The NVIDIA HGX H200 is a premier accelerated scale-up platform designed for the most demanding generative AI, data analytics, and high-performance computing (HPC) workloads. It integrates eight NVIDIA H200 Tensor Core GPUs with high-speed NVLink interconnects to deliver unprecedented performance and memory bandwidth for large-scale model training and inference.
- Features 8 NVIDIA H200 Tensor Core GPUs with SXM form factor.
- Utilizes NVIDIA NVLink and NVSwitch technology for high-speed GPU-to-GPU communication.
- Supports up to 900GB/s of aggregate bidirectional GPU-to-GPU bandwidth.
- Optimized for large language models (LLMs) and complex AI reasoning tasks.
- Compatible with NVIDIA Quantum-2 InfiniBand and Spectrum-4 Ethernet networking.
- Includes support for NVIDIA BlueField-3 DPUs for cloud networking and security.
- Delivers massive HBM3e memory capacity and bandwidth for data-intensive applications.
- Fully integrated with the NVIDIA AI Enterprise software suite.
- Designed for seamless integration into standard data center rack environments.
Related Products
A relatable product