Unverified50% confidenceFactExact time
相较于用GPU执行AI推理，NPU的功耗通常只有前者的1/5到1/10
1
Sources
50%
Confidence
Long-term
Relevance
7/6/2026
First Seen
Sources
Related Claims
UnverifiedIn a complete AI application system, GPU inference typically accounts for only 30%–40% of total compute, with the remaining 60%–70% of the compute load borne by the CPU.77% similarUnverifiedAI推理的边际成本不趋于零，因为每次推理需在GPU显存中加载完整模型权重，且每次生成的输出唯一无法缓存复用76% similarUnverifiedDuring large model inference, actual GPU utilization is often less than 10%.73% similarUnverifiednvidia-smi报告的GPU利用率本质上只统计采样时间窗口内GPU上至少有一个内核在执行的时间占比，而非计算资源的实际占用率71% similarUnverifiedDuring the autoregressive decoding phase, GPU compute utilization is often less than 10%, making it a memory-intensive operation.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/113486API
curl https://kongchang.com/api/v1/knowledge/claims/113486MCP
get_claim(id=113486)