Unverified50% confidenceFactExact time
NVIDIA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
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/21/2026
First Seen
Valid until: 10/19/2026
Sources
Related Claims
UnverifiedNVIDIA H100 GPU原生支持FP8格式，其吞吐量相较FP16提升约2倍77% similarUnverifiedNVIDIA H100芯片的FP8训练吞吐量相较上一代A100提升了近6倍77% similarUnverifiedNVIDIA的cuFFT库在处理大规模多维FFT时性能可达CPU实现的数十倍77% similarVerifiedNVIDIA H100采用专为AI工作负载设计的Transformer Engine，在FP8精度下实现近4000 TFLOPS算力，比A100提升约3倍76% similarUnverified混合精度训练由 NVIDIA 在 2018 年提出，同时使用 FP32 和 FP16 两种精度75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/584828API
curl https://kongchang.com/api/v1/knowledge/claims/584828MCP
get_claim(id=584828)