Corbenic 推 SSD 缓存 KV，14 倍加速遭质疑
TechRadar 报道，比利时初创公司 Corbenic AI 的 Galahad 于 2026 年 10 月 1 日进入测试版，可把模型的 KV 缓存写入加密的本地 NVMe 固态盘，在 GPU 显存被清空或服务器重启后再次加载，从而不必重新读取同一段文本。创始人 Sietse Schelpe 称只需 Linux、Nvidia GPU 和普通 NVMe 盘，不需要特殊硬件。但报道指出，论文中约 14 倍的整体加速主要来自检索组件 Blaise（它只把约 668 个 token 的相关片段交给模型），而非从 SSD 载入；单用 KV 组件 Taliesin 时，vLLM 每题耗时 15.56 秒，反而慢于 8.11 秒的基线，SGLang 为 22.02 秒对 10.29 秒。报道还提到 LMCache、Nvidia Dynamo、WEKA、Phison 等已有类似方案，且 Corbenic 的论文数据均由创始人本人在租用云 GPU 上测得、为自发布预印本，原始日志尚未公开。
报道指出，反复让 GPU 重读同一批合同、手册和代码确实存在浪费，而 LMCache、Nvidia Dynamo 等方案已能将 KV 缓存持久化到存储，因此 Galahad 的差异化在于精确、失败即回退的单库复用，而非速度数字本身。
- Corbenic AI's Galahad beta saves a model's KV cache to encrypted local NVMe so servers can skip rereading text after GPU memory is cleared or after a restart
- Founder Sietse Schelpe tells TechRadar Pro no special hardware is needed, but the paper credits most of the 14x speedup to Blaise, a retrieval component, not SSD loading
- LMCache, Nvidia Dynamo and CMX, WEKA and Phison already persist KV cache to storage; Galahad's edge is exact, fail-closed reuse in one small library
Belgian startup Corbenic AI says it can stop GPUs from rereading documents they have already processed, and its founder told TechRadar Pro the trick needs nothing more exotic than the NVMe drive already sitting in the server.
Galahad, which entered beta on October 1 2026, saves a model's key-value (KV) cache, the attention state built while a model reads a prompt, to encrypted local storage and reloads it the next time the same text appears.
In an email to TechRadar Pro, Corbenic introduced Galahad with the line "the GPU never reads the same text twice," and founder and CEO Sietse Schelpe summed it up as "the SSD a server already has becomes part of the AI's working memory."
#Corbenic AI#Galahad#Taliesin#Blaise#LMCache#Nvidia#Phison#vLLM#SGLang#Gemma 4 31B#Tensormesh