Architecture of Meta's First-Generation AI Inference Accelerator

More details on MTIA, Meta Training and Inference Accelerator ➡️ bit.ly/3IUyYXd
A closer look at the architecture of Meta's first-generation AI inference accelerator. MTIA v1 is a custom-designed ASIC built with next-generation recommendation model requirements in mind.
The accelerator is built on TSMC's 7nm process at a 25W TDP and provides 102.4 TOPS at INT8 precision and 51.2 TFLOPS at FP16 precision.
We're sharing more details on this work at #ISCA2023.
--
Subscribe: www.youtube.co...
Learn more about our work: ai.facebook.com
Follow us on Twitter: / aiatmeta
Follow us on Facebook: / aiatmeta
Connect with us on LinkedIn: / aiatmeta
Working on AI at Meta, we're bringing the world together by advancing AI, powering meaningful and safe experiences, and conducting open research.