Post #4277908
2026-07-31 12:17 UTC
@mia@movsw.0x0.st i think the igpu actually performs better at ai inferrence too. No one understands the purpose of those coprocessors
Replies (1)
-
@mia@movsw.0x0.st 2026-07-31 13:01
@snacks@netzsphaere.xyz something like, the npu only accelerates time to first token before the gpu kicks in? idk funny to go from my permanently bandwidth-starved intel igpu to a platform with 200-500GB/s memory that can be dynamically pinned for GPU use it is so stupidly fast for a laptop