AWS ditches Nvidia for in-house 'Inferentia' silicon

Alexa queries and facial recognition data will now be processed by Amazon's own chips

An Amazon Eco Dot being pressed by a user

Amazon Web Services (AWS) will ditch Nvidia chips responsible for the processing of Alexa queries and will instead use its own in-house silicon, the company confirmed on Friday.

The cloud giant will also be shifting data processing for its cloud-based facial recognition system, 'Rekognition', over to these in-house chips, according to Reuters.

Alexa queries, issued through Amazon's Echo line of smart speakers, are sent through the company's data centres where they undergo several stages of processing before coming back to users with an answer, including translating the processed text into audible speech.

The company said that the "majority" of this processing will now be handled using Amazon's own "Inferentia" computing chips. These were first launched in 2018 as Amazon's first custom silicon-designed chips for accelerating deep learning workloads.

Amazon has said that the shift to Inferentia for Alexa processing had resulted in a 25% latency boost and 30% lower cost. The firm hopes the same will happen with its Rekognition system, which has also started to adopt the Inferentia chip.

The cloud giant didn't specify which company previously handled Rekognition processing, but the service has come under some scrutiny from civil rights groups for its involvement with law enforcement. Police were temporarily banned from using it earlier in the year, following the Black Lives Matter protests.

Nvidia and Intel are two of the biggest providers of computing chips, often for data centres, with companies like Amazon and Microsoft included in their clientele. However, a number of firms have begun to move away from vendors and are bringing the technology in-house. For example, Apple has recently moved away from Intel chips in favour of the A14 Bionic processors, which will be used going forward.

Featured Resources

Managing security risk and compliance in a challenging landscape

How key technology partners grow with your organisation

Download now

Evaluate your order-to-cash process

15 recommended metrics to benchmark your O2C operations

Download now

AI 360: Hold, fold, or double down?

How AI can benefit your business

Download now

Getting started with Azure Red Hat OpenShift

A developer’s guide to improving application building and deployment capabilities

Download now

Recommended

How to become a machine learning engineer
Careers & training

How to become a machine learning engineer

23 Dec 2020
Data science fails: Building AI you can trust
Whitepaper

Data science fails: Building AI you can trust

2 Dec 2020
MLOps 101: The foundation for your AI strategy
Whitepaper

MLOps 101: The foundation for your AI strategy

2 Dec 2020
Realising the benefits of automated machine learning
Whitepaper

Realising the benefits of automated machine learning

2 Dec 2020

Most Popular

IT retailer faces €10.4m GDPR fine for employee surveillance
General Data Protection Regulation (GDPR)

IT retailer faces €10.4m GDPR fine for employee surveillance

18 Jan 2021
Citrix buys Slack competitor Wrike in record $2.25bn deal
collaboration

Citrix buys Slack competitor Wrike in record $2.25bn deal

19 Jan 2021
Should IT departments call time on WhatsApp?
communications

Should IT departments call time on WhatsApp?

15 Jan 2021