NVIDIA H100 80GB HBM2e PCIe Tensor Core GPU AI HPC 699-21010-0200-000

NVIDIA H100 80GB HBM2e PCIe Tensor Core GPU 699-21010-0200-000: Hopper-Powered Acceleration for AI, LLMs, and HPC

The NVIDIA® H100 80GB HBM2e PCIe Tensor Core GPU delivers unprecedented acceleration for the world's most demanding artificial intelligence, large language model (LLM), and high-performance computing (HPC) workloads. Built on the NVIDIA Hopper™ architecture (GPU SKU GH100-200), this dual-slot PCIe Gen5 accelerator combines 80GB of HBM2e memory, 14,592 CUDA cores, and 456 fourth-generation Tensor Cores to move your largest models from training to inference at scale.

Key Features and Benefits of the NVIDIA H100 80GB PCIe Tensor Core GPU

Transformational AI and LLM Performance

Fourth-generation Tensor Cores and a dedicated Transformer Engine accelerate every math precision, from FP64 for scientific computing to FP8 for generative AI. With sparsity, the H100 PCIe delivers up to 3,026 TFLOPS of FP8 and 1,513 TFLOPS of FP16 Tensor Core performance*, dramatically speeding transformer-based model training and inference.

Massive Memory and Bandwidth

Equipped with 80GB of HBM2e memory delivering more than 2,000 GB/s of memory bandwidth - the highest of any PCIe card at launch - the H100 handles the largest models and datasets with ease. The PCIe Gen5 x16 host interface provides 128 GB/s of throughput, while up to three NVIDIA NVLink® bridges connect two H100 PCIe cards at 900 GB/s for 5x the bandwidth of PCIe Gen5.

Multi-Instance GPU (MIG) Flexibility

Multi-Instance GPU technology partitions the H100 into as many as seven fully isolated GPU instances, letting data centers right-size acceleration for every job and maximize utilization across mixed workloads with guaranteed quality of service.

Energy Efficiency and Thermal Management

Engineered for mainstream servers, the H100 PCIe operates unconstrained up to its 350 W thermal design power. Its passive, dual-slot full-height, full-length (10.5") design relies on system airflow with bi-directional airflow support for optimal thermal management in dense deployments.

Enterprise-Ready Virtualization and Software

The H100 for mainstream servers includes a five-year subscription to the NVIDIA AI Enterprise software suite, streamlining development and deployment of AI chatbots, recommendation engines, vision AI, and more. It is certified for NVIDIA-Certified Systems from leading server vendors.

Reliability and Security

With ECC memory, Secure Boot, SR-IOV, and hardware-isolated MIG instances, the H100 ensures data integrity and secure multi-tenant operation for mission-critical, production AI.

Upgrade your data center with the NVIDIA H100 80GB HBM2e PCIe Tensor Core GPU 699-21010-0200-000 and experience the future of accelerated computing.

* Shown with sparsity. Specifications one-half lower without sparsity.