Unverified50% confidenceFactExact time
英伟达H100基于Hopper架构，采用HBM3内存带宽高达3.35TB/s，单卡FP16算力可达近2000 TFLOPS
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
DeepSeek自研AI芯片：算力自主背后的战略逻辑与挑战
hackernewshackernews7/9/2026
Related Claims
Verified英伟达H100采用的HBM3内存带宽高达3.35TB/s84% similarPartially VerifiedA100 80GB采用HBM2e显存，带宽约2039GB/s；H100 SXM采用HBM3，带宽约3.35TB/s，FP64 Tensor Core算力约67 TFLOPS，是A100（约19.5 TFLOPS）的3倍以上81% similarVerifiedH100 SXM5的HBM3带宽约为3.35TB/s，FP16峰值算力约989TFLOPS81% similarUnverifiedTPU v3-8 采用脉动阵列架构，包含 8 个 TPU 核心，总计提供 320GB 高带宽内存（HBM）78% similarUnverifiedGB300 集成 HBM3e 高带宽内存，单卡容量高达288GB，内存带宽超过10TB/s77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489799API
curl https://kongchang.com/api/v1/knowledge/claims/489799MCP
get_claim(id=489799)