Loading...
Foresight News reported that cloud computing platform CoreWeave (NASDAQ: CRWV) announced the successful deployment of a super large scale computing cluster with 2496 NVIDIA Blackwell GPUs in collaboration with NVIDIA and IBM, setting a record for the largest submission in MLPerf benchmark testing history. The cluster used 2496 NVIDIA Blackwell GPUs and ran on CoreWeave's AI optimized cloud platform. The cluster size is 34 times larger than other cloud service providers. In performance testing, the cluster completed the training of the large language model Llama 3.1 405B in 27.3 minutes, which is more than twice as fast as similar sized clusters.