Feed Entities Podcasts Deep Dives Developers About

[KongchangAI]

Real-time AI knowledge base. Every claim is traceable to its source, with data freshness under 24 hours.

Powered by 240+ first-hand AI sources

Knowledge

Feed
Hot Entities
Podcasts

Developers

CLI
MCP Server
REST API
About

© 2026 KongchangAI. The freshest, most accurate AI knowledge

Unverified50% confidenceFactExact time

vLLM的PagedAttention将KV缓存切分为固定大小物理块（通常16个token），使显存利用率从不足40%提升至接近90%

1

Sources

50%

Confidence

Long-term

Relevance

7/19/2026

First Seen

Sources

块低秩压缩：内存受限GPU上的大模型推理加速指南

Related Claims

UnverifiedvLLM的PagedAttention将KV Cache划分为固定大小物理块，通过块表实现地址映射，将碎片化率从60-80%降至不足4%，吞吐量提升可达24倍85% similar UnverifiedPagedAttention将KV Cache拆分为固定大小的物理块通过块表映射，GPU利用率可从低于40%提升至90%以上83% similar UnverifiedvLLM通过PagedAttention技术将KV Cache的内存碎片率从60%以上压缩至接近零，在高并发场景下可将推理吞吐量提升数倍78% similar UnverifiedvLLM引入PagedAttention技术，将KV Cache分割为固定大小物理块按需动态分配，多并发场景下显存利用率可比朴素实现提升3倍以上77% similar UnverifiedvLLM的PagedAttention将KV Cache碎片化率从传统方案的60-80%降至不足4%，吞吐量提升可达24倍75% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/561356

API

curl https://kongchang.com/api/v1/knowledge/claims/561356

MCP

get_claim(id=561356)