Verified80% confidenceFactTime unknown
NVIDIA H100单卡FP64双精度浮点性能达到约34 TFLOPS
4
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedNVIDIA A100的理论FP16峰值算力约为312 TFLOPS，但朴素LLM推理场景下GPU利用率常不足30%82% similarUnverifiedNVIDIA H100的屋脊点约为597 FLOP/Byte（约2000 TFLOPS ÷ 3.35 TB/s）80% similarVerifiedNVIDIA A100单卡拥有80GB HBM2e显存和312 TFLOPS的FP16算力79% similarUnverifiedTPU v4单芯片峰值算力约275 TFLOPS（BF16精度），TPU v4-32 pod由32个芯片通过ICI高速互联组成，总算力约8800 TFLOPS79% similarVerifiedNVIDIA H20 的 BF16 峰值算力约为 296 TFLOPS，相比 H100 的 989 TFLOPS 大幅削减，但保留了 141GB HBM3 显存与 4.0 TB/s 显存带宽76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/21249API
curl https://kongchang.com/api/v1/knowledge/claims/21249MCP
get_claim(id=21249)