Intel recently held an event to announce a new China-specific version of its Gaudi 2 processor for artificial intelligence (AI) workloads. This new version complies with the latest US sanctions on AI and GPU processors for China. The new version will limit the AI server performance on the base somewhat due to the reduced scale-up interconnection of the cards. This would somewhat fill a huge shortage of AI GPUs such as Nvidia’s A100 and H100, as well as his scaled-down version of the A800, which Nvidia made specifically for Chinese customers.
Intel has two Gaudi 2 versions in the 600W OCP model form factor.called normal version Gaudi 2 HL-225Hwith a 24 x 100Gbps RDMA (RoCE v2) interconnect, but a cut-down version of this product Gaudi 2 HL-225B, with 21 x 100Gbps RDMA (RoCE v2) scale-up interconnects. Reducing the number of scale-up interfaces reduces the performance of AI servers based on this version of Gaudi 2, but given the huge demand for AI capabilities in China, it may still gain popularity.
Intel describes the Chinese market version of the Gaudi 2 accelerator as “The HLB-225B processor is compliant with: US BIS regulationWhen it comes to supercomputers and semiconductors,
The availability of Intel’s Habana Gaudi 2 in China should ease the GPU server shortage, but note that applications already using Nvidia’s GPUs are unlikely to be ported to Gaudi. is needed. Therefore, existing Nvidia customers should continue to purchase from his Nvidia.
“On July 11, Intel held an event for customers, partners and local media in the Chinese market,” said an Intel spokesperson. “As part of the event program, Intel provided attendees with updates on our AI strategy and unique AI product portfolio, and announced the availability of Gaudi2 for Chinese customers. The offering continues Intel’s nearly 40-year history of servicing, delivering innovative yet compliant products to this key growth market.”
