On: AMD acquires Taalas to boost inference performance by etching models in silicon

[trebligdivad]

What, even if it means you can run models without relying on the currently backlogged DRAM production?

[adgjlsfhk1 → trebligdivad]

The size of model we're talking about running doesn't need much if any dram.

[trollbridge → adgjlsfhk1]

The chatjimmy demo is using a model that needs 6-18GB of VRAM. That's not exactly trivial.

I could see it being feasible to get a Qwen-3.6-27b type of model done on something like this. Qwen-3.6-27b at 18tok/s would be a game changer.

[adgjlsfhk1 → trollbridge]

right, but that's a reticle size chip. to put something in a phone it has to be ~10-30x smaller