Expired50% confidenceBenchmarkExact time
The theoretical inference throughput of the NVIDIA H100 and H800 at FP8 precision is 3–4 times higher than that of the A100.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/7/2026
First Seen
Valid until: 10/5/2026(expired)
Sources
GLM 5.2来袭：AI利润率崩塌前夜的产业变局
hackernewshackernews7/6/2026
Related Claims
Unverified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73% similarUnverifiedNVIDIA's iteration cycle from A100 to H100 and then to B100 is approximately two years, with each generation delivering roughly a 3–4× improvement in FP8 mixed-precision compute.72% similarUnverifiedThe H100's fourth-generation Tensor Core natively supports FP8, with a peak compute of 3958 TFLOPS, which is twice that of FP16.72% similarUnverifiedNVIDIA H100芯片的FP8训练吞吐量相较上一代A100提升了近6倍70% similarUnverifiedRunning the model in full precision requires several H100s.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/203109API
curl https://kongchang.com/api/v1/knowledge/claims/203109MCP
get_claim(id=203109)