{"type":"link","version":"1.0","title":"Offloading key-value cache only helps while the per-step transfer overlaps with compute; past that point it subtracts throughput","author_name":"AI Archs","author_url":"https://ai-arch.pages.dev","provider_name":"AI Archs","provider_url":"https://ai-arch.pages.dev","url":"https://ai-arch.pages.dev/n/risk-cross-node-kv-transfer-rate-caps-offload-benefit","thumbnail_url":"https://ai-arch.pages.dev/og/risk-cross-node-kv-transfer-rate-caps-offload-benefit.png","thumbnail_width":1200,"thumbnail_height":630}