Faceswap Forum

I just did an experiment, and I have to ask, Is it possible using the CPU to mask is actually faster than the GPU?

I was testing a 4k movie because it seemed to me using the CPU last night to "re-mask" while I was playing video games was actually slightly faster than using the GPU. Is that possible? Or was it just a lucky draw? I'm using a 3080ti and a i7 8700k.

Thoughts?

Which mask were you using? If landmarks based mask, then the GPU isn't used at all. The mask is generated from landmarks.

If it was an NN based mask, it would seem unlikely. However, let me know which one, because if it is as fast on CPU, then that is something I would want to look at.

It was on Bisenet, but now I'm unsure, and I'm going to test this further before I make a fool of myself and insist I'm right. Lol. If it's not faster or the same it's definitely close (a couple minutes slower perhaps.)

Okay, I did a comparison of masking GPU vs. CPU. Unfortunately I was unable to get the exact time for CPU usage as I walked away and forgot to make a note. I can say with some accuracy the CPU was between 1 hour 10 minutes and 1 hour 20 minutes (probably around 1hr 15 min.) The GPU was 1 hour 16 minutes. This is updating 33,801 masks on a 59,192 frame 4k movie, Bisenet.

So CPU masking (on an i7 8700k, w/32g memory) is absolutely comparable to using the GPU (Evga 3080ti hybrid.). If not it's within 5 minutes at the most. (And, if it is 5 minutes extra on a CPU, that's not so bad if I want to do other stuff while waiting like gaming, etc.)

Ok, good to know. Thanks. I will look at the potential for choosing to put that on cpu.