The NVIDIA H100 Tensor Core GPU is the ultimate data centre GPU for large-scale AI and HPC. It has unprecedented performance, scalability, and security, and comes with the NVIDIA AI Enterprise software suite to make AI development and deployment easier. H100 can handle exascale workloads with a dedicated Transformer Engine for massive language models. It can also be split into right-sized Multi-Instance GPU (MIG) partitions for small jobs. With Hopper Confidential Computing, H100 can protect sensitive applications on shared data centre infrastructure. H100 is the most powerful end-to-end AI and HPC data centre platform with NVIDIA AI Enterprise.
The NVIDIA Hopper architecture is the foundation of H100. It has new compute core capabilities, such as the Transformer Engine, and faster networking to deliver an order of magnitude speedup over the previous generation. NVIDIA NVLink enables ultra-high bandwidth and low latency between two H100 boards, and supports memory pooling and performance scaling (application support required). Second-generation MIG allows secure partitioning of the GPU into isolated instances for 7x more secured tenants. NVIDIA AI Enterprise (exclusive to the H100 PCIe) is a software suite that optimizes the development and deployment of accelerated AI workflows, and leverages the new H100 architectural innovations. These technology breakthroughs make the H100 Tensor Core GPU – the world’s most advanced GPU ever built.