Unverified50% confidenceFactExact time
When scaling to tens of thousands of GPUs, overall compute efficiency can plummet below 40%, with more than half of compute power consumed by data transfer and waiting — a phenomenon known as the 'communication wall.'
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Related Claims
Unverified当AI集群规模扩展到数万张GPU时，整体效率可能暴跌到40%以下，超过一半的算力被消耗在数据传输和等待过程中80% similarUnverified对于MuJoCo、Atari等环境，单个环境的采样速度往往跟不上GPU计算能力，导致GPU利用率长期停留在20%–40%甚至更低68% similarUnverified自回归解码阶段GPU算力利用率往往不足10%，属于内存密集型操作65% similarUnverifiedGPU利用率可从低流量时的20%~30%提升至高流量时的70%~80%，因请求量越大越容易聚合为更大计算批次64% similarUnverified流量规模化后GPU批处理利用率可从低流量时的20%~30%提升至高流量时的70%~80%，形成规模经济和规模壁垒62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/49977API
curl https://kongchang.com/api/v1/knowledge/claims/49977MCP
get_claim(id=49977)