NVIDIA H100 Tensor Core GPU Architecture Whitepaper

This document details the architecture of the NVIDIA H100 Tensor Core GPU, a next-generation processor designed for the most demanding data center workloads. Built upon the NVIDIA Hopper GPU architecture, the H100 delivers exceptional performance, scalability, and security for artificial intelligence (AI), high-performance computing (HPC), and data analytics applications.

The H100 GPU represents a significant leap forward, offering advanced features and capabilities that accelerate complex computations. It is engineered to power the future of cloud computing, enabling breakthroughs in scientific research, enterprise AI, and large-scale data processing.

Explore the innovations within the H100, including its enhanced Tensor Cores, new Transformer Engine, advanced memory subsystems, and improved interconnect technologies like NVLink and NVSwitch. This architecture is crucial for handling the exponential growth in AI model complexity and data volumes, positioning NVIDIA at the forefront of accelerated computing.

	NVIDIA H100 Tensor Core GPU Datasheet for AI and HPC Datasheet detailing the NVIDIA H100 Tensor Core GPU, featuring unprecedented performance, scalability, and security for data centers. Highlights include Hopper architecture, Transformer Engine, NVLink, and accelerated AI/HPC workloads.
	NVIDIA H100 Tensor Core GPU Datasheet - High-Performance AI and HPC Acceleration Detailed datasheet for the NVIDIA H100 Tensor Core GPU, highlighting its unprecedented performance, scalability, and security for AI and HPC workloads. Features include the Hopper architecture, Transformer Engine, NVLink Switch System, and Confidential Computing.
	NVIDIA H100 PCIe GPU Product Brief: Specifications and Features Detailed product brief for the NVIDIA H100 PCIe GPU, covering specifications, features, NVLink support, power connectors, AI enterprise software, and support information.
	NVIDIA H100 PCIe GPU Product Brief Detailed product brief for the NVIDIA H100 PCIe GPU, covering its specifications, features, NVLink support, power requirements, NVIDIA AI Enterprise software integration, and support information.
	NVIDIA H100 NVL GPU Product Brief A product brief detailing the NVIDIA H100 NVL GPU, its specifications, features, and support information for data center applications in AI, data analytics, and high-performance computing (HPC).
	NVIDIA Data Center GPU Driver Release Notes v535.161.08/538.46 Official release notes for the NVIDIA Data Center GPU Driver, detailing version highlights, fixed issues, known issues, and hardware/software support for Linux and Windows operating systems.
	NVIDIA Data Center Platform: GPU Portfolio and Workload Acceleration Explore the NVIDIA Data Center Platform, featuring Hopper and Ada Lovelace architectures. This guide details GPU capabilities for AI, HPC, NLP, inference, training, and more, helping select the ideal GPU for specific workloads.
	NVIDIA DGX SuperPOD: Next-Generation AI Infrastructure Reference Architecture This document outlines the reference architecture for the NVIDIA DGX SuperPOD, a scalable infrastructure designed for AI leadership. It details the key components, network fabrics, storage architecture, and software stack, including NVIDIA DGX GB200 systems, InfiniBand, NVLink, and Mission Control software, to power next-generation AI factories.

Related Documents