The NVIDIA HGX H100 is a premier accelerated scale-up platform designed for the most demanding generative AI, data analytics, and HPC workloads. It integrates eight NVIDIA H100 Tensor Core GPUs with high-speed interconnects to deliver unprecedented performance for modern data centers.
- Features 8x NVIDIA H100 Tensor Core GPUs with SXM5 form factor.
- Utilizes fourth-generation NVIDIA NVLink for high-speed GPU-to-GPU communication.
- Delivers up to 32 petaFLOPS of FP8 training performance.
- Provides 640GB of total HBM3 memory with up to 24TB/s of aggregate bandwidth.
- Supports NVIDIA NVSwitch technology for seamless multi-GPU scaling.
- Integrated PCIe Gen5 support for high-bandwidth CPU-to-GPU connectivity.
- Optimized for large-scale AI model training and complex scientific simulations.
- Compatible with NVIDIA InfiniBand and Ethernet networking up to 400Gb/s.
- Advanced thermal management designed for high-density server environments.
- Fully supported by the NVIDIA AI Enterprise software suite.
Related Products
A relatable product