Performance reports or announcements from NVIDIA or xAI

NVIDIA and xAI Build World's Largest AI Supercomputer with 100,000 GPUs in Memphis in 122 Days

Oct 28, 2024, 07:12 PM

NVIDIA and Elon Musk's xAI have built the world's largest AI supercomputer, named Colossus, in Memphis, Tennessee. This supercomputer, which houses 100,000 NVIDIA Hopper GPUs, was constructed in just 122 days and achieved 95% data throughput with zero latency issues using NVIDIA's Spectrum-X Ethernet networking platform. The project set a record by installing 100,000 Nvidia H100 chips in just 19 days. Jensen Huang, CEO of NVIDIA, praised the rapid installation and performance of the system. xAI plans to double the size of Colossus to 200,000 GPUs in the coming months. The supercomputer, which uses liquid cooling, is expected to significantly advance AI training capabilities. Elon Musk described Colossus as the most powerful training system in the world. The system will power xAI's AI model, Grok.