The NVIDIA HGX H100 combines 8 NVIDIA H100 Tensor Core GPUs linked via 900 GB/s NVLink connectivity and NVSwitch technology. Engineered for generative AI, large language models (LLMs), and high-performance computing (HPC) simulations, this server platform represents the pinnacle of modern enterprise artificial intelligence infrastructure.
Technical Specifications
Technical Specifications
8 Parameters Listed
GPU Architecture
8x NVIDIA H100 SXM5 GPUs (80GB HBM3 memory per GPU)
Interconnect
NVIDIA NVSwitch at 900 GB/s bidirectional bandwidth per GPU
FP8 AI Performance
Up to 32 PetaFLOPS FP8 AI compute
System Processors
Dual AMD EPYC 9004 Series or Dual 5th Gen Intel Xeon Scalable