Expired50% confidenceFactExact time
During testing, only 1 of the 4 GPUs reached a utilization rate of approximately 10%–13%, while the rest were nearly idle, and CPU usage was as high as approximately 50%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/30/2026(expired)
Sources
The Complete Guide to Running GLM-5.2 Locally with 4 RTX 3090s for $4,000
bilibili程序员-智能译站6/24/2026
Related Claims
Unverified单条推理通常只能占用GPU SM利用率的5%～15%，而合理的批次大小可将利用率提升至60%～90%74% similarUnverifiedIn 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.74% similarUnverifiedDuring the autoregressive decoding phase, GPU compute utilization is often less than 10%, making it a memory-intensive operation.74% 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73% similarUnverifiedWithout a sufficiently optimized software stack, actual GPU utilization often reaches only 30–50% of the theoretical peak.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/40847API
curl https://kongchang.com/api/v1/knowledge/claims/40847MCP
get_claim(id=40847)