NVIDIA H800
芯片
GPU
历史
The export-compliant H100 for China — and the chip DeepSeek trained V3 on.
您正在查看一条芯片记录：该处理器是什么、其关键规格、谁设计它、谁制造它、谁供应其零部件，以及它被用在哪里。
The H800 kept the H100's FP8 tensor throughput but cut NVLink bandwidth to about 400 GB/s and sharply reduced FP64 to satisfy the 2022 export rules. DeepSeek's V3 technical report describes training on a 2,048-GPU H800 cluster, engineering around the interconnect limits — the chip is now shorthand for how constrained hardware shaped Chinese frontier labs. Banned by the October 2023 rules.
规格与关键参数
每个数据均标注来源依据：厂商规格、厂商声明、公开报道、估算或官方数据。"未公开披露"即字面含义——我们从不捏造数值。
Overview
Designer
NVIDIA
Architecture
Hopper
Introduced
March 2023, China market
Status
Historical — barred by the October 2023 US export rules
Physical
Interconnect
NVLink reduced to ~400 GB/sReported
算力
FP64
~1 TFLOPS (sharply reduced vs H100)Reported
FP8/BF16 tensor
Matches H100 published figuresReported
它如何与整个技术栈相连
本记录的关联链接一览——点击分组可直接跳转。
设计方
由……制造
用于训练
基于
来源
DeepSeek-V3 technical report (H800 cluster) ↗
记录审阅截至 2026-01-15 — 查看 方法论.