Microsoft Azure Shows Off World’s First Custom Nvidia Blackwell Racks – Tests Its Own 32 GPU Server Rack and Liquid Cooling Solution

When you buy from our articles through links, Future and its syndication partners may earn a commission.

Nvidia’s Blackwell processors are among the most sought-after data center hardware today as companies seek to train large language models (LLMs) with an ever-increasing number of parameters. Rumor had it that Microsoft was the first company to get their hands on Blackwell servers, but that was unofficial information. Today the company said that not only had it acquired Nvidia Blackwell parts, but that they were also operational.

“Microsoft Azure is the first cloud (service provider) to run Nvidia’s Blackwell system with GB200-powered AI servers,” reads a post from Microsoft Azure at floating, using Infiniband networking and innovative closed-loop liquid cooling Read more on MS Ignite.”

So Microsoft Azure has at least one GB200-based server rack with an unknown number of B200 processors – probably 32. It uses a very advanced liquid cooling system. The machine isn’t Nvidia’s NVL72 GB200, which Microsoft reportedly prefers to lower-density variants. This particular rack will likely be used for testing purposes (both Nvidia Blackwell GPUs and the liquid cooling system), and in the coming months Microsoft will deploy Blackwell-based servers for commercial workloads.

It is expected that an NVL72 GB200 machine with 72 B200 graphics processors will consume and dissipate approximately 120 kW of power, making liquid cooling mandatory for such machines. Therefore, it is a good idea for Microsoft to test its own liquid cooling solution before deploying Blackwell-based cases.

Nvidia’s B200 GPU offers 2.5 times the performance compared to the H100 processor when it comes to FP8/INT8 performance (4,500 TFLOPS/TOPS vs. 1,980 TOPS). With the FP4 data format, Nvidia’s B200 offers a whopping 9 PFLOPS of performance, opening the doors for the training of highly advanced LLMs that could enable new use models for AI in general.

“Our long-standing partnership with NVIDIA and deep innovation continue to lead the industry, powering the most advanced AI workloads,” said Satya Nadella, CEO of Microsoft, in another X post.

Microsoft is expected to share more details about its Blackwell-based machines and AI projects in general at its annual Ignite conference. The MS Ignite conference will take place in Chicago from November 18 to 22, 2024. Large-scale deployment of Blackwell server installations is expected to begin late this year or early 2025.