Google Leverages NVIDIA Blackwell GB200 NVL Racks For Its AI Cloud Platform, Liquid Cooled Computing

Oct 16, 2024 at 11:45am EDT

Google is yet another player to use the custom NVIDIA Blackwell GB200 NVL racks for improved computing performance for its Cloud platform.

Google Cloud shows off its first NVIDIA Blackwell GB200 NVL Rack, featuring liquid-cooled B200 GPUs

The search engine giant, Google today showed off its first GB200 NVL-based server on X, deployed to power its Cloud platform. The company has a big 12% share in the cloud system market worldwide, and being the third largest cloud service provider, it now wants to expand its operations by offering faster services.

Related Story Tensordyne’s 3nm Napier AI Chip Promises 13x Higher Token Throughput Than Blackwell & Blazes Past Rubin With 1000 Tokens/s In Multi-Trillion Parameter Models

The official Google Cloud account on X announced that it has begun deploying NVIDIA's GB200 NVL racks for its cloud platform. It posted a pic, showing a sever with several GB200 NVL racks, which features liquid-cooling for the GB200 high-performance chips. Each GB200 chip contains one Grace CPU and one B200 Blackwell-based data center graphics chip for an incredible 90 TFLOPS of FP64 performance.

It should be kept in mind that Google announced that it is using 'custom' GB200 NVL racks and we aren't sure if this brings the same configuration as the GB200 NVL72 with 32 Grace CPUs and 72 B200 GPUs through a 72-GPU NVLink domain.

The official NVIDIA GB200 NVL72 brings each rack featuring 36 Grace CPUs paired with 72 Blackwell GPUs, which can offer a bandwidth of up to 130 TB/s. Each rack with 36 Grace CPUs and 72 B200 GPUs can result in a staggering 3240 TFLOPS of FP64 performance and provide up to 13.5 TB of HBM3e memory.

Google isn't the first one to deploy the GB200 NVL racks or GB200 chips as several other companies have already started using GB200 chips, including giants such as Microsoft, which posted a similar GB200-based server with liquid cooling.

Then there is Foxconn, a Taiwanese manufacturer, which has begun deploying the GB200 NVL72 racks to build the fastest supercomputer in Taiwan. Through the powerful computing performance of GB200 chips, several companies are on their way to upgrading their existing systems for various uses, including LLMs, medical research, cloud storage, AI workloads, etc. NVIDIA's Blackwell is now in full production so expect more systems to utilize the industry-leading AI chips.

News Source: Google Cloud

About the author: Sarfraz Khan is a hardware reporter with a focus on PC components and the builder community. With years of experience writing about PC hardware and laptops, his work has been featured on several reputable technology publications. Sarfraz's hands-on experience is demonstrated through his first-person accounts of using and comparing different hardware configurations, providing practical and relatable insights for everyday users. His technical analysis is respected by peers in the enthusiast community and has been cited by specialized hardware sites such as Germany's Igor's Lab.

Follow Wccftech on Google to get more of our news coverage in your feeds.