Introducing NVIDIA HGX H100: An Accelerated Server Platform for AI and High-Performance Computing

Originally published at: https://developer.nvidia.com/blog/introducing-nvidia-hgx-h100-an-accelerated-server-platform-for-ai-and-high-performance-computing/

Introducing the NVIDIA HGX H100, a key GPU server building block powered by the Hopper architecture.

Why does not put NVSwitches into the 4H100 Server in order to increase the P2P bandwidth from 300GB/s to 900GB/s?

why not using on of those HGX systems to enhance the block diagram images in this post?