Microsoft.

The two are discussing Microsoft and AMD's longstanding partnership, with Boyd saying Microsoft has been using "several generations of Instinct".

"It's been a key part of our inferencing platform, and we've integrated ROCm to our inferencing stack, making it really easy for us to take and deploy the models on the platform," he says.

Boppana is now talking about what he describes as a new opportunity for driving down inference costs: distributed inference.

"In any LLM serving application, there are two phases. There's the pre fill phase and there's decode phase," explains Boppana. "These two phases of the model are typically handled on the same GPU, but now if you [use] the same GPU, it often becomes a bottleneck for large models, or when demand spikes happen and you can get limited in performance or flexibility."

Disaggregating the pre fill and decode phase can "significantly improve throughput, reduce cost, and boost responsiveness", he says.

Here is that solution: ROCm Enterprise AI. "ROCm Enterprise AI makes it easy to deploy AI solutions with new cluster management software, it ensures reliable, scalable and efficient operation of AI cluster," says Boppana.

"Our mlops platform allows fine tuning and distillation of models with your own data and a growing catalog of models will come for specific industries," he adds.

Another announcement from Boppana: AMD Developer Cloud and Developer credits. All developers who are here will have a credit voucher in their email inbox right now, he says, which elicits cheers from the assembled crowd.

Boppana hands over to Anush Elangovan, CVP of software development, and Sharon Zhou, CVP of AI. Elangovan channels Steve Balmer: "Developers! Developers! Developers!" to yet more cheers – although he stopped after just three.

"We are serious about bringing rocket everywhere and to everyone, from client to the cloud," he continues. "In AI, speed is your mode. Access to compute is paramount. We've been delivering on speed, so now let's get you access to compute."

"With the developer cloud, anyone with a GitHub ID or an email address can get access to an instant GPU with just a few clicks," he says, directing developers who want to give it a go to dev.amd.com.

"We've also included a lot of easy to use frameworks like vlln, SG, Lang, pytorch, etc," he says. "You just select one of those frameworks, add your SSH key and then create and you're set."

Zhou briefly encourages developers in the crowd to "come say hi" at hackathons, university labs and elsewhere, before handing back to Boppana, who reveals the next of today's announcements: The ability to access ROCm 7 on Windows or Linux, on device or on client.

Boppana hands over to Forrest Norrod, EVP and GM of data center solutions, who is underlining the importance of agentic AI for AMD, as well as the performance improvements of fifth-generation Epyc CPUs.

He also talks up what AMD says is the continued importance of CPUs.

"Some will naively tell you that CPUs are less important in the age of AI, but that's not correct," he says. :With agentic AI, we see an explosion of autonomous agents accessing data and enterprise applications. This increases the needs for efficient, high performance x86 compute across the data center, then within the AI server itself, the CPU serves the demands of pre processing and workload orchestration to keep the GPUs working efficiently."

"Let me show you a few examples of how the right CPU can manage GPUs work better," he says. "Choosing poorly can create bottlenecks and stranded valuable resources. As you can see across a range of models and use cases, our fifth generation epic CPUs can boost the inference performance of the entire system from six to 17% that makes a huge impact on the overall TCO and performance of The AI deployment, and it's a critical element in designing the best possible AI system. So get much more out of your GPUs with the right CPU."

Lisa Su is back on stage to give the audience "a first look at the next big step for [AMD's] AI roadmap, the MI400 series". She quickly pivots from there to introduce Helios AI Rack.

"Think of Helios as really a rack that functions like a single, massive compute engine," Su says. " It connects up to 72 GPUs with 260 terabytes per second of scale up bandwidth. It enables 2.9 exaflops of FP four performance, and that is a great number."

Naturally, Su also takes a moment to make comparisons with AMD's competition, saying Helios supports "50% more HBM ... memory bandwidth, and scale out bandwidth".

And now, a surprise guest, "someone who's really an icon in AI", says Su: Sam Altman, CEO of OpenAI.

Altman talks about his company is seeing in the market.

"I think the models have gotten good enough that people have been able to build really great products, text, images, voice, all kinds of reasoning capabilities," he says.

"We've seen extremely quick adoption in the enterprise now, coding has been one area people talk a lot about, but I think what we're hearing again and again in all these different ways, is that these tools have gone from things that were, you know, fun and curious to truly useful," he says.

Asked by Su how compute demands are changing, Altman says: "One of the biggest differences has been, we've moved to these reasoning models. So we have these very long roll outs where a model that often think about come back with a better answer.

"This is really the pressure on model efficiency and long context for all this, we need tons of computers, tons of memory, tons of CPUs as well. And our infrastructure ramp over the last year and what we're looking over the next year has just been a crazy, crazy thing to watch."

Turning to the hardware used to support OpenAI's platforms like ChatGPT, Altman says the company is already running some work on the MI300x, "but we are extremely excited for the MI450, the memory architecture is great for inference. It can be an incredible option for training as well".

Su asks Altman what he sees in the future for AI. "How will the workloads evolve? What happens with – quote, unquote – AGI (artificial general intelligence)," she says.

"I think we're going to maintain the same rate of progress rated models for the second half of the decade as we did for the first," Altman responds. "I wasn't so sure about that a couple of years ago for a new research thing to figure out, but now it looks like we'll be able to deliver on that."

"So if you think forward to 2030 ... these systems will be capable of remarkable new stuff, about scientific discovery, running extremely complex functions throughout society, and things that we just couldn't even imagine," he continues. "It's really going to take these are huge systems now, very complex engineering projects, very complex research, and to keep on this [path of] scaling, we've got to work together across research, engineering, hardware, how we're going to deliver these systems and products.

"This has gotten quite complex, but if we can do on that, if we if we can deliver on that drives collaboration across the whole industry, we will keep this curve going," he concludes.

Altman leaves the stage and Su delivers some final thoughts to the audience.

"When I think about this past year, it's really redefined what progress in AI looks like. It's really moved at a pace unlike anything that we have seen in modern computing ... anything that we've seen in our careers, and, frankly, anything that we've seen in our lifetime," Su says.

"The future of AI is not going to be built by any one company or in a closed ecosystem. It's going to be shaped by open collaboration across the industry. It's going to be shaped because everyone is bringing their best ideas, and it's going to be shaped because we're innovating together. So on behalf of all of us at AMD, we look forward to changing the world with you together," Su concludes, and with that the keynote is over.

Geekom is another Chinese OEM carving out a name for itself in the Mini PC market. Unlike the mini PC boxes we've looked at recently from relative newcomers Acemagic and Beelink which are both entry level devices built around Intel's N150 processor the Geekom Mini IT13 2025 Edition, to give it its full title, runs on the altogether more potent Core i9-13900HK CPU.

Naturally, that puts the new Geekom box into a different price category from its Twin Lake juniors. At the time of writing, Geekom is selling the 1TB model for £649 rather than the "usual" RRP of £849 and the 2TB model for £699 rather than, oddly, £799. It's good to see a manufacturer not gouging the public for extra storage capacity. Certain famous computer makers with fruit logos could take note. Both models come with 32GB of RAM.

That price puts the Mini IT13 in direct competition with the likes of the Asus PN53 and Lenovo ThinkCentre M70q Gen 4 Tiny, though both those machines are bigger and heavier than the Geekom. What we have here, then, is a grown-up mini PC squeezed into the footprint of a small, budget mini PC.

Geekom Mini IT13: Design

The Geekon Mini IT13 2025 edition looks very much like the Mini IT13 which in turn bore a remarkable similarity to the IT12 and IT11. Clearly "if it ain't broke, don't fix it" is the order of the day and Geekom and I say "fair enough" to that because, as mini PCs go there is nothing much to criticise in the design.

In size and shape the Mini IT13 2025 is par for the NUC course, weighing 652g and measuring up at 117 x 112 x 49mm. The exterior is a solid metal box finished in an appealing grey-green paint with the makers name painted on the lid in white.

Naturally, given the diminutive size of the Mini IT13, the PSU is external, but the 120W rat-and-tail affair is small enough to be tucked away out of sight and mind and comes with a usefully long cable. Geekmon also bundles a VESA mounting bracket and an HDMI cable.

Geekom has done sterling work squeezing a wide selection of ports into the Mini IT13's case and making full use of the chipsets 28 PCIe lanes to provide excellent levels of connectivity.

On the front panel along with the on/off switch you'll find two 10Gbps USB-A ports and a 3.5mm combo audio jack, while on the left side, there is a full-sized SD card slot and on the right a Kensington security lock.

But it's around the back that things get really busy. Here you'll find two HDMI 2.0 ports, two USB-C 4.0 ports both supporting DisplayPort video, another 10Gbps USB-A port, a USB-A 2.0 port, a 2.5GbE Ethernet jack and the barrel jack for the PSU.

That's a very snug arrangement, and removing things when all the ports are occupied involves taking some care, but you have to pat Geekom on the head for squeezing in the maximum number of connectors into such a small area.

Getting inside the Mini 13 is easy enough, just undo the four Philips screws that sit inside the rubber feet that are fixed to the base and pull the body up from the base. Be careful, though, because there's a ribbon cable that connects the motherboard to the SATA3 connector that is attached to the base. The base plate also features a slide-in cradle for a 2.5-inch HDD, which makes installing said a cinch.

Once you've separated base from body, there is easy access to the two SODIMM slots as well as the occupied PCIe Gen 4 M.2 2280 and unoccupied 2240 SSD mounts. Our test box came with two 16GB RAM sticks fitted.

According to Geekom's specification sheet, wireless communications are handled by an Intel AX211 card, but our unit had a MediaTek MT7992 component. Both cards support 6Ghz Wi-Fi 6E and Bluetooth 5.3 so we don't regard this as a major cause for concern.

Geekom Mini IT13: Specs and Performance

The Core i9-13900HK chip inside the Mini IT13 is, of course, the 14-core, 20-thread mobile chip, here with a 65W maximum TDP, not the 125W, 24-core desktop model. The mobile i9 is still a strong performer though with 6 Performance cores and a maximum turbo clock of 5.4GHz.

The 2024 model of the Mini IT13 used the Core i9-13900H CPU which only gives away between 8 and 10 per cent in multi-thread performance so there's really no need to upgrade if you have one of those on your desk.

The 2025 update of the Mini IT13 still uses DDR4 rather than DDR5 memory, despite the processor supporting the latter, which offers greater bandwidth and higher speed. A move to DDR5 would also boost the performance of the Iris Xe iGPU, which shares the 32GB of system memory.

In our standard 4K multi-media benchmark, the Mini IT13 scored 262 points. To put that into context, the Asus PN53, which is built around an 8-core AMD Ryzen 6900HX CPU with 32GB of DDR5 RAM, scored 298 points while Lenovo's ThinkCentre M70q Gen 4 Tiny, which uses a desktop Core i7-13700T with 16GB of DDR4 RAM, scored 303. The 41-point difference between the Geekom and Lenovo machines would be very difficult to discern in everyday use.

In the GeekBench 6 CPU test, the Geekom and Lenovo machines ran neck-and-neck, the Mini IT13 scoring 2,694 single-core and 10,286 multi-core to the M70q's 2,585 and 10,941. It was a similar story in the PCMark 10 test with the Lenovo just shading the Geekom box by 6,066 to 6,002.

Comparing the Asus, Lenovo and Geekom machines using the Cinebench R23 multi-core test, we get 11,345 vs. 12,914 vs. 12,127, respectively. Given that the Asus and Lenovo boxes are both bigger and heavier than the Geekom, this has to be seen as a good set of results for the diminutive Mini IT13.

The Xe graphics integrated GPU rather lets the side down as it simply can't compete with the latest Arc and Radeon iGPUs from Intel and AMD. To put that into context, the SPECviewperf 3dsmax 3D modelling benchmark ran at a pedestrian 15fps. Intel's Arc 140V iGPU can hit 25fps in the same test while AMD's Radeon 890M can get close to 40fps.

The Radeon 680M iGPU in the Asus PN53 ran the same SPECviewperf test at 28fps, so if outright graphics performance is a requirement but you still fancy the look of the Mini IT13, it could be worth looking at the Geekom AX8 Pro which features an AMD Ryzen 7 8845HS CPU with Radeon 780M graphics.

Of course, you could always invest in an external GPU enclosure and connect it to the Mini IT13 via one of the 40Gpbs Type-C ports. That would theoretically turn it into a very potent graphics and gaming platform.

The 1TB Kingston SSD in our review unit turned in sequential read and write speeds of 3,312MB/s and 2,026MB/s, respectively, which is a pretty unremarkable level of performance, though sufficient for the kind of general computing jobs the Mini IT13 is likely to be pressed into.

The system is cooled by a single fan that sits above the motherboard. It runs very quietly even at full speed, but the absence of an air intake grille above the fan does seem to limit performance under extreme stress: After a 30-minute stress test, CPU utilisation dropped to 60 per cent indicating some thermal restrictions.

The Mini IT13 comes loaded with Windows 11 Pro, but I found that everything worked as it should when running Ubuntu 24.10, making this a good platform for anyone wanting to walk the open-source path. It's a shame you can't buy the Mini IT13 without an operating system preinstalled to save some money.

Geekom Mini IT13: Is it worth it?

Given the wide range of I/O ports and the ease with which you can add more memory and storage, it's hard to argue with the Mini IT13's value offering, especially at the current "offer" price direct from the manufacturer. Baseline performance is good too, easily matching the more well-known offerings we've tested from the likes of Asus and Lenovo.

Graphics performance does let the Mini IT13 down somewhat, but that's more due to the leap forward that Intel has recently made with its integrated GPUs than any inherent fault with the chipset inside the Mini IT13. If you're worried about buying from a relatively unknown name, Geekom advertises a 30-day money-back guarantee and a 3-year warranty.

Geekom Mini IT13 specifications

Microsoft is entering the mini desktop PC market with the availability of its business-centric Windows 365 Link thin client.

The small, neat-looking black box has a shiny Windows logo on the top and will be available for $349.99 in several countries (including the US and UK).

Windows 365 Link was announced last November as a continuation of Microsoft's Cloud PC push. The device is easier for IT departments to manage than a traditional computer, reducing the need for hands-on support.

The Link won't appear in shops, as such, as it is marketed for businesses. To purchase, customers will likely need to buy in bulk from a Microsoft account team or an authorised reseller. Regional availability includes the US, Canada, Australia, Germany, Japan, New Zealand, and the UK.

Modern thin clients

The Link is a modern version of what is known as a 'thin client', a PC that runs a virtualized desktop off an on-site server. The difference here is that the 365 Link runs Windows over the internet so that the user can work from virtually anywhere. Also, Microsoft says it has been built to boot within seconds, which again, differentiates it from traditional thin clients.

The Link has dual 4K monitor support, four USB ports, an Ethernet port, Wi-Fi 6E, and Bluetooth 5.3, as well as seamless connectivity with both wired and wireless peripherals.

What's more, the device's local processing capabilities offer high-performance video playback and video conferencing with Microsoft Teams meetings. For other meeting services, Microsoft said it was working with partner solutions such as Webex by Cisco, which might be available in the near future.

Microsoft had tested the Link with partners and selected customers before pushing it out for sale, taking in feedback for security. As such, the tech giant said the device "prioritizes security" and features a locked-down operating system with no local data or apps, and no local admin users.

Eliminating local data and apps "significantly reduces" the attack surface, according to Microsoft, making it harder for malicious actors to compromise the Windows 365 Link. The device will also come with passwordless authentication through Microsoft Entra ID, allowing users to sign in with multifactor authentication.

MORE FROM ITPRO

• Amazon's thin client for business
• Can thin clients be an antidote to sprawling attack surfaces?
• Windows 365

It’s been a whirlwind two years since Sonya Mathieu joined NetApp. Leading the UK & Ireland channel division has placed the industry veteran at the forefront of the firm’s push to secure its position as the leading storage provider in the age of AI.

Having joined during the embryonic stages of the generative AI ‘boom’, the company’s efforts to forge closer ties with partners and deliver success for customers has been a protracted process - but one that’s now paying dividends.

The company unveiled sweeping changes to its channel strategy in 2023, centered firmly around increasing rebate payouts and streamlining distribution agreements.

Relationships with leading - or ‘prestige’ - partners was also a key focus in this pivot in a bid to foster closer alignment on messaging, Mathieu told ITPro.

“It’s been really positive,” Mathieu told ITPro.

“We’ve got six of what we call ‘main managed partners’, which are our prestige partners in the UK, and we pride ourselves on having very close relationships at every level within those organizations to make sure we’re aligning our focus for messaging into what is top of mind for those organizations.”

“That’s been really successful over the last two years,” she added.

Similarly, on the distribution front, the storage giant has made some significant investments, according to Mathieu, largely focusing on muscling in on burgeoning growth areas, such as SAN.

“We've seen some really significant gains over the last year, which is exactly what we wanted to see.”

NetApp capitalizing on SAN

The growing SAN market is an area in which NetApp believes has both huge potential and a domain where it can pry custom from competitors. In the UK, the SAN market is £300 million-plus, Mathieu noted, and while the firm currently commands a 10% share, the plan is to delve further into this space.

“A big focus for us is around what we call ‘secure SAN’. We’ve made some significant advancements around security, and we are the most secure SAN storage vendor in the market,” she said.

Mathieu noted that, among partners leaning into SAN or its storage as a service model, those partners have doubled their business with the company over a two year period, underlining the success of this strategy so far.

Capitalizing on this is a challenge, Mathieu admitted, not least of all due to the confluence of issues affecting customer adoption rates and hardware refresh programs.

“Security is top of mind for all of our partners and our customers,” she said. “If you think about the process that organizations are going through to determine, you know, what hypervisor they want to move forward with, they’re looking at modernizing that infrastructure.”

Dialing in on the security impetus is vital here, Mathieu added. Enterprises embarking on IT modernization programs can ill afford to skimp on this area.

It’s here that NetApp has a “really unique differentiator” by being able to offer anti-ransomware protection, according to Mathieu, and it’s a message that resonates with partners and customers alike when combined with the company’s close integration with major cloud providers.

Safe, guiding hands are critical when enterprises face such high stakes.

“When we are able to lead with the security messaging and our ability to leverage cloud services, we have no competition,” she said. "It just depends on when we get into that battle and that conversation; whether it's at the beginning or the middle or the end, which always can make things a little bit challenging.”

“But there are a number of opportunities where we've been engaged and we've won and were brought in at the last minute because of the way that we challenged our customers to look at the importance around security and our key differentiators.”

Collaboration on AI is vital for NetApp

Generative AI is an unavoidable topic that has seeped its way into both the storage and security domains over the last two and a half years. NetApp has made no secret of its position on the technology, framing itself as a leading storage provider to support innovation in this space.

CEO George Kurian told ITPro in late 2023 that the company’s efforts in driving hybrid cloud adoption would be a key differentiator for customers adopting AI tools, and so far this has proved a highly successful strategy.

With regard to partner collaboration on this front, Mathieu said the company’s two-pronged approach to messaging has been crucial. This centers around highlighting the potential opportunities for partners and customers alike, but also addressing growing concerns over sustainability.

“The market opportunities are there,” she said. “The spend will be there. So it’s about making sure that we are out having conversations and enabling our customers and partners to understand the value that we bring.”

“A lot of these workloads will start in the cloud, you’ll then want to move them back on-prem, and we’re the best data management company to help them do that,” Mathieu added.

MORE FROM ITPRO

• NetApp is on a mission to drive hybrid cloud adoption
• The tech industry “needs to do a better job” on AI sustainability, says NetApp CEO
• NetApp wants to talk about cold data and its impact on sustainability

Qnap's TS-h1290FX all-Flash NAS impressed us when we tested it as this compact desktop appliance delivered a performance that would appeal to teams running demanding workloads such as real-time 4K/8K video editing. Its AMD EPYC power, massive memory capacity, and integral support for U.2 PCIe NVMe Gen 4 and SATA SSDs did come with a high price tag and for those with more modest budgets, Qnap now offers the TS-h1277AFX.

At first glance, the TS-h1277AFX looks identical to its predecessor as it uses the same sleek desktop chassis with twelve SFF hot-swap drive bays at the front. It's all change inside though, as this model supports SATA SSDs only, is powered by an 8-core 3.8GHz AMD Ryzen 7 9700X CPU and its base 32GB of memory is the faster DDR5 variety which can be upgraded to a very usable 192GB.

You still get dual 2.5GbE multi-Gig network ports but the embedded 25GbE SFP28 ports in the TS-h1290FX have been replaced with copper 10GbE ports which are also multi-Gig. The PCIe Gen 4 slot count has been reduced from four to three but that still offers plenty of expansion potential with the appliance supporting Qnap's dual-port 10GbE, 25GbE, and 100GbE cards.

Qnap TS-h1277AFX review: Build quality

The TS-h1277AFX looks good on the desktop and the panel above the drive bays incorporates a backlit LCD screen showing details on all network ports and their IP addresses. The chassis is constructed of sturdy pressed steel and removing the metal cover reveals the Ryzen CPU in the center and mounted by a very large active heatsink.

Usefully, Qnap has provided the base 32GB of memory on a single stick and the four DIMM slots means you can upgrade to 128GB without having to replace it. Cooling is handled by two 9cms diameter fans and during all our performance tests, we found the appliance to be whisper quiet.

At the rear, you have Type-A and Type-C USB 3.2 Gen 2 10Gbps ports and another new feature is video from the CPU's embedded Radeon iGPU is piped through to an HDMI 1.4b port. The appliance is also compatible with a wide range of GPGPU cards and Qnap has even fitted two extra cables to provide additional power for them.

Qnap TS-h1277AFX review: Software features

The 'h' in its name indicates the TS-h1277AFX is aimed at Qnap's QuTS hero OS which is a superior choice over QTS if you want the best data protection features. They're both very easy to use and present almost identical web interfaces but the ZFS-based QuTS offers near unlimited NAS share and iSCSI LUN snapshots, end-to-end checksums, encryption, and two WORM (write once, read many) policies for protecting NAS share data from tampering, and ransomware attacks.

It includes compression and the base memory enables deduplication which requires a minimum of 16GB. RAID options are more extensive as along with the usual suspects, QuTS supports triple parity RAID to protect against three drive failures and triple mirroring which stores identical copies of data on three drives.

Apps are in abundance with over 110 available for the TS-h1277AFX. All security and backup apps are available and include Qnap's Security Center, Hybrid Backup Sync (HBS) 3, and Hyper Data Protector (HDP) with the latter now linking up with the free NetBak PC agent for block-level backups of Windows servers and PCs.

Along with optional cloud storage, the myQNAPcloud service provides remote access and administration of registered NAS appliances and once you've created a Qnap ID, you can use the new Hybrid Backup Center (HBC) cloud service. Currently, in beta, HBC provides centralized remote management of HBS backup and restore tasks for multiple appliances.

Each appliance requires HBS 3 and the HBC agent installed which allows you to create backup and sync jobs from the cloud portal, choose sources and destinations, schedule them, monitor them in real-time, and load a topology view to see all backup activity. A smart feature is the option to declare remote appliances as targets and if Qnap's RTRR (real-time remote replication) service isn't loaded, the HBC agent will install it for you.

Qnap TS-h1277AFX review: 10GbE performance

For our 10GbE lab performance tests, we fitted eight 1TB Western Digital Red SA500 SATA SSDs and created a single RAID5 array. With a NAS share mapped to a Dell PowerEdge R760xs Windows Server 2022 host, we recorded 9.2Gbits/sec and 9.3Gbits/sec for sequential reads and writes with the same results returned for random operations.

IP SAN speeds over 10GbE are equally impressive with a 1TB iSCSI target also returning Iometer sequential read and write speeds of 9.2Gbits/sec and 9.3Gbits/sec and the same rates for random operations. Moving up to a dual 10GbE MPIO link to the iSCSI target delivered sequential read and writes of 18Gbits/sec and 18.3Gbits/sec while random operations recorded 18Gbits/sec and 17.8Gbits/sec

Swapping to 4K block sizes delivered solid I/O throughputs for the MPIO iSCSI link with Iometer reporting 294,200 and 313,900 IOPS for sequential reads and writes and 294,050 and 82,900 IOPS for random reads and writes. We noted our random write IOPS test pushed the Ryzen CPU to around 90% utilisation but with all other tests, it never strayed above 20%.

Qnap TS-h1277AFX review: Is it worth it

Value is comparatively good as the price for the TS-h1277AFX with 32GB of memory starts at a shade over £3k – around £1,300 less than the entry-level TS-h1290FX. It's also a more sensible choice for businesses that want to use cheaper SATA SSDs as they won't be paying extra for the privilege of support for expensive PCIe NVMe SSDs.

It still represents a considerable investment but the appliance's powerful combination of AMD Ryzen 7 9700X CPU and DDR5 memory delivers extremely good performance over 10GbE with high IOPS rates for random reads and writes. The TS-h1277AFX is a solid choice for content creators and video editors that need fast access to data and Qnap's QuTS hero software delivers a wealth of features including great data protection services.

Qnap TS-h1277AFX specifications

The Vista V1 Mini PC is the second mini PC ITPro has tested built around Intel's new N150 CPU. Like the Beelink S13, it's an ultra-compact device designed for lightweight general use or with embedded systems, or as a NAS device or a home theatre PC. Again like the Beelink S13, it comes preloaded with a legit version of Windows 11 Pro and is available from Amazon.

Acemagic may not be a name overly familiar in Europe. Still, alongside Beelink, Chuwi, Geekom, and GMKTek it's one of an increasing number of Chinese OEMs who are carving out an ever-increasing slice of the mini PC market.

The Vista V1 is available in three versions, the cheapest comes with 8GB of DDR4 RAM and a 256GB SSD and an RRP of £229.99 including VAT, next comes a model with 16GB of RAM and a 512GB SSD for £259.99 while at the top of the tree sits a 16GB / 1TB model which costs £279.99. Inside the box along with the PC itself, you get a VESA mourning bracket and screwns and an HDMI cable.

Before we get into the review proper one further issue needs to be addressed. It only takes a quick internet search to discover that Shenzhen Shanminheng Technology Co. Ltd. the company that owns several different brands including Acemagic, Kamrui, NiPoGi, and CTONE has had issues in the past with malware being found on some of its Windows PCs. With this in mind, we thoroughly scanned our review sample and found nothing untoward and have no reason to believe that this is still an issue with Acemagic PCs.

Acemagic Vista V1: Design

Tiny, miniscule, petite, diminutive. All these words can be used to describe the Acermagic Vista V1 because boy, it is small. At 100 x 100 x 32mm and weighing just 245g it's smaller and lighter than both the Beelink S13 and the GMKtec NucBox G1. If there's a smaller Windows mini PC on the market we have yet to stumble across it.

All parts of the external case are made from plastic but it feels solid and sturdy and the silver paint finish makes it look and feel like a more expensive bit of kit than it actually is.

The range and positioning of the ports are typical for the breed with two USB-A ports on the front along with a power button and 3.5 audio jack while around the back you'll find two more USB-A ports, HDMI 2.0 and DisplayPort 1.4 – both capable of 4K 60fs – video outputs, an RJ-45 Gigabit LAN port, a Kensington lock, and the DC-in port.

The differences between the Vista V1 and Beelink S13 are small but worth mentioning. The S13 features two HDMI ports but no DisplayPort while all four of the Beelink's USB-A ports are 3.2 Gen 2 or 10Gbit/s specification while the two rear ports on the Vista V1 are USB 2.0 spec.

Given that the rear data ports are likely to be used to attach a keyboard and mouse their data speed to arguably irrelevant so the Vista V1's inclusion of a DisplayPort socket makes it the more versatile of the two machines in our opinion. Wireless communications are managed by the Realtek RTL8821CE card which supports Wi-Fi 6 over the 2.4 and 5Ghz wavelengths and Bluetooth 5.2.

The RTL8821CE has a reputation for not playing nice with some Linux distros, unlike the Intel AX101 card in the Beelink box, however, when we booted the Vista V1 into Ubuntu 21.10 everything worked perfectly including both the Wi-Fi and Bluetooth radios.

Taking the Vista V1 apart involves prising off the four glued-on rubber feet to access the four Philips screws that hold the chassis and case together. The screws are at the bottom of a long, thin hole so you will need a correspondingly long, thin screwdriver to get at them.

Once you've removed the screws and slid the base unit out of the case you can easily access the M.2 2280 SSD mount and SODIMM slot. Given that there's only one of each opening up the Vista V1 is arguably pointless unless you've bought the cheapest model and feel the need to upgrade either the memory or storage cards.

The Beelink S13 offers no better expansion option when it comes to memory – the N150 chipset only supports single-channel RAM – but you can fit it with a second SSD.

Acemagic Vista V1: Specs and performance

The new Intel N150 processor, commonly known as Twin Lake is based on the Alder Lake design but with minor clock speed improvements. The N150 is a four-core four-thread processor with a base TDP of 6W, a maximum turbo clock of 3.6GHz, and an integrated GPU with 24 execution units and a 1Ghz clock. It's intended for low-power and low-cost implementations.

Though the Vista V1 shares a nearly identical basic specification to the Beelink S13 it proved faster in some of our benchtests. In ITPro's in-house multi-media test, the Vista V1 scored 65 points to the Beelink's 50.5, a surprising 30% advantage, while in the PCMark 10 productivity benchmark, the Vista V1 again beat the Beelink machine, scoring 3,252 to the S13's 2,984.

Looking at the results from the Geekbench 6 and Cinebench R23 tests there's little to choose between the Vista V1 and S13 in terms of CPU performance, the Vista V1 scoring 2,750 and 2,740 in the respective multi-core tests, the S13 scoring 3,029 and 2,890. In the Geekbench 6 Open CL test though the Vista V1 is the winner scoring 4,267 to the S13's 3,293.

That suggests that it's in the GPU department that the Vista V1 has the edge, something borne out by the GFXBench benchmark. In the Car Chase and Manhatten off-screen tests, the Vista V1 scored 32.9 and 69.1fs while the S13 scored 25.6 and 56.6fs.

Strangely the GPU memory clock is higher in the S13 at 1,596MHz to the Vista V1's 1,330MHz which suggests the S13 should be faster rather than the other way around.

Of course, none of these different scores means much in the real world: Both the Vista V1 and the S13 deliver the sort of performance you could expect from a 2017-vintage dual-core Intel Core i5-7500U processor and will be all at sea when faced with anything other than low-power tasks.

You'll also see higher scores from any mini PC built around Intel's 15W TDP N95 or 12W N97 processors. Still, the fact remains that for reasons we can't nail down the Vista V1 is the more potent of the two.

Like the S13 the Vista V1 can happily run for prolonged periods with both the CPU and GPU maxed out without any thermal issues. The Vista V1 is actively cooled but unless you checked the specification sheet you'd never know: Even under heavy stress, the fan is completely silent.

One feature shared by both the Beelink S13 and Acemagic Vista V1 is a glacially slow SSD. The Vista V1's SATA 3 drive – made by Rayson – recorded average sequential read and write speeds of just 509MB/s and 465MB/s respectively: Dismal, but within a margin of error no worse than the Beelink S13 which packs an equally pedestrian drive.

Acemagic Vista V1: Is it worth it?

Taken on its merits the Vista V1 is an easy piece of kit to recommend. The size, weight, and presence of a DisplayPort video output give it an advantage over the Beelink S13 if you want a very small mini PC to carry around with you on your travels with the widest range of video connectivity options.

The Beelink S13 counters that with the option to add a second SSD and by being – at the time of writing at least – cheaper in the UK though as with most of these Chinese products the price can vary dramatically depending on when and where you look. At the end of the day, they are both decent little units that do their intended jobs without drama and both represent good value for money.

Acemagic Vista V1 specifications

Data storage capacity requirements are set to surge by 150% in the next couple of years, new research from Hitachi Vantara has revealed.

While the average large organization currently holds 150 petabytes (PB) of data, organizations are expected to be storing over 300PB of data by the end of 2026.

This is matched by an uptick in investment, with organizations' financial commitments in data storage set to more than double over the same period at an increase of 224%. Investments in AI and processing power, meanwhile, are also set to more than double.

According to the study, 31% of IT leaders said data storage capacity was a serious concern as the rate of growth is placing strain on critical assets, while some also complained of the increasing complexity of data.

More than one-third (76%) of those surveyed said more than half of their data is now unstructured.

While the demand for larger, more complex data storage is rising, the storage industry isn’t ready for this increase, Seagate CCO B.S. Teh said.

“As AI matures and scales, the value of data will increase, leading us to store more data for longer,” Teh told ITPro.

“However, the storage install base is forecasted to have a 17% CAGR - therefore at a significantly slower pace than the growth in data generated. And it takes a whole year to build a hard drive,” he added.

This growth rate disparity will disrupt global storage supply, Teh said, and businesses will need to build long-term plans around capacity to ensure storage supply as generative AI use becomes more strategic.

The AI data problem

The driving force behind this looming data problem is AI. As a technology, it demands not just access to larger quantities of data, but also lengthy storage periods.

“Whether it’s from capturing training checkpoints to saving source data sets, the more data we retain during the process, the more we can validate AI as trustworthy,” Seagate CEO Dave Mosley told ITPro.

“That data therefore needs to be available long term, not just to comply with evolving legal requirements but also to ensure that inference is explainable,” he added.

This ever-growing volume of data can cause different problems for organizations, depending on what route they choose for storage.

“While outsourcing storage comes with its own security, compliance, and potential risk, holding the data for AI internally can become eye-wateringly expensive,” Paul Mackay, RVP of cloud in EMEA and APAC at Cloudera, told ITPro.

What can businesses do?

To avoid feeling the sting of increasing data storage requirements, businesses will have to focus more keenly on data management strategies and data architectures, experts have told ITPro.

Chris Harris, VP of global field engineering at Couchbase, told ITPro that for organizations to fully unlock value from generative AI, infrastructure modernization projects must be conducted in parallel to the development of a robust data management strategy.

“To capitalize on AI investments now and in the future, organizations must ensure they can control data storage, access, and usage, enable real-time data sharing, and maintain a consolidated database infrastructure to prevent multiple versions of data,” he said.

As AI is increasingly integrated into applications, Couchbase’s VP of AI and Edge Mohan Varthakavi said, data architectures will also need to be redesigned to support new workloads.

“Companies will implement new data architectures that go beyond simple record storage to capture the ‘intelligence history’ and thought processes of AI systems,” Varthakavi told ITPro.

“They will need to simplify complex architectures, including consolidation of platforms, and eliminate data silos to create trustworthy data,” he added.

The ongoing AI boom has precipitated a change in computational methods from the CPU to GPU, and storage architectures will need to adapt to meet the new demands of AI workloads, according to a Huawei exec.

As illustrated by the unprecedented rise of Nvidia to become the AI chipmaker par excellence, the industry has transposed much of its computation demands on GPU data centers to train their AI models.

Peter Zhou, president of data storage product line at Huawei, thinks there needs to be a corresponding change in storage systems to accommodate the needs of these GPU-based facilities.

Speaking to ITPro, Zhou explained how the storage industry has remained relatively stagnant for the last decade, but advancements in machine learning, generative AI, and big data will continue to disrupt the space and force vendors to innovate.

“There were not really any big changes happening in that domain , the industry like EMC or NetApp, they have been doing what they have been doing for 10 years. Their product has not really been changed for about 10 years, but when big data and AI came to reality, everything changed” he explained.

“First of all the value of data changed…Data is not just used for recording, data becomes an asset that machines can use to study [the world], it’s the source of their knowledge and can enable their ‘thinking’, and now the data storage industry has to be changed”.

The IT industry, therefore, has to focus on ensuring their storage solutions can both accommodate the new volume of data being collected and ensure it is made available at the speeds these systems require.

Data mobility is storage industry’s biggest challenge

Zhou detailed how storage has a vital role to play in driving the efficiency of these GPU-based computation systems, noting that key considerations such as energy consumption and performance are vital for the modern business engaging in AI development.

“In today’s GPU the work efficiency is less than about 50%, which means that half the time the GPU is just waiting there and wasting energy. The only way to change this is by making the data storage system more efficient, and then the efficiency of these GPUs can be improved by about 30%, which is a lot of energy.”

But to achieve the level of efficiency these GPU clusters need, the underlying architecture of storage systems needs to evolve, according to Zhou.

“I think the architecture of data storage has to be changed. Today, out of the box data storage, the architecture is exactly the same as the normal CPU-centric architecture, which is very old,” he noted.

“The CPU is there, there’s memory directly connecting to the memory, the memory is so small and the data cannot be stored in the memory long term. And then they use the I/O interface to connect all these disks, and then when the CPU is processing the data, this data needs to be brought to the memory, and then and then back to the processor etc.”

Zhou stated roughly 60% of energy consumed by these storage devices is used for data mobility, and reducing this is the challenge storage vendors will need to overcome in coming years.

Huawei’s in-house chipset capabilities could be its greatest strength for storage innovation

The challenge for the industry, in Zhou’s view, is that the hardware needs to be changed to meet these new efficiency requirements, but the major players in data storage aren’t designing or producing their own chipsets.

This consigns them to using general purpose CPUs and other components, leaving potential efficiency gains on the table - and this is where Zhou argued Huawei have an advantage due to their in-house chip production.

“We are designing and producing our own chipsets, which means we have the capability to change the architecture of the computational system.”

He ran through some changes the company has made to move to what he describes as a ‘data-centric’ architecture. Earlier this year, for example, Huawei announced it would move away from I/O interfacing with its storage architectures with all of its components now connected to a universal data bus, leveraging Intel’s Compute Express Link (CXL) to enable faster communication with the GPU.

Zhou also raised decoupling the data plane from the control plane in storage systems, providing benefits in terms of the scalability, performance, flexibility, and management, thereby helping to avoid performance bottlenecks when tasked with large amounts of data and particularly complex AI workloads.

Many of these changes follow similar adaptations Nvidia has made on the computational side of the equation, Zhou said, adding that storage architectures will inevitably need to follow its example to adapt to the evolving needs of the industry.

“If you look into the Nvidia GPU cards, you can sense the architecture of the computational system is changing. This means data storage has to be changed to reflect this.”

More from CloudPro

• How Snowflake is helping customers navigate new and emerging challenges
• AWS CISO Chris Betz wants greater industry collaboration for the multi-cloud era
• Partnerships are Oracle's key to success in the "multi-cloud era"

The Lenovo ThinkCentre M90a Gen 5 stands out in the all-in-one space. With a 13th-generation Intel Raptor Lake CPU, a wide range of ports, decent quality screen and speaker systems, and a design that aspires to be something more than downright ugly it will happily grace even the most stylish offices at work or home.

The price for the Gen 5 all-in-one starts at £870 inc VAT and runs to £1,660 if you want the most potent silicon and maxed-out memory and storage. The model we were sent to test will set you back £1,240 which is the price we'll be basing our value judgments on.

Lenovo ThinkCentre M90a Gen 5: Design

Things made by Lenovo with the word Think in the title tend to be black and so it is with the M90a Gen 5, the only color highlight being the red channel along the front of the stand that can be used to store pens, pencils, and mobile phones.

The plastic cabinet that houses the display and electronics is less than 55mm at its deepest which makes the whole system no larger than a 24-inch monitor which is the point behind an all-in-one. At just over 8Kg it's rather heavier than your average 24-inch monitor.

Perhaps a little unnecessary for an office-bound device the M90a Gen 5 meets the US Department of Defense's MIL-STD 810H standards for resistance to temperature, pressure, humidity, vibration, and more. Lenovo says it has tested it from the "icy expanses of the Arctic tundra to dust-laden desert winds" no less.

The bezels above and to the side of the 24-inch (23.8-inch to be exact) screen are nicely slender at 8mm and while the chin below is much deeper at 48mm it houses the speaker system so it's that size for a good reason. The webcam that sits atop the display looks a little like an afterthought but that's presumably a side effect of it being an optional extra.

Lenovo offers various types of stands including the Full Function Monitor Stand we were supplied with. This offers 110mm of height adjustment, 90 degrees of clockwise pivot, 45 degrees of swivel to left and right as well as tilt from -5 degrees to +20 degrees. That's a very healthy range and the equal of any PC monitor.

At 420 x 190mm, the stand has a large footprint but most of the space is taken up by a shallow tray that is very handy for keeping stuff in. Without an hour of setting the system up the tray on ours housed a sound level meter, colorimeter, car keys, debit card, two pens, and a ruler. We're starting to wonder how we lived without it. The stand is attached to the cabinet by a conventional quick-release bracket that covers a 100 x 100m VESA mount.

Lenovo has gone to town with the ports, the M90a Gen 5 boasting no less than 6 USB-A ports – three 10Gbps ports on the right edge of the cabinet and three 5Gbps ports on the rear. Also on the right, you'll find a 3-in-1 SD card slot and a 3.5mm audio jack while the rear also hosts a 10Gbps Type-C port (data only), Gigabit Ethernet, HDMI 2.0 in/out and DisplayPort 1.4 output. There is space on the left side of the cabinet for an optical disc drive and at the back for a serial port but my review machine had a blanking plate across both slots.

What physical controls the M90a has are all on the right-hand side of the cabinet. At the bottom is a large power button while at the top is a microphone mute-button and a display switch that turns the unit into a dumb monitor for whatever is plugged into the HDMI port.

Oddly the M90a Gen 5 lacks the little 4-way joystick that is a feature of the M90a Gen 4 models and which lets you quickly change screen settings such as the brightness levels. With the Gen 5 model, you are left at the mercy of the display and audio settings available on the connected device.

The wireless modem in our unit was the ever-reliable Intel AX201 which supports 2.4/5Ghz Wi-Fi 6 and Bluetooth 5.3. The AX211 card which supports 6Ghz Wi-Fi 6E is a £20 optional extra. 

Lenovo ThinkCentre M90a Gen 5: Accessories

The webcam fitted to our review machine proved perfectly capable of handling the video and audio basics, shooting crisp and well-balanced 1080p, 30f/s video even in less-than-ideal light.

Our camera didn't support Windows Hello IR facial recognition, which is a £10 optional extra, while the model with a radar scanner that runs the smart human presence detection system will set you back a further £30. All camera models have a privacy shutter switch but it's rather stiff and not particularly easy to use.

Since the M90a Gen 5 isn't driven by a Core Ultra processor you don't get the Windows Studio Effects camera features but you can access the same set of features such as gaze tracking and background blur through the Lenovo Smart Meeting app.

The bundled keyboard and mouse are both cabled affairs and of a slightly higher quality than you usually get with AiOs. The keyboard especially is a bulky and solid unit with a full-sized numeric keypad and full-sized arrow keys. It lacks a backlight but it is otherwise ideal for anyone destined to spend serious time at their keyboard. The mouse is a basic affair but the Lenovo-red scroll wheel has a perfectly weighted ratchet effect and the click-action is impossible to fault.

Lenovo ThinkCentre M90a Gen 5: Performance

Intel's Core i5-13400 CPU is a solid performer thanks to its 10-core (6 Performance, 4 Efficiency) design and maximum 4.60GHz clock speed. Faced with the CineBench R23 multi-core CPU test it scored 11,797 while managing 305 points in ITPro's own 4K multi-media benchmark. 

The last all-in-one we tested, the considerably bigger and more expensive Intel Core i7-1360P-based Acer Aspire S32 scored 12,843 and 248 respectively giving the two machines broadly similar – and high – levels of performance in everyday tasks.

When it comes to graphics performance M90a Gen 5 is let down by its integrated graphics chip, the Intel UHD Graphics 730 iGPU, which boasts a paltry 24 execution units. 

This can't hold a candle to the Iris Xe iGPU with its 96 execution units let alone the new Arc chips that appear on the new Core Ultra SoCs. Case in point, the SPECviewperf 3dsmax 3D modelling test will run at around 16fps on an Iris Xe iGPU but the UHD Graphics 730 only managed a pedestrian 10.6fps.

Running the GFXBench Car Chase test offscreen produced a score of just 53.8fps which is the lowest score we've seen from a PC for a good number of years. The same goes for the 3DMark Time Spy benchmark which scored just 750 points to the Acer S32's 1,410. The takeaway here is that if you need good graphics performance, the M90a Gen 5 is not the place to look.

At least the M90a Gen 5 can consistently run both the CPU and GPU flat out without any thermal issues. We inadvertently left the stress test apps running overnight but come the morning both were still happily chugging along at 100% utilisation and the fan noise registered at 42.7dBA which is barely more than the ambient background noise level in the office. The transfer speeds of the 512GB Samsung-made SSD were good with sequential read and write speeds of 4,817MB/s and 3,724MB/s respectively. 

Lenovo ThinkCentre M90a Gen 5: Display 

While it's possible to have your M90a Gen 5 with a touchscreen for an extra £100, our unit came with the standard 1,920 x 1,080 IPS panel. The peak brightness of 290cd/m2 proved more than adequate for indoor use while the color gamut coverage was good for an office AiO with 104.7% sRGB, 74.1% DCI-P3, and 71.2% Adobe RGB. 

With a contrast ratio of 1065:1, Delta E colour variance of 1.9 vs the sRGB profile, and a Gamma of exactly 2.2 the screen looked both colorful and natural to the naked eye. There's no support for HDR content and the refresh rate is a standard 60Hz but we expected no more. Overall it's a surprisingly good display for an all-in-one.

Buried inside the cabinet below the screen are two 3W forward-facing Dolby Atmos-accredited speakers. They are not the loudest around, producing 73dBA as measured against a pink noise source at a 1m distance, but the sound they make is rich, tuneful, and underpinned by a solid bass line. 

If you want an all-in-one that can double up as a smart TV without the need to add external speakers then the M90a Gen 5 is perfectly suited to the job.

Lenovo ThinkCentre M90a Gen 5: Is it Worth it?

Just as long as graphics capability is not an issue the new ThinkCentre M90a Gen 5 is an impressive package. Its media playback ability is excellent thanks to the good-quality display and tuneful speakers while the plethora of ports make it very versatile. We were also happy to see that Lenovo has included the option to use the M90a Gen 5 as a dumb monitor which is handy for anyone who wants to plug something like a games console into it.

We'd call the M90a Gen 5 a solid value proposition too, especially given the vast amount of choice the buyer is faced with, including nine different processors, one or two SSDs of varying capacities, and an OS-free choice for those who prefer a Linux distro to Windows 11 (Home or Pro).

The plethora of options brings into stark contrast the one thing you can't change, the GPU. The m90a Gen 5 is screaming out for the option of even a relatively mild – and cheap – discrete GPU like the Intel Arc 530M. If you could make that choice the M90a Gen 5 would be a solid five-star system, but as it stands it's just a four, bordering on four-point-five.

Lenovo ThinkCentre M90a Gen 5 - Specifications

Last year’s Pure Accelerate was a flash frenzy. Riding high on the benefits of its ‘all flash’ approach, Pure Storage went all in on the sustainability and efficiency gains with flash as compared to traditional disk drives. 

Skip forward a year to Pure Accelerate 2024 and, while flash isn’t the central topic – AI has taken that title as with the majority of tech events – it's still top of mind for the firm. This is particularly true for the firm's founder and chief visionary officer John ‘Coz’ Colgrove, who has been banging the drum on the benefits of flash since founding Pure Storage alongside John Hayes purein 2009. 

From his pivotal position in the market, Colgrove is under no illusions as to the life cycle of flash, and he makes this clear in conversation with ITPro. Like any technology, Colgrove says, flash is bound to peak and return fewer efficiency and productivity gains.

“At some point in the future – far distant future – it will flatten out, because no technology goes forever,” Colgrove says. He adds it's important to think of the journey of disk technology here, by way of comparison. 

Disk storage “kept improving” for longer than anyone believed possible, Colgrove says, and only when it “flattened out and stopped improving” did there come an opportunity for new technology to take the helm, in this case flash. 

“Disk’s been on a 70-year improvement; every technology that’s tried to replace it had to get to a decent scale and then run this really gruelling race to try to catch up,” Colgrove says. “The disc out-distanced all of those until flash, and now flash is running faster than disk”.

Flash still has a long development curve

Right now, there’s no reason to think that flash is near the end of its usefulness. It’s clear flash is still improving at a rapid pace – with innovation driven by companies such as Pure Storage – and Colgrove honestly admits he has “no clue” what its eventual replacement will be.

“At some point beyond the horizon … flash will stop improving, or it will slow down,” Colgrove says. “At that point in time, that's when the next technology will have an opportunity to come along”.

That day is more than a decade away, Colgrove expects, with a real long-term storage alternative unlikely to emerge until the 2030s.

“The new technology is going to have to get to a decent scale and then run a long grueling race to catch up. When flash slows down, that's when you catch it from behind,” Colgrove adds. 

Crucially, there are still plenty of opportunities for flash to improve. “The flash guys keep figuring out ways to go in. Sometimes they shrink the cell size, sometimes they stack more layers,” Colgrove says. It's clear that flash still has a lot of avenues for innovation. 

Pure Storage’s role in flash going forward

Colgrove has some insights on Pure Storage’s future in flash specifically. “One of the things we do is we've pulled the flash management out of the direct flash module into the host, so we do it across all of the flash modules in the host”.

One of Pure Storage’s advantages, he says, lies in its ability to actively monitor decaying flash instances and to leverage insights into the write life of flash across an enterprise’s estate. This focus on oversight and storage simplicity has been key at Pure Accelerate 2024.

“It's really doing this across all the flash management that helps us be more efficient and gain a better life,” Colgrove adds. 

As  AI rears its head, Pure Storage may have to lean into the benefits of flash even more. For example, Colgrove is careful to point out the detrimental impact of AI on the environment and sustainability strategies. “AI destroys sustainability because the GPUs use so much power,” Colgrove says. 

Efficient storage is key here, Colgrove says, well-trodden ground for Pure Storage for which sustainability is a key issue. Much of the efficiencies to be gained in the storage space with regard to AI will come from more efficient data management when it comes to using data to train AI and for inferencing models.

While Pure Storage is responding to the training issue through its new AI-focused solutions, it’s a more niche concern, something Charlie Giancarlo, CEO at Pure Storage, comments on to ITPro. “Most of the training is going to happen in hyperscale clouds, a few very large tech companies, and GPU clouds, because I don't see ordinary companies wanting to do the training,” Colgrove says. 

Much of the firm’s gains can be made when it comes to inference, as companies are going to need responsive and scalable access to data to avoid unecessary data redundancy. Giancarlo says that on top of this, firms will seek to add AI access to that data on top of “all the normal access” required.

Flash can play a clear role here and Pure Storage can in turn take advantage of this demand.

“Efficient storage that is resilient, reliable, available, durable … becomes more important for that because your data’s more valuable,” Colgrove says. “The inferencing for us today is where the big opportunity for data is,” he adds, with Pure Storage’s smaller, direct flash offerings filling this niche.

Once a disruptive startup in a venture capital conference room, Pure Storage has grown into a global leader in data storage technology. Founded with the ambitious goal of transforming enterprise storage, the company introduced all-flash arrays that challenged conventional norms in the industry. Its pioneering efforts laid the foundation for a new era in data management, one where speed, scalability, and reliability are paramount. This early focus on innovation and performance has propelled Pure Storage into an indispensable role within the tech infrastructure of countless industries.

Over the years, the company's solutions have evolved to address complex demands across various sectors. Universities rely on their technology to support breakthroughs in research, while financial institutions depend on Pure Storage for robust systems that ensure secure and seamless transactions. Other prominent users include Sirius XM, which leverages the company's data platform to deliver uninterrupted streaming to millions of subscribers. Pure Storage is no longer just a storage solution; it has become a critical enabler of everyday operations in diverse fields.

As the world shifts toward artificial intelligence (AI) driven processes and sustainable technology, Pure Storage has positioned itself as a key player in this transformation. By aligning its offerings to support AI workloads, the company ensures that enterprises can harness the immense computational power required for machine learning applications. Sustainability also remains a cornerstone of its strategy, with initiatives to reduce energy consumption and environmental impact. Pure Storage is making strides not only in advancing technology but also in fostering a greener future.

This journey of growth and innovation has also brought financial success. With a market cap fluctuating between $16 and $22 billion in 2025, and annual revenue reaching $3.2 billion in the 2025, an impressive increase from $2.8 billion the previous year Pure Storage exemplifies the resilience and potential of modern tech companies. From its humble beginnings to its current status as an industry leader, the company's story is a testament to the power of vision and dedication in reshaping the technological landscape.

Here is a broad overview of the path the company has taken to become one of the most prominent names in storage, with a market cap of $21.75 billion to match.

A quick history of Pure Storage

Pure Storage was founded with venture capital funding from Sutter Hill and an ambitious goal: to define the "Modern Data Experience." Its introduction of flash storage solutions, such as FlashArray and FlashBlade, revolutionized enterprise environments, particularly data centers. By addressing the inefficiencies of traditional hard disk drives (HDDs) with modern solid-state drives (SSDs), Pure Storage established itself as a pioneer in data storage innovation.

The company entered the market at a time when businesses were transitioning to more advanced storage technologies but lacked the infrastructure to support significant change. John 'Coz' Colgrove, a founding engineer from Veritas, and John Hayes, a former Yahoo CTO, identified this technological gap, termed the "Flash Chasm," and built solutions to bridge it. Their vision enabled Pure Storage to become a significant player in the enterprise storage market.

Pure Storage went public in 2015, further cementing its influence in the industry. That same year, the company settled a patent dispute with EMC for $30 million, which demonstrated its willingness to protect its intellectual property and market position.

The annual Pure//Accelerate conference, launched in 2016, became a signature event to showcase its latest advancements, including its focus on AI and cloud integration. These conferences continue to highlight Pure's strategy to align with emerging technological trends.

As the importance of generative AI has grown, Pure Storage has positioned itself as a critical infrastructure provider for these demanding workloads. Its flash storage systems are designed to scale to the petabyte levels required for AI applications, supporting businesses in their digital transformation journeys. CEO Charles Giancarlo has frequently emphasized how Pure's solutions cater to the unique storage needs of AI, which demand speed, scalability, and sustainability.

To reinforce this position, the company has made strategic investments in AI infrastructure, such as its partnership with CoreWeave, an AI cloud provider. By combining cutting-edge technology with a focus on sustainability, Pure Storage continues to meet the evolving needs of enterprises while maintaining its leadership in the competitive storage market.

What does Pure Storage sell?

Pure Storage provides a wide range of flash storage solutions designed for data centres, complemented by software and platform tools that optimize data storage and usage. Its product lineup includes the FlashArray series, ideal for block storage, and FlashBlade, which caters to file and object data.

The company has placed significant emphasis on delivering its offerings as a service. Customers can choose cloud storage options tailored to private cloud, public cloud, and hybrid cloud. environments, including integration with AWS and Azure. A standout offering, the Evergreen subscription model, ensures seamless upgrades to both hardware and software, fostering sustainability and prolonged hardware lifespans. This service model competes with the likes of HPE GreenLake and Dell Apex.

Recent developments have introduced features aimed at making storage systems more intelligent and automated. During its annual conference, the company unveiled Evergreen//One for AI, a pioneering storage-as-a-service product for AI applications, and an AI-driven assistant for storage management that leverages insights from customer systems to streamline operations.

Pure Storage works at the enterprise level across the private and public sectors, having worked with ScotGov to overhaul its storage approach and embrace containers, and with Mississippi’s tax authority to support critical apps.

Beyond physical arrays, Pure Storage offers storage and disaster recovery as a service products, data security solutions, AI-focused offerings, and its ‘Storage-as-Code’ options. Its tagline is “uncomplicate data storage, forever”.

Its consumption-based storage subscription service Evergreen competes with the likes of HPE GreenLake and Dell Apex.

Pure Storage's mergers and acquisitions

Before its initial public offering in 2015, Pure Storage had already secured substantial funding, including one of the largest investment rounds in its sector at the time. The company now trades on the New York Stock Exchange under the ticker PSTG.

Over the years, Pure Storage has strategically acquired several companies to strengthen its portfolio. These include:

• StorReduce, a software company, for an unannounced amount in 2018
• Compuverde, a Swedish software company, for an undisclosed amount in 2019
• Portworx, a Kubernetes data services platform, for $370 million in 2020

Doing Business with Pure Storage

Based in Silicon Valley, Pure Storage serves prominent clients such as Ford, Meta, JP Morgan Chase, and NASA. Its operations extend globally, with offices across Europe, Latin America, and the Asia-Pacific regions.

In 2025, the company revamped its partner program to enhance profitability and provide partners with tools to drive customer adoption of its platforms, particularly for artificial intelligence and cyber resilience. The updated program includes an automated system for managing service agreements and new incentives to encourage the replacement of outdated disk storage systems.

From an internal policy perspective, Pure Storage remains committed to its goal of achieving net-zero greenhouse gas emissions for its direct operations by 2040. A key selling point for the company is the sustainability of its products; it claims its all-flash solutions can significantly reduce data center energy consumption by up to 85% compared to competitor systems.

The company's 2024 ESG report highlights its commitment to helping customers lessen their environmental impact through reduced energy and space usage, as well as minimizing e-waste. Pure Storage plans to submit its science-based emissions reduction targets for approval in July 2024.

Windows all-in-ones (AiOs) may be the PC product type most difficult to get excited about. Designed as mundane and functional items, no major manufacturer has created anything half as interesting as Apple's 24-inch iMac. That's a darned shame because all you'd need to do is package the components in a thin case, fit a 4K display, and let customers choose from a selection of colors beyond black.

That brings us neatly to Acer's Aspire S32, a Windows AiO available in any color you want, as long as it's black. Built around a mobile Intel Raptor Lake CPU and aimed at the general home and business users, the S32, as the name suggests, is somewhat larger than the average AiO in this price bracket, with 27-inch displays being more common.

The Aspire S32 also features a range of modular accessories, including a monitor light, two webcams, and a detachable touchpad made from ocean-bound recycled plastic.

Acer Aspire S32-1856: Design

While the S32 won't keep the design team at Apple awake at night, it's a decent design for something made entirely of black plastic. The monitor part of the affair is only 10mm thick, thanks to all the PC gubbins built into the stand rather than the monitor cabinet. The bezels around the display are also unusually thin, 8mm on the top and sides and just 10mm at the bottom. The stand's base is thick and bulky, but that's because it houses a detachable 210 x 110mm touchpad and a fingerprint scanner. 

The S32 weighs 11 kg, which makes it easy enough to move around; however, because the PC hardware is in the stand, there's no option to mount it on a VESA bracket. The base also has a Kensington lock at the back for theft prevention.

The support pillar houses all the ports and controls. On the right side, you'll find two 5Gbits/sec USB-A ports above the power button, while on the opposite side is an SD card reader, a 10Gbits/sec Type-C port that also supports DP Alt Mode video output and 100W charging, and lastly, an input selector for when you use the S32 as a dumb monitor.

On the back of the pillar are two 10Gbits/sec Type-A ports and two HDMI ports, one for input so you can use the unit as a monitor and one for output. The rear is also home to the DC-in socket and a Gigabit Ethernet port. That's a solid selection, though the layout means none of the ports are easy to access, especially those on the back of the pillar. Wireless communications – WiFi 6 and Bluetooth 5.3 – are handled by Intel's ubiquitous AX201 card.

One area where the S32's design fails is adjustability, with only tilt (between -10° and +20°) available. The absence of swivel or pivot movement isn't an issue, but the lack of height adjustment is, especially when using the top-mounted webcam. To put ourselves anywhere neat center screen, we had to tilt the display forward as far as it would go or push back quite a way from the desk. 

Acer Aspire S32-1856: Accessories

The S32 comes with a standard plastic 2.4Ghz wireless mouse and keyboard. Just plug the USB dongle into one of the S32's USB-A ports, and you are ready to rock and roll. The keyboard lacks a backlight, but the white-on-black keycap graphics are easy to read, even in low light. 

The keyboard has a full-sized numeric keypad and cursor keys, making it perfect for data entry. Overall, we found the mouse and keypad a pretty decent package and had no complaints about the typing action.

The other bundled accessories are rather more interesting. Our review sample included a clip-on light bar and a 1440p webcam. Both accessories are attached to the monitor cabinet by a magnetic pogo-pin clasp. The magnet attraction is strong enough to locate without the need to get up and look for the mount; you can do it by touch.

The 35cm long light bar has adjustable brightness levels and hue, so you can select anything from a low beige to a searing white. The webcam, meanwhile, produces crisp 1440p video and has a three-stage ring light to keep you well-lit. Usefully, both accessories can be attached to the S32 via a Type-C cable if you want to use them in a location other than the top of the cabinet.

The swappable webcams (a 1080p model without floodlight) also feature Acer's EyeSense system, which blurs and/or dims the screen when you move away and lets you replace yourself with an avatar in video calls.

The detachable touchpad makes a great little graphics tablet for impromptu creative work or handwriting input with the bundled active stylus pen. The top of the touchpad also lets you adjust the system volume and brightness and summon an on-screen keyboard. Like the webcam and light, the touchpad can also be connected via a Type-C cable for prolonged use away from its docked position.

The touchpad has two modes of use, toggled by a lock switch on the front. Unlocked, it works like a graphics tablet but locked, it reverts to a conventional multi-touch trackpad. This is handy for controlling your PC if you are kicking back in your chair rather than sitting up at your desk.

Acer Aspire S32-1856: Performance

Our review sample wasn't the latest word in S32 modernity, lacking the 6Ghz Wi-Fi, 1TB SSD, 32GB of RAM, and Iris Xe iPGU Acer offers. Instead, we had 512GB of SSD storage divided into two 256GB partitions, 8GB of RAM, 2.4/5GHz Wi-Fi, and the Intel UHD iGPU. All iterations of the S32 use the same 12-core (4 Performance, 8 Efficient) mobile Core i7-1360P processor.

We cannot see any way of accessing the S32 internals. So ensure you get the specification you want when you purchase it, or you'll be stuck with it. That's par for the course with AiOs, no matter the manufacturer.

The Aspire S32 scored 248 points in our in-house 4K multi-media benchmark and 5,670 in the PCMark10 productivity benchmark. That means it has enough potency to chew through everyday computing tasks quickly. Those scores also put the Aspire S32 comfortably ahead of the HP All-in-One 27 we tested earlier in the year.

Graphics are, of course, a different matter. The Intel UHD iGPU only scored 1,410 in the 3DMark Time Spy test, meaning this is not a system for graphically intense workloads like 3D modeling, let alone 3D gaming of any serious description.

The S32 is actively cooled, though we can't tell you how many fans it has. The cooling system is good enough to let the system run the CPU and GPU at 100% for prolonged periods without overheating or the fan(s) making an intrusive noise.

Acer Aspire S32-1856: Display

The S32's 31.5-inch screen has a resolution of 2,560 x 1,440, which results in a pixel density of just 93dpi, considerably below the 137dpi you get from a FullHD 16-inch laptop or the 214dpi the iMac delivers. This means that unless you are sat a long, long way away from the screen, you will notice that the image isn't as crisp or as sharp as you are probably used to on your laptop, let alone smartphone.

That's the bad news. The good news is that the display is bright enough, peaking at 340cd/m2, and usefully colorful, with gamut volumes of 132.9% sRGB, 94.1% DCI-P3, and 91.6% AdobeRGB. There's no option to lock the screen to any industry-standard color profiles. Still, the Delta E variances of 2.45 vs sRGB and 2.8 vs AdobeRGB are not too bad and mean the general user is presented with a fairly wide, well-balanced, and accurate color palette.

This being a bog-standard IPS screen with no HDR capability, you'll have to make do with a rather middling contrast ratio of 950:1. The visual daylight temperature was a little on the warm side at 5526K. Gamma was fine at 2.12, but motion fidelity was poor, with lots of ghosting and smearing visible in testing. The last is hardly unexpected, given the 60Hz refresh rate and lack of pretensions to gaming capabilities.

Our review machine had a Micron SSD installed, and it proved to be a reasonable rather than exceptional performer. It recorded sequential read and write speeds of just over 3,000MB/s, which is more than adequate for a PC of this type.

Acer Aspire S32-1856: Is it worth it?

Acer will relieve you of £1,599 in VAT (£1,332 ex VAT) for the Aspire S32. That's for the model with a 1TB SSD, 32GB of RAM, and Iris Xe iGPU, making it more capable than the model we were sent for review. Acer's product page does not mention the detachable webcam and downlight, but Acer UK has confirmed that you will find them inside the box.

For the money, it's a comprehensive package, with the detachable touchpad being a unique feature that adds some genuinely enhanced usability. The only serious criticism we can level is that given the 32-inch screen size, a resolution closer to 4K would be useful for anyone wanting the highest degree of clarity and sharpness. That said, given the price, WQHD is the best you can seriously hope for.

Acer Aspire S32-1856 specifications

Magnetic tape storage might be one of the oldest storage technologies, but it appears to have found a new lease of life thanks to the generative AI boom.

A record 152.9 exabytes of total (compressed) tape capacity was shipped in 2023, according to the report from the companies behind the Linear Tape-Open Technology (LTO) program, namely Hewlett Packard Enterprise (HPE), IBM, and Quantum.

That’s growth of just over 3% compared to 2022, which the report puts down to rapid growth in data generation and the increased infrastructure needs of hyperscaler cloud providers and enterprise customers.

It seems that tape is turning out to be useful for storing the huge amounts of unstructured data that result from the adoption of new technologies such as AI.

“As AI and related technologies cause increases in storage requirements and costs, LTO tape technology makes the management of vast amounts of data more cost-effective and sustainable while delivering comprehensive protection against data breaches, which continue to grow in number and sophistication,” the report said.

The vast majority of enterprise data is unstructured and often ignored; however, the rise of generative AI has made this sort of data potentially extremely valuable, either for training AI models or to mine through with AI to find insights.

Some tech investment firms have even suggested that desire to get hold of this unstructured data could lead to a corporate acquisition boom. Either way, finding a way to host these gigantic datasets has become more important for enterprise customers.

Why magnetic tape storage is still in vogue

Magnetic tape has been used to store data since the 1950s, although the LTO format itself only dates back to the 1990s. 

Even if it’s an old technology, it is widely used for good reason. Tape is considered to be low cost, highly portable, and can store very large amounts of data. It’s an extremely durable storage choice which can last for up to 30 years.

LTO also features error correction algorithms with an error rate of 1 in 10^19. This, the LTO companies said, would mean it would take a whole year, with 130 tape drives writing data continually, to encounter an error that could not be fixed by LTO’s error correction technology.

The current, ninth generation, technology can store up to 18 terabytes (TB) of data on a single cartridge (45TB compressed); the LTO roadmap plans to increase capacity up to 144TB by generation 12. Since the first LTO products arrived, over 5.6 million drives, 371 million cartridges and 517,958 billion GB of media capacity have been shipped, the companies said.

The companies argue that tape has an 86% lower total cost of ownership than disk and a 66% lower total cost of ownership with cloud storage over a ten-year basis. 

The major disadvantage of tape is, of course, that it’s much slower at transferring data which is why it is often used for offline backup situations where reliable storage is more important than speed.

“The probability of losing data through degradation, however, is far less than it being maliciously encrypted in a ransomware attack, or destroyed by fire, flood or acts of war in the data center,” the companies note.

Indeed, one of the other reasons for the ongoing popularity of tape is as a backup technology, something that has become essential as a result of the rise of ransomware. Many companies will use the 3-2-1, or even the 3-2-1-1 backup rule, which means having at least three copies data stored on two different types of storage mediums, one of which is off-site and (for addition peace of mind) one that is off-line: tape is usually the ‘1’ in either case.

“LTO tape will endure as a critical component of storage architectures across the enterprise, especially as tape technology itself continues to improve to handle even larger volumes of data and more intensive workloads,” said Tom Coughlin, storage analyst and president of Coughlin Associates, quoted in the LTO companies’ press release

“The latest annual tape capacity shipment media report signifies the continued importance of LTO tape technology in the digital storage hierarchy not only today, but also far into the future as storage needs evolve in complexity and cost,” he said.

With the exception of Apple's colorful 24-inch iMac, All-in-Ones are mundane and functional items bought for value and convenience rather than performance or style.  Previous Windows All-in-Ones that we've looked at include Acer's C24 and MSI's AM241, perfectly effective workaday tools but utterly devoid of inspiration and cursed by low-rent displays and speakers: Call center fodder, nothing more.

HP's new All-in-One 27 (officially called the cr0014na) is a cut above the Acer and MSI machines. If not quite up to iMac levels of desirability and capability, it looks, feels and performs like something you might buy yourself rather than have to endure because the IT department got a good deal from the manufacturer.

A "good deal" is actually what I'd call HP's pricing of this 27-inch AiO. The usual price is £899 inc VAT, but it's currently on offer for £820 inc VAT or £683 ex. VAT. The HP may not match the iMac for looks or performance, but it doesn't have to because it's £600 cheaper.

HP All-in-One 27: Design

By the standards of Windows All-in-Ones, the HP 27 is a looker thanks to its all-white livery and silver stand. If you look closely at the body, you'll see it has a flecked finish not unlike that used on Philip's new Evnia gaming monitors. It's a good look, and we like it a lot. 

At 32mm thick in the middle, it's much thicker than the iMac, but the top and side bezels are a mere 7mm thick, and the chin at the bottom is not excessive at 24mm. The whole thing feels very solidly built, and at 6.72Kg it's not too difficult to move around.

The stand connects to the PC by way of a bespoke quick-release bracket, and there is nothing in the box to facilitate attaching the body to a VESA mount, so you'll need to buy an adapter like this. The stand doesn't offer much adjustment with no rotation or pivot, just tilt between -5 and +20° and 100mm of height adjustment. The top of the impressively small base is coated in a felt material, which stops things sliding off and offers a soft surface if you want to put your phone screen down on it. 

The HP has a wide range of ports with a single Type-C and two Type-A ports supporting 5Gpbs data and another two Type-As running at v2.0 speeds. The Type-C port does not support video output, for that, you must rely on the HDMI 1.4 output. A 3.5mm audio jack and a Gigabit RJ45 Ethernet port round things out. The only control on the cabinet is a small on/off button in the bottom right corner of the back.

We only have one issue with the design: all the ports are on the back toward the bottom and spread on both sides of the stand bracket. If you continually swap out USB dongles or cables, you will soon start to curse the absence of side or front-mounted ports.

For wireless comms, you get 2.4/5Ghz W-Fi 6 (6Ghz would have been a stretch at this price point) and Bluetooth 5.3, which is handy if you plan on listening to music wirelessly, thanks to its improved support of low latency audio streams.

According to the specification sheet, there are two expansion slots inside, one for an SSD and one for a WLAN card. As there is no way to access the internals, we can't tell you if those two slots are in addition to the ones already occupied by the SK Hynix SSD and Realtek wireless card. We would guess that they are not and that the contents of those slots and the two SODIMM RAM mounts must be specified prior to purchase.

HP All-in-One 27: Accessories

The peripherals that ship with Windows all-in-one PCs are usually pretty grim, but the keyboard and mouse HP has bundled are not too shabby. To start with, they are all-white to match the main unit, so they don't look cheap. Secondly, they are wireless and connect to a single 2.4 GHz Wi-Fi dongle, making setup and connection a cinch and keeping your desk cable-free.

The mouse is a large, chunky, ambidextrous affair with a positive click-action and a well-calibrated scroll wheel. It's battery-powered, but HP bundles a Duracell AA for it along with a brace of AAAs for the keyboard.

The keyboard is also on the bulky side, but the typing action is good, the layout is faultless, and it has a numeric keypad and full-sized cursor keys. As a typing tool, it gives surprisingly little away to the Cherry keyboard we use, though it is rather noisier and lacks anything in the way of a backlight.

The HP 27's party piece is a concealed tiltable webcam that retracts into the top of the cabinet. It's a good quality 1080p affair that captures a sharp and colorful image even in low light and supports Windows Hello facial recognition.

The camera is mounted on a swivel, so you can adjust the vertical angle to around 20° below the horizontal, which is handy if you want to use the camera while the stand is at maximum height.

HP All-in-One 27: Performance

The HP 27 is built around an Intel Core i7-1355U processor, a 10-core, 12-thread component with a maximum clock speed of 5GHz. Accompanying that, you get 16GB of dual-channel but rather slow DDR4-3200MHz RAM, a 512GB SK Hynix SSD, and Intel's integrated XE graphics solution.

The HP 27 scored 5,657 in the PCMark10 benchmark and 170 in ITPro's in-house 4K multi-media benchmark. Those numbers are broadly what you should expect from a high-end ultra-compact laptop like the Asus Zenbook S 13. 

The Cinebench R23 scores of 1,709 single core and 6,799  multi-core were also similar to the Zenbook S 13 and highlight the main drawback of this U-series i5 chip: It only has two performance cores. That means it is enough power for everyday computing, entertainment, and admin tasks, but not enough for more demanding tasks or gaming.

While the HP may be short of out-and-out grunt, it runs quietly, even under heavy stress. Using the FurMark and Prime 95 apps to put the maximum stress on the CPU and the GPU, the cooling fans were inaudible unless we put an ear right up against the rear of the cabinet. Of course, part of the reason was that the CPU utilization dropped to 60% after a few minutes and then stayed there.

The SK Hynix SSD turned in solid rather than spectacular results with average read and write speeds of 3820MB/s and 2,415MB/s, respectively. Those are more than adequate speeds for general home or office usage.

HP All-in-One 27: Display and Audio

The display is not the sharpest you'll ever see. A 1,920 x 1,080 matrix combined with a 27-inch diagonal results in a lackluster 81.6dpi perilously close to a third of the 218dpi the iMac offers. Look closely at the HP 27's display, and you can discern the pixel structure. Things improve when you increase the Windows scale to 125% or 150%, but the HP 27 is still not an ideal choice if you want to look closely at big spreadsheets or pages of small copy.

In other areas, the HP 27's IPS panel makes a much stronger case for itself. Our colorimeter recorded the maximum brightness at a healthy 327cdm/2, though the contrast ratio turned out to be 1023:1, a little low thanks to a rather high black level luminance of 0.32cd/m2. There is plenty of color available with gamut volumes of 107.5% sRGB, 74.1% Adobe RGB, and 76.1% DCI-P3.

Measured against the sRGB profile, the Delta E came in at 2.06, which is a commendable result for what is, at the end of the day, a budget IPS display. Anything at or below 2 means the screen is good for color-critical work just as long as you don't require wide gamut support.

Watching our usual slate of test videos on the HP 27, the screen was nicely saturated and impressively colorful, though there is no support for HDR content. Given the mundane 60Hz refresh rate, motion handling was poor, but as this is certainly not a gaming rig, we can let that slide.

Out of the box, the panel's gamma was 2.24, and the visual daylight temperature 6496K, both close enough to the ideals of 2.2 and 6500K to be practically perfect. We expected the 2 x 2W speaker system to be pretty poor, but we could not have been more wrong. There was ample volume (73.5 dBA as measured against a pink noise source at a 1m distance) and plenty of bass combining to create a detailed, punchy, and melodic soundscape. 

HP All-in-One 27: Is it Worth it?

When we pulled the HP All-in-One 27 out of its box, we expected another dull-as-dishwater Windows AiO that could only justify itself in the most brutally pragmatic fashion. Imagine our surprise then when it turned out to be a rather stylish and practical answer to the computing needs of many home and office users.

The main issue is the FullHD 27-inch screen. If you plan on looking at small text, 81.6dpi is a bit on the low side. Sadly, HP doesn't make a QHD version of this model, but there is a 23.8-inch version identical in all ways apart from the screen size, which gives a sharper 92.5dpi. At £779 inc VAT, it's not a lot cheaper, though.

Resolution aside, the display is good, while the speakers are excellent. Performance is also more than acceptable given the tasks the HP 27 is likely to be employed in, and the peripherals it ships with are better quality than you might expect. 

For the asking price, it's hard not to see the HP 27 as good value and the perfect PC for anyone wanting a simple, compact, capable, and stylish one-box solution to their general computing needs at home or at work. 

HP All-in-One 27 specifications

Global distribution giant TD Synnex has signed a deal to distribute the full range of Nvidia enterprise software and accelerated computing products in the UK. 

The new agreement means channel partners can now tackle their Nvidia product and platform requirements through TD Synnex, including Nvidia AI Enterprise and Nvidia Omniverse Enterprise software licenses for developers.

Businesses will also have access to the chipmaker’s high-performance networking adapters, accelerated computing GPUs, and superchips.

The distributor is also applying its accelerated delivery, flexible credit, and finance options, as well as its technical and professional services to the Nvidia range to help customers maximize their sales potential.

In an announcement, Simon Bennett, managing director of TD Synnex’s Advanced Solutions UK and Ireland business, said the company will provide the necessary support to help partners optimize Nvidia products within their infrastructures.

“Many organizations today depend on complex, hybrid infrastructures that encompass multiple cloud and on-premises resources,” he said.

“Nvidia technologies help drive the next-level graphics, HPC and AI performance essential for these intensive workloads.

“We will be working closely with Nvidia to make sure that our team is ready to provide partners with the technical and professional support and services they need to optimize Nvidia’s high-performance solutions, from the cloud to the data center and right out to the edge.”

A prominent global IT distributor and solutions integrator, TD Synnex serves over 150,000 customers across more than 100 countries around the world.

The firm’s portfolio pulls together IT products, services, and solutions from over 1,500 technology vendors, covering high-growth tech segments such as cloud, cyber security, big data and analytics, AI, IoT, mobility, and everything as a service.

With its new Nvidia distribution agreement, TD Synnex said it is expanding its expertise to meet UK partners’ full range of Nvidia-related needs, whether deploying solutions as part of their own designs or as a part of other systems’ brands.

“Nvidia technologies are used within the solutions from many leading vendors that we partner with, such as Dell, HPE and Lenovo, and now, with the extended distribution agreement, we can also provide the full Nvidia portfolio directly to partners as well,” Bennett added.

Content creators seeking a powerful storage system for their video masterpieces may find Qnap's TVS-h874T could be just the ticket. Proudly claimed by Qnap as the world's first Thunderbolt 4 NAS, this 8-bay desktop box sports a pair of these 40Gbits/sec ports for those who want fast access for 4K video production.

There's plenty of power to hand as Qnap offers a choice of Intel's Gen12 Core i7 or Core i9 CPUs. We have the cheaper option on review which sports a 12-core Core i7-12700E CPU offering eight 2.1GHz Performance cores, four 1.6GHz Efficient cores, and an integrated Intel UHD Graphics 770 GPU.

The appliance comes with 32GB of DDR4 memory upgradeable to 64GB – an important feature as the 'h' in the model name indicates Qnap is aiming this NAS primarily at its QuTS hero OS. This is more demanding than Qnap's QTS as it requires at least 16GB for its inline data deduplication feature and 32GB to get the best performance from 10GbE upgrades.

Qnap TVS-h874T review: Hardware features

The appliance is encased in a sturdy steel chassis with solid plastic lockable tool-free drive carriers and removing the metal lid reveals a tidy interior with Qnap's QXP-T42P Thunderbolt 4 card fitted in the lower of the two PCIe slots. Space is limited here as accessing the upper X16 slot requires an L-shaped screwdriver, otherwise the power supply will have to be removed to gain access to its bracket screw. 

The motherboard sits in the right side of the chassis with the CPU fitted with a large finned passive heatsink and partnered by a small cooling fan further back. General chassis cooling is handled by two 12cms diameter rear fans and during testing, we were impressed with how quiet the appliance was making it a great choice for small offices and home workers.

The 32GB of base memory is supplied as two 16GB DDR4 SO-DIMM modules so any upgrades will require one or both of them to be replaced and for reasons that escape us, Qnap has placed warranty stickers over the edge of each stick stating it will be voided if they are removed. Below the memory modules is a pair of easily accessible M.2 2280 NVMe SSD slots and these are the PCIe Gen 4 variety.

Qnap TVS-h874T review: 10GbE NAS and IP SAN performance

We fitted a pair of Kingston 480GB M.2 NVMe SSDs and after following the web console's wizard to install QuTS hero, we used them as a mirrored array for the initial system pool. For general data storage, we fitted a quartet of 22TB Western Digital Red Pro NAS drives and used the Storage & Snapshots app to create a big 58TB RAID5 pool. 

For our 10GbE performance tests, we installed a dual-port 10GbE SFP+ fibre adapter in the appliance and called up the lab's HPE ProLiant DL380 Gen11 rack server for host duties. This was equipped with dual 32-core 2.1GHz Xeon Scalable Gold 6430 CPUs, 256GB of DDR5 memory plus a dual-port 10GbE SFP+ adapter and running Windows Server 2022. 

With a NAS share on the RAID5 array mapped over 10GbE to the server, we recorded top Iometer sequential read and write rates of 9.3 and 9.2Gbits/sec, and swapping over to random operations returned very similar results. These numbers highlight an important feature of QuTS hero as its ZFS file system's ARC (adaptive read cache) and ZIL (ZFS intent log) features are already very efficient making SSD caches unnecessary. 

Moving over to IP SANs delivered equally impressive results with a 1TB target recording high sequential and random read and write rates of between 9.3Gbits/sec and 9.1Gbits/sec. We ramped up the pressure with a dual-10GbE MPIO link to the target and watched Iometer sequential and random read rates increase to 18.2Gbits/sec and 17.8Gbits/sec with write operations settling at 18Gbits/sec and 17.7Gbits/sec respectively.

I/O throughput for the MPIO iSCSI connection was also excellent. Dropping to Iometer 4KB block sizes delivered sequential read and write rates of 210,700 and 206,800 IOPS and changing to random operations saw read and write throughputs of 202,300 and 162,600 IOPS. We noted during all our tests that the QuTS resource monitor showed CPU utilization never going above 12 per cent so we can't see it having any problems delivering Thunderbolt 4 speeds

Qnap TVS-h874T review: QuTS hero features

QuTS hero offers superior data protection features to QTS with its ZFS copy-on-write delivering near unlimited NAS and iSCSI LUN snapshots along with inline data deduplication and compression. End-to-end checksums handle self-healing of data corruption while the SnapSync app provides fast snapshot backups to remote QuTS appliances using block-level replication run to a schedule or in real-time. 

Along with the standard RAID arrays, you can also opt for triple parity RAID to protect against three drive failures or triple mirroring which stores identical copies of data on three drives. A choice of two WORM (write once read many) policies can also be applied to NAS shares during creation to protect their data from tampering or accidental deletion.

App choices for both OSes are equally impressive and include key backup utilities such as Hybrid Backup Sync (HBS) 3 which helps create a 3-2-1 appliance protection strategy and now supports the new myQNAPcloud hosted storage. Virtualization is on the menu with the Container Station, Ubuntu Linux Station, and Virtualization Station apps all present while Hyper Data Protector keeps your VMware and Hyper-V VMs safe by backing them up regularly.

Qnap TVS-h874T review: Is it worth it?

With the model on review costing around £2,700 excluding VAT, the TVS-h874T represents a considerable investment but its powerful Core i7-12700E CPU provides a good return as performance is extremely good. The appliance's high memory capacity makes it a perfect platform for Qnap's QuTS hero OS which offers a wealth of enterprise-class data protection features and it's a good choice for video editors that want super-fast Thunderbolt 4 ports and plenty of storage potential. 

Qnap TVS-h874T specifications