Meta MTIA v2
芯片
AI加速器
发货中
Meta's recommendation-inference chip — custom silicon for the ads engine.
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。
MTIA is Meta's inference accelerator for the ranking and recommendation models that drive its feed and ads. Version 2 (2024) tripled compute over v1 within a 90 W envelope using large on-chip SRAM instead of HBM — a different design point from LLM chips, tuned for Meta's own workloads. Training-class successors have been reported as in development.
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
Meta
Announced
April 2024
Manufacturing
Foundry / process
TSMC N5Vendor spec
算力
INT8
354 (dense) / 708 (sparse) TOPSVendor spec
内存
On-chip memory
256 MB SRAM; LPDDR5 off-chipVendor spec
电力
TDP
90 WVendor 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
来源
记录审阅截至 2026-01-15 — 查看 方法论.