Home › Chip › NVIDIA H100

NVIDIA H100

Chip GPU Spedizione

The GPU of the ChatGPT era — the unit of account for the 2023–25 AI build-out.

Stai visualizzando un record di chip: cos'è questo processore, le sue specifiche principali, chi lo progetta, lo produce e ne fornisce i componenti, e dove viene utilizzato.

Hopper's H100 became the defining commodity of the AI boom: clusters were sized in H100s, startups raised funding denominated in them, and export policy was written around them. Its Transformer Engine pairs fourth-generation tensor cores with FP8 precision, roughly tripling A100 training throughput per chip.

xAI's Colossus assembled roughly 100,000 of them in 122 days; Meta, Microsoft, OpenAI, Google and every major cloud deployed them in the hundreds of thousands. Blackwell succeeded it in 2025, but H100/H200 fleets still carry a large share of world AI compute.

Specifiche e dati principali

Ogni dato è etichettato in base alla fonte: specifiche del produttore, dichiarazione del produttore, riportato, stima o ufficiale. "Non divulgato pubblicamente" significa esattamente questo — non inventiamo mai valori.

Overview

Designer NVIDIA

Architecture Hopper (GH100)

Announced March 2022 (GTC); shipping volume from early 2023

Status Shipping — succeeded by Blackwell at the frontier

Manufacturing

Foundry / process TSMC 4N (custom 5 nm-class)Vendor spec

Transistors 80 billionVendor spec

Die size 814 mm²Vendor spec

Packaging CoWoS-S with 5 or 6 HBM stacksVendor spec

Memoria

Memory 80 GB HBM3 (SXM); 94 GB HBM3 (NVL); 80 GB HBM2e (PCIe)Vendor spec

Memory bandwidth 3.35 (SXM) / 3.9 (NVL) TB/sVendor spec

Calcolo

TF32 tensor (dense) 495 TFLOPSVendor spec NVIDIA datasheets often headline throughput with structured sparsity; figures here are dense unless noted.

BF16/FP16 tensor (dense) 990 TFLOPSVendor spec

FP8 tensor (dense) 1,979 TFLOPSVendor spec FP8 via the Transformer Engine was Hopper's signature feature

INT8 tensor (dense) 1,979 TOPSVendor spec

FP64 tensor 67 TFLOPSVendor spec

AI Performance

MLPerf Set the reference results for training and inference benchmarks across 2023–24 submissionsMLPerf

Energia

TDP up to 700 (SXM) / 350 (PCIe) WVendor spec

Physical

Form factors SXM5, PCIe, NVL (paired PCIe)

Interconnect NVLink 4 — 900 GB/s; PCIe 5.0Vendor spec

Prezzo

Launch price Not publicly disclosed

Market price ~$25,000–40,000 per GPU at the 2023–24 peakReported

Cloud rental Roughly $2–7 per GPU-hour across major and specialist clouds during 2025 — falling steadilyReported varies with commitment and provider

Come si collega al resto dello stack

I collegamenti di questa scheda a colpo d'occhio — clicca un gruppo per raggiungerlo. +11 altri gruppi più in basso.

Progettato da

NVIDIA

Prodotto da

TSMC

Nodo di processo

TSMC N5 / 4N (5 nm class)

Packaging

CoWoS

Fornitore di packaging

TSMC

Tecnologia di memoria

HBM3

Memoria fornita da

SK hynix

Interconnessione e networking

NVLink & NVSwitch InfiniBand

Presente in

Dell PowerEdge XE9680/XE9712 Supermicro NVL72 & 4U Liquid-Cooled Systems NVIDIA HGX H100/H200

Esecuzioni

Triton (OpenAI) PyTorch Megatron-Core vLLM

Supportato da

CUDA

Usato per addestrare

Llama 3.1 405B Llama 4 Grok 3

Disponibile da

Amazon Web Services Microsoft Google (Alphabet) Oracle CoreWeave Lambda

Distribuito presso

xAI Colossus

Base per

NVIDIA H800

Sostituito da

NVIDIA H200

Successore di

NVIDIA A100

È in concorrenza con

AMD Instinct MI300X Google TPU v5p Groq LPU AWS Trainium2 Intel Gaudi 3 Huawei Ascend 910C

Correlati

Export Controls & the Chip War Training & Pre-training MLPerf

Fonti

NVIDIA H100 page ↗ · Hopper architecture whitepaper ↗

Record revisionato al 2026-01-15 — vedi Metodologia.