跳到正文
原文
SemiAnalysis_· @SemiAnalysis_ · X·· 1 天前AI 评分50

Google TPUv7 FP8 vLLM 性价比超 GB300

AI 导读

突发新闻🚨:Google 已向我们的 Agentic Inference Benchmark 提交了首批 TPUv7 FP8 vLLM 结果,其每美元性能比 GB300 NVL72 FP4 还高出 1.67 倍。🚀 看到 Google 全力推进,将 TPU 的能力扩展到 Google 之外,令人惊叹。👇️ (1/4)🧵

正文 · 原文

BREAKING NEWS🚨: Google has submitted its FIRST TPUv7 FP8 vLLM results to our Agentic Inference Benchmark and has demonstrated 1.67x better perf per dollar than even GB300 NVL72 FP4. 🚀

It is amazing to see Google go full steam ahead with externalizing the power of TPUs outside of Google.👇️ (1/4)🧵

来源:SemiAnalysis_ · x.com