Shulou(Shulou.com)11/24 Report--

CTOnews.com Microsoft Azure AI infrastructure has been upgraded with the launch of ND H100 v5 virtual machine series, which is equipped with NVIDIA H100 Tensor Core graphics processing unit (GPU) and low latency network.

Microsoft has released the Azure ND H100 v5 Virtual Machine (VM) family, which is now available in the Eastern and Central South US Azure regions.

The series features the latest Nvidia H100 Tensor Core GPUs and Nvidia Quantum-2 InfiniBand networks to address the increasing complexity of AI workloads. As part of Microsoft's deep and ongoing investment in AI, Microsoft is leveraging an AI-optimized cluster of 4K GPUs that will scale to hundreds of thousands of brand new GPUs over the next year.

CTOnews.com summarizes the features of the ND H100 v5 VM family:

AI Supercomputing GPU: Equipped with eight NVIDIA H100 Tensor Core GPUs. Compared to the previous generation of virtual machines, this series can significantly speed up AI models.

Next-generation computer processing units (CPUs): CPU performance is critical to AI training and reasoning, and Microsoft chose the fourth generation Intel Xeon Scalable processor.

Low-latency networking: 400Gb / s per GPU and 3.2 Tb / s cross-node bandwidth per VM with NVIDIA Quantum-2 ConnectX-7 InfiniBand.

Optimize host-to-GPU performance: 64GB/s bandwidth per GPU via PCIe Gen5.

Massive memory and memory bandwidth: DDR5 memory is at the heart of this family of virtual machines, providing higher data transfer speeds and efficiency.

Microsoft says the new 8-bit FP8 floating-point data type is six times faster for matrix multiplication operations than its predecessor, the FP16. ND H100 v5 VM helps large language models (such as BLOOM 175B end-to-end model) achieve up to twice the inference speed.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.