Microsoft has announced during the Build 2020 developer conference that it is building its own supercomputer focused on its own OpenAI API to focus on General Artificial Intelligence, or IAG.
This new supercomputer will be part of the Microsoft Azure cloud infrastructure, and unfortunately its specifications are vague, as they only mention the use of 285,000 CPU cores , the use of 10,000 graphics processors , and a network connectivity of 400 Gb / s for each GPU server.
Obviously, it is logical to expect that processors have trusted AMD EPYC knowing that the last update of their Azure servers trusted AMD, while in GPUs, the logical thing would be Nvidia, but who knows if AMD made any juicy offer for the company to acquire its GPUs in addition to CPUs; Although with the Nvidia A100 already available, it is a very remote possibility, and more so when they indicate that it will be among the top 5 supercomputers in the world, so the most logical thing is AMD EPYC + Nvidia A100. It will be necessary to wait for the company to reveal more information.
“The supercomputer developed for OpenAI is a unique system with over 285,000 CPU cores, 10,000 GPUs, and 400 gigabits per second of network connectivity for each GPU server. Compared to other machines featured in the world’s top 500 supercomputers, it is in the top five, “according to Microsoft.
“Hosted on Azure, the supercomputer also benefits from all the capabilities of robust and modern cloud infrastructure, including rapid deployment, sustainable data centers, and access to Azure services.”