Groq LPU
芯片
AI加速器
发货中
The deterministic inference engine famous for extreme tokens-per-second.
您正在查看一条芯片记录：该处理器是什么、其关键规格、谁设计它、谁制造它、谁供应其零部件，以及它被用在哪里。
Groq's Language Processing Unit strips out caches and dynamic scheduling: the compiler knows exactly where every byte is every cycle, using 230 MB of on-chip SRAM at 80 TB/s. Made on GlobalFoundries 14 nm — proof that architecture, not process node, drives its speed — LPU racks serve open models at hundreds to thousands of tokens per second. Founded by TPU creator Jonathan Ross; valued in the billions on 2024–25 raises with a major Saudi deployment deal.
规格与关键参数
每个数据均标注来源依据：厂商规格、厂商声明、公开报道、估算或官方数据。"未公开披露"即字面含义——我们从不捏造数值。
Overview
Designer
Groq
Architecture
Tensor streaming, deterministic
Scale model
Hundreds of chips pipelined per model; no HBM anywhereVendor spec
Manufacturing
Foundry / process
GlobalFoundries 14 nmVendor spec
内存
On-chip memory
230 MB SRAM, ~80 TB/sVendor spec
算力
INT8
750 TOPSVendor spec
它如何与整个技术栈相连
本记录的关联链接一览——点击分组可直接跳转。
设计方
由……制造
相关
来源
记录审阅截至 2026-01-15 — 查看 方法论.