Unverified50% confidenceBenchmarkExact time
NVIDIA H100 GPU原生支持FP8格式，其吞吐量相较FP16提升约2倍
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/7/2026
First Seen
Valid until: 10/5/2026
Sources
NVIDIA AI-Q蓝图×OCI：企业级AI智能体生产部署实践指南
rss6/26/2026
Related Claims
Verified在配备 Tensor Core 的 NVIDIA GPU（如 A100、H100）上，FP16/BF16 矩阵运算的吞吐量可以达到 FP32 的 2-8 倍84% similarUnverified现代GPU如NVIDIA H100、H800原生支持FP8计算，理论上可实现约2倍吞吐量提升79% similarUnverifiedNVIDIA的Blackwell架构（B100/B200）在硬件层面原生支持MXFP4格式的矩阵运算，吞吐量相比FP8可提升近一倍78% similarPartially VerifiedNVIDIA的MIG技术可在A100/H100等高端GPU上实现硬件级别的资源隔离78% similarUnverifiedNVIDIA从A100到H100再到B100的迭代周期约两年，每代FP8混合精度算力提升约3-4倍77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/185584API
curl https://kongchang.com/api/v1/knowledge/claims/185584MCP
get_claim(id=185584)