Microsoft Azure-Based Machine Ranked No. 10 in Supercomputing List


Microsoft this week credited the use of Nvidia A100 Tensor Core 80-GB graphics processing units (GPUs) in Azure virtual machines for getting four of its machines placed in the TOP500 supercomputing list.

The TOP500 organization publishes an annual supercomputing list, with its latest findings announced on Nov. 15. The Fugaku supercomputer at the Riken Center for Computational Science in Kobe, Japan was the top performer, delivering 442 petaflops per second. The machine was codeveloped by Riken and Fujitsu using a custom ARM A64FX processor built by Fujitsu. Its first-place standing was a repeat performance.

The only shakeup in TOP500 rankings this year happened with Microsoft's new Voyager-EUS2 machine running on Microsoft Azure infrastructure, according to the TOP500 announcement. It claimed the No. 10 spot, delivering 30.05 petaflops per second.

Here's the TOP500 description of Microsoft's machine:

Voyager-EUS2, a Microsoft Azure system installed at Microsoft in the U.S., is the only new system in the TOP10. It achieved 30.05 Pflop/s and is listed at No. 10. This architecture is based on an AMD EPYC processor with 48 cores and 2.45GHz working together with an NVIDIA A100 GPU with 80 G.B. memory and utilizing a Mellanox HDR Infiniband for data transfer.

Microsoft's credited its success to the use of Nvidia's A100 Tensor Core with 80-GB GPUs, a doubling of its past use of 40-GB GPUs.

AMD processor use was also deemed a factor, per the TOP500 announcement:

Like the last list, AMD processors are seeing a lot of success. Frontera, which has a Xeon Platinum 8280 processor, got bumped by Voyager-EUS2, which has an AMD EPYC processor. What’s more, all of the new Top15 machines described above have AMD processors.

Microsoft described its efforts as building "public AI supercomputers" that can be leveraged by organizations to train AI models. Meta (formerly "Facebook") is using Microsoft's supercomputers for research. Nuance is using them for voice-enabled Dragon solutions and Microsoft's internal teams are using them for "large scale cognitive science model training."

Microsoft has also claimed top status based on its collaboration with OpenAI, announced about a year-and-a-half ago, on an Azure-based supercomputer. The supercomputer is being used model things like language translation, text search, speech recognition and computer vision. Microsoft invested $1 billion in OpenAI a couple of years ago, and OpenAI uses Azure as its service provider.

At this month's Microsoft Ignite conference, Microsoft described a new Azure OpenAI service providing access to OpenAI's GPT-3 natural language models. GPT-3 lets organizations train models using 175 billion training parameters.

China is the country with the most supercomputers in the TOP500 list at 173 machines. The United States is No. 2 with 150 machines. These two countries account for nearly two-thirds of the machines in the TOP500 list.

The TOP500 list just measures a systems ability to "solve a linear set of equations." It's not a measure of performance. The list should not be used to select a supercomputer to run applications, the TOP500 organization explained, in its site description.

About the Author

Kurt Mackie is senior news producer for 1105 Media's Converge360 group.


comments powered by Disqus

Subscribe on YouTube