Description

The NVIDIA H100 Tensor Core GPU, built on the Hopper architecture, delivers next-generation AI and HPC acceleration for data centers. With 80 billion transistors and up to 80GB of high-bandwidth HBM2e or HBM3 memory, the H100 enables breakthrough performance across FP64 to FP8 precision workloads. It features the advanced Transformer Engine for AI model training and inference, second-generation Multi-Instance GPU (MIG) for secure multi-tenant environments, and NVLink support for ultra-fast GPU-to-GPU communication. The H100 powers large-scale deep learning, data analytics, scientific computing, and confidential computing workloads in the most demanding enterprise environments.