>>106314284
It works but it's slow.
GGUF Q6 Wan 2.2 with 4+4 step turbo model takes about 360s with 81 frames at 640x480 81 frames.
There's no practical SageAttention or XFormers tricks to speed things up and save memory.
Also currently in comfy chroma/flux is broken unless you select slightly slower quad cross attention. Running Linux of course.
4chan Search
2 results for "3f6cfa443738dfbc6f1d8a0c2229bb66"
>>105933364
they ran so out of money that they could only afford engineers, and as such they started making useful products instead of marketable ones
they're currently on the other side of the bell curve now tho, AMD is getting fat. they spend like 100x more on R&D now than when they developed zen, yet radeon and zen are making less progress at longer cadence
they ran so out of money that they could only afford engineers, and as such they started making useful products instead of marketable ones
they're currently on the other side of the bell curve now tho, AMD is getting fat. they spend like 100x more on R&D now than when they developed zen, yet radeon and zen are making less progress at longer cadence