Microsoft announced the NDm A100 v4 Public AI supercomputing

Microsoft announced its latest supercomputer NDm A100 v4 Public AI on its official blog yesterday, and stated that the supercomputer was ranked in the top 10 in the recent supercomputing TOP 500 list.

Microsoft’s official blog that Kevin Raines, a marketer in Microsoft’s Azure cloud computing department, said on the official blog that at the Microsoft Ignite conference five years ago, Microsoft announced the global cloud-based AI supercomputer, and it will be released in 2020. At the Build 2020 developer conference held in May of 2005, it was once again announced that it would cooperate with OpenAI to create a supercomputer comparable to the world's five largest supercomputers at the time, and Microsoft has been trying to continuously challenge the limit so far.

In 2021, Microsoft's Azure cloud service platform took public cloud services to a new level and showed public AI supercomputers to the outside world. In June 2021, the supercomputing TOP 500 list won from 26th to No. 29 in four consecutive positions, these supercomputer clusters are part of the ND A100 v4 series, and each of the Microsoft supercomputers ranked in the supercomputing TOP 500 has reached 16.59 PetaFLOPS on the HPL benchmark. In the latest supercomputing TOP 500 list recently announced by "Supercomputing 21", Microsoft Azure Cloud Platform's NDm A100 v4 Public AI Supercomputing successfully joined the TOP 10 list with a score of 30.05 PetaFLOPS.

Ian Buck, Vice President and General Manager of the Accelerated Computing Department of Nvidia, said: “The integration of high-performance computing and artificial intelligence will be a revolution. It will provide obvious acceleration effects for every simulation and promote the entire science. And the development of the industrial field." The Microsoft Azure cloud platform supercomputing NDm A100 v4 example demonstrates the NVIDIA GPU acceleration function and the strength of the NVIDIA InfiniBand network, allowing researchers to discover new content faster and promote scientific development.

Microsoft NDm A100 v4 Public AI Supercomputing is equipped with eight Nvidia A100 Tensor Core GPUs , each GPU has 432 Nvidia Tensor Cores and 6912 Nvidia CUDA Cores, each GPU supports HDR 200G InfiniBand connection, which can be connected to thousands in Azure cloud Establish a fast connection between two GPUs.

Post a Comment