NVIDIA - HGX GH200 - Grace Hopper AI Supercomputing Platform

The NVIDIA HGX GH200 is a premier accelerated scale-up platform designed for the most demanding generative AI, data analytics, and HPC workloads. It integrates NVIDIA Grace Hopper Superchips with high-speed interconnects to deliver unprecedented performance and memory capacity for large-scale AI models. - Features NVIDIA Grace Hopper Superchips combining Grace CPU and Hopper GPU architectures. - High-bandwidth NVLink-C2C interconnect providing 900 GB/s bidirectional bandwidth between CPU and GPU. - Unified memory architecture with high-capacity LPDDR5X and HBM3/HBM3e memory subsystems. - Supports NVIDIA NVLink Switch System for seamless multi-node scaling and GPU-to-GPU communication. - Integrated NVIDIA BlueField-3 DPU for cloud networking, storage, and zero-trust security. - Optimized Transformer Engine to accelerate AI inference and training performance. - Delivers up to 4 PFLOPS of FP8 performance per superchip for advanced AI reasoning. - Compatible with NVIDIA Quantum-2 InfiniBand and Spectrum-4 Ethernet networking fabrics. - Designed for massive-scale model training and high-throughput agentic AI workloads. - Full-stack support with NVIDIA AI Enterprise and optimized HPC software environments.