Cradle Codec: GPU-Native KV Cache Compression, Like NCCL Over Ethernet
knowrohit07 · x · 2026-07-29
Cradle Codec is a GPU-native KV cache compression technology that addresses the bottleneck of KV cache storage and cross-node transfer. It enables communication similar to NCCL over Ethernet, allowing large states from reusable prefixes to be moved efficiently across nodes without NVLink, potentially reducing storage and transfer costs.
Related event: Cradle Codec: GPU-native KV Cache Compression for Ethernet Transfer(4 posts)→
More from Infra
- Power-management chips: the invisible layer deciding AI data center energy efficiency — Olivier__OG · 2026-10-17
- MiniCPM-V 4.7 runs a 35B multimodal model in 3GB of memory, fully on-device — alifcoder · 2026-10-17
- Full FP8 n-gram table visibly improves local Qwen3.8-Flash-Next at zero speed cost — sillynoobhorse · 2026-10-17
- Local Nemotron 3.5 on one A100 hits 169 tok/s, processes 100 clinical notes for 2.4 cents — MaziyarPanahi · 2026-10-17
- NVIDIA's open-source AIPerf benchmarking tool and Triton GPU tutorials for inference measurement — techNmak · 2026-10-17
- Official Triton GPU Programming Tutorials: Vector Add to Fused Attention — techNmak · 2026-10-17