Nvidia and Oracle scale Vera Rubin in the OCI cloud to multiple gigawatts
Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .
Nvidia's newest chips go straight into the cloud in large clusters, and there is more compute available for training models.
Nvidia said Vera Rubin NVL72 systems and Vera processors are already generally available in Oracle Cloud Infrastructure, and OCI superclusters are expanding to multiple gigawatts; each rack links 72 accelerators via NVLink. HGX Rubin NVL8 is planned to be added: eight Rubin accelerators, two 64-core Intel Xeon 6 processors, 4 TB of DDR5 memory and 6.4 Tbps RDMA networking. An OpenAI representative said Vera Rubin on OCI provides the power to train and serve models; at Abilene, OpenAI runs more than 100,000 Blackwell accelerators.
Sources 1