NVIDIA HGX H200 8-GPU AI Platform

NVIDIA HGX H200 8-GPU AI Platform

Enterprise AI and High-Performance Computing

The NVIDIA HGX H200 platform combines eight NVIDIA H200 Tensor Core GPUs with high-speed GPU-to-GPU interconnect technology to accelerate demanding AI and scientific computing workloads.

Designed for enterprise AI infrastructure, the platform supports large language model training, fine-tuning, inference, data analytics and high-performance computing. Its high-capacity HBM3e memory and high-bandwidth GPU interconnects help process large datasets and complex models efficiently.

Key Features

• Eight NVIDIA H200 Tensor Core GPUs
• 141 GB HBM3e memory per GPU
• 1,128 GB total GPU memory across eight GPUs
• 4.8 TB/s memory bandwidth per GPU
• Up to 38.4 TB/s aggregate GPU memory bandwidth across eight GPUs
• NVIDIA NVLink and NVSwitch interconnect technology
• Designed for enterprise AI training and inference
• Suitable for scientific computing and HPC workloads

Compatibility and Configuration

HGX H200 is a GPU computing platform intended for compatible server systems. CPU configuration, system memory, networking, storage, power delivery, cooling and chassis design depend on the complete server configuration.

Please contact us to discuss platform availability, compatible server configurations and delivery options.