Hacker News new | past | comments | ask | show | jobs | submit
mdp2021 | on: AMD acquires Taalas to boost inference performance by etching models in silicon
> A small model would be [mobile size]

A ~30mm side for the HC1 tech for an 8b model (still unclear the planned HC2)?

teaearlgraycold | parent
Is that analogue or are they baking floating points into the silicon?
AlotOfReading | root | parent
It's entirely possible they're using something like block floating point, where most of the hardware is simply fixed point. AMD's NPU does this, for example.