AMD and Microsoft cement relationship with cloud collaborations

Microsoft is offering new instances on its Azure cloud platform optimized for artificial intelligence (AI) and high-intensity workloads that are powered by chips from AMD.

The Azure ND MI300X v5 virtual machine (VM) offers 1.5TB of high bandwidth memory (HBM) – described as “unprecedented” by Microsoft – and is powered by eight AMD Instinct MI300X GPUs interconnected via Infinity Fabric 3 in each VM.

This architecture allows customers to process large AI models and is, according to Microsoft, “ideal for AI applications that need to quickly process vast amounts of data”.

“ND MI300X v5 has the most HBM capacity available, 1.5TB per VM or 192GB of HBM per GPU, the highest HBM capacity available in the cloud,” said Marc Charest, program manager at Microsoft, in a blog post. “This lets customers run the largest, most advanced AI models faster, and with fewer GPUs. In the end, you save power, cost, and time to solution.”

Vamsi Boppana, senior vice president for AI at AMD added: "AMD is fostering AI everywhere – from the cloud, to the enterprise and end point devices – all powered by our CPUs, GPUs, accelerators and AI engines."

"Together with Microsoft and a rapidly growing ecosystem of software and hardware partners, AMD is accelerating innovation to bring the benefits of AI to a broad portfolio of compute engines, with expanding software capabilities."

In addition to the Azure ND MI300X v5 virtual machine, Microsoft and AMD have also collaborated on a series of new memory-optimized, compute-optimized, general purpose VMs based on the chip maker’s 4th generation EPYC CPUs.

Dalsv6, Dasv6, Easv6, Falsv6, Fasv6, and Famsv6, as they are called, will be made available in public preview during the first quarter of 2024.

“These new VMs offer improved price/performance versus the prior Dasv5 and Easv5 VMs and NVMe connectivity for faster local and remote storage access,” said Lauren Britton, Senior Product Manager at Microsoft, in a blog post. 

According to AMD, this means up to 20% better performance for both general purpose and memory-intensive VMs, and up to twice the CPU performance for compute-optimized VMs compared to the previous generation.