From Nvidia: 2024-11-18 13:30:13

The NVIDIA Hopper architecture has revolutionized AI and HPC, offering higher performance and energy efficiency. The new NVIDIA H200 NVL PCIe GPU is tailored for data centers seeking lower-power, air-cooled designs for all AI and HPC workloads. With 1.5x memory increase and 1.2x bandwidth, H200 NVL delivers faster inference and 1.3x performance boost for HPC.

NVIDIA H200 NVL comes with NVLink technology for 7x faster GPU-to-GPU communication, enhancing performance for HPC and AI workloads. Enterprises can utilize NVIDIA AI Enterprise software for accelerated applications, including high-performance AI model inference. Companies like Dropbox and the University of New Mexico are leveraging H200 NVL for their AI and HPC initiatives.

Dell Technologies, HPE, Lenovo, and Supermicro will offer configurations supporting H200 NVL, with other platforms from Aivres, ASRock Rack, ASUS, GIGABYTE, and more. Some systems will be based on NVIDIA MGX modular architecture for flexible data center infrastructure designs. Availability from global partners begins in December, with NVIDIA developing an Enterprise Reference Architecture for H200 NVL systems.



Read more at Nvidia: NVIDIA H200 NVL PCIe GPU Accelerates AI and HPC Applications