News for hackers. Stuff that matters.

A comment on A 10 year old Xeon is all you need (for 26B-A4B MTP Drafters without GPU). Parent comment

Re:A 10 year old Xeon is all you need (for 26B-A4B MTP Draft... by Havoc on Monday June 01, 2026 @04:52AM (#48354225)
In general you’re mem bandwidth constrained so cpu vs gpu often ends up similar on APUs
[ Parent | on HN ]
Re:A 10 year old Xeon is all you need (for 26B-A4B MTP Draft... by fulafel on Monday June 01, 2026 @05:35AM (#48354570)
There are ways to trade off compute power for memory bandwidth (like MTP and other speculative decoding approaches). The CPU and GPU would need to be able to share the same cache for this to work. In the Strix Halo case the GPU has a private cache on the GPU die I think, which is the snag.
[ Parent | on HN ]