You're viewing a single thread.
I think they just got a huge boost from the vulkan thing and with open Nvidia modules. But the tests seemed to indicate it was still something like 60% of the performance and only on some specific cards. Sad.
It would be so nice if the Nvidia situation was as good as AMD situation or especially Intel situation right now.
And the worst part: if you have actual needs for ML stuff - you kinda are stuck with nvidia cards, so you can't even switch when the need for upgrades arise.
You can do that on AMD but it is a bit slower and takes a little more time
Not in my experience. ComfyUI segfaults on a 7900 XT, EndeavourOS, no matter what I do.
I'm not sure if we're ever going to get the full potential of a card out of NVK, but the fact that running them above base clock is now possible is pretty huge already.
(Unless I misunderstood something, but I believe that's now a thing)