Unverified50% confidenceFactExact time
H100 GPU采用英伟达Hopper架构，单卡峰值算力达3,958 TFLOPS（FP16精度）
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
马斯克承诺不切断Anthropic：AI云基础设施的信任博弈
rss7/9/2026
Related Claims
UnverifiedH100 GPU基于Hopper架构，其Transformer引擎能在FP8和FP16精度间动态切换，将吞吐量提升至A100的约2-3倍83% similarUnverifiedA100是英伟达数据中心级GPU，单卡算力约312 TFLOPS（FP16）81% similarUnverifiedV100的Tensor Core在FP16精度下峰值算力可达125 TFLOPS，是前代Pascal架构的5倍以上79% similarUnverifiedNVIDIA H100是基于Hopper架构的数据中心GPU，其SXM5版本TDP高达700W77% similarUnverified英伟达H100/H200 GPU的算力是台积电先进制程工艺与英伟达CUDA架构的协同产物77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/470992API
curl https://kongchang.com/api/v1/knowledge/claims/470992MCP
get_claim(id=470992)