Microsoft cooperated with Nvidia in March this year and announced the release of the preview version of the Azure ND H100 v5 VM virtual machine; after several months of testing, Microsoft officially launched the virtual machine today to help enterprises process generation more efficiently. AI tasks.
It was previously reported that ND H100 v5 VM is Azure’s most powerful and highly scalable AI virtual machine series so far.
The virtual machine supports an on-demand configuration of up to 8 to thousands of NVIDIA H100 GPUs interconnected through a Quantum-2 InfiniBand network, significantly improving the performance of AI models. Compared with the previous generation ND A100 v4 VM, the virtual machine launched this time includes the following innovative technologies:
8 NVIDIA H100 Tensor Core GPUs, interconnected by next-generation NV Switch and NV Link 4.0.
400 Gb/s NVIDIA Quantum-2 CX7 InfiniBand per GPU and 3.2Tb/s performance per VM in a non-blocking fat-tree network.
The 8 local GPUs in each VM are interconnected via 3.8Tb/s pair split bandwidth NV Switch and NV Link 4.0.
4th generation Intel Xeon Scalable processors.
PCIE Gen 5 host-to-GPU interconnect with 64Gb/s bandwidth per GPU.
16 channels of 4800 MHz DDR5 memory.
The new virtual machines use NVIDIA’s NVIDIA Tensor Core H100 GPUs and NVIDIA Quantum-2 InfiniBand networking, and access is currently limited to Azure Enterprise in the East and West US.
Microsoft added that the company is working to add hundreds of thousands of new GPUs to its virtual machines over the next year.
In addition to officially announcing the Azure virtual machine, Microsoft also announced the expansion of its Azure OpenAI service, which is available in Canada East, East United States 2 (East United States 2), Japan East and UK South regions.
Microsoft said its Azure OpenAI service has more than 11,000 customers, attracting an average of 100 new customers per day this quarter.
This remarkable growth is a testament to the value and scalability our services bring to businesses eager to leverage the potential of AI for their unique needs.
Microsoft also added that GPT-4, the most powerful version of the ChatGPT AI platform, will also be available in the new Azure OpenAI service region.