Nov. 15, 2023 — Right now at Microsoft Ignite, AMD (NASDAQ: AMD) and Microsoft featured AMD merchandise, together with the upcoming Intuition MI300X GPU accelerator, EPYC and Ryzen CPUs with AI engines.
Microsoft’s new Azure ND MI300x v5 Digital Machine (VM) sequence optimized for AI workloads options the MI300X, making Azure the primary cloud to make use of the brand new accelerator. The brand new VMs are a part of Microsoft’s infrastructure supporting AI for enterprises in all places giving prospects extra selection in effectivity and scalability.
In a weblog on the VM, Microsoft stated it has 1.5 TB of excessive bandwidth reminiscence (HBM) that leverages the MI300X. “What differentiates a ND MI300X v5 VM from our present household of ND-series VMs is the inclusion of 8 x AMD Intuition MI300X GPUs interconnected by way of Infinity Cloth 3.0 in every VM,” Microsoft stated. “MI300X affords trade main HBM capability and bandwidth, 192 GB of HBM3 reminiscence per GPU able to speeds as much as 5.2 TB/s.”
The VMs embrace:
- 400 Gb/s NVIDIA Quantum-2 CX7 InfiniBand per GPU with 3.2Tb/s per VM
- 4th Gen Intel Xeon Scalable processors
- PCIE Gen5 host-to-GPU interconnect with 64GB/s bandwidth per GPU
- 16 Channels of DDR5 DIMMs
HBM is crucial for AI functions resulting from its excessive bandwidth, low energy consumption, and compact dimension, Microsoft stated. “It’s very best for AI functions that have to shortly course of huge quantities of information. ND MI300X v5 has probably the most HBM capability accessible, 1.5 TB per VM or 192 GB of HBM per GPU, the very best HBM capability accessible within the cloud. This lets prospects run the most important, most superior AI fashions quicker, and with fewer GPUs.”
4th Gen AMD EPYC processors are additionally now getting used to run a brand new era of normal goal, memory-intensive and compute-optimized VMs. These new VMs from Microsoft proceed to showcase the expansion and demand of AMD EPYC processors within the cloud and can present higher value choices, as much as 20% higher efficiency for normal goal and memory-intensive VMs, and as much as 2x the CPU efficiency for compute-optimized VMs versus the earlier era of AMD EPYC-powered VMs. The brand new VM sequence might be made accessible in public preview within the first quarter of 2024.
At Ignite, AMD can also be highlighting the Azure NGads V620 sequence of VMs, now basically availability. It’s a new addition to the Azure N-series household of GPU-accelerated VMs for visualization. These VMs are powered by a number of AMD applied sciences – AMD Radeon™ PRO V620 GPUs and threerd Gen AMD EPYC CPUs. The NGads V620 sequence is meant for workloads that require comparatively extra GPU sources per VM, to assist extra demanding workloads or extra customers per VM. The V620 delivers an awesome expertise for a number of courses of GPU workloads, together with gaming, VDI and rendering. NG sequence customers get an optimized driver via the brand new AMD Software program: Cloud Version.
“AMD is fostering AI in all places – from the cloud, to the enterprise and finish level units – all powered by our CPUs, GPUs, accelerators and AI engines,” stated Vamsi Boppana, Senior Vice President, AI, AMD. “Along with Microsoft and a quickly rising ecosystem of software program and {hardware} companions, AMD is accelerating innovation to convey the advantages of AI to a broad portfolio of compute engines, with increasing software program capabilities.”
AMD additionally showcased the continued momentum of Ryzen AI, the primary devoted AI accelerator accessible on an x86 processor. With over 50 methods accessible now with Ryzen 7000 Sequence processors with Ryzen AI inbuilt, there are thousands and thousands of AMD AI PCs accessible in market. AMD and Microsoft have already collectively enabled Home windows Studio Results on Ryzen AI PCs, and AMD continues to work with software program builders, to convey increasing AI workloads and generative AI experiences to customers.
