HGX & OAM Server

Servers with SXM modules that interconnect 8 or 16 GPUs achieve the best possible performance/watt for GPU workloads and gpu-to-gpu bandwidth to drive LLMs and supercomputing.