flowframes icon indicating copy to clipboard operation
flowframes copied to clipboard

Flowframes won't utilise all 3 of my GPU's

Open agentdark45 opened this issue 3 years ago • 8 comments

Hi,

I'm currently interpolating some 4K frames using the XVFI model and no matter what I set in the dedicated GPU options (0,1,2 - in both fields), only one of my 3x3090 GPU's is utilised for the interpolation.

If anyone could point me in the right direction that would be great.

I'm using the latest build on Patreon.

Thanks in advance for any help.

agentdark45 avatar Feb 21 '22 12:02 agentdark45

Multi-GPU setups aren't supported. You could however use multiple instances, each running on a different GPU.

PweSol avatar Mar 14 '22 08:03 PweSol

Thanks for the tipp PweSol! Because, yeah i noticed that to. No Multi-GPU even if there is a setting with multi-GPU. When I write "0" it uses my main GPU, if I write "1" it uses my second GPU. If i write 0,1 it uses my main GPU and if I write 1,0 it uses my second GPU... Strange setting. Seems useless :D

But nice software

Platorius avatar Apr 07 '22 16:04 Platorius

Ok, this does not work for me. I tried to use 2 instances, but when i change the GPU Priority in one window, it changes it too in the other.

How do you do that, PweSol?

Platorius avatar Apr 07 '22 18:04 Platorius

Have you tried duplicating the flowframes install folder? (copy it and paste in the same place

itsTyrion avatar Apr 27 '22 19:04 itsTyrion

I have the same issue that only 1 GPU is used, even though I specified multiple ones in the setting.

edwin-yan avatar Jul 16 '22 05:07 edwin-yan

I can confirm that the dual GPU works with RIFE NCCN but not RIFE CUDA. I am using the latest version (1.35.0)

edwin-yan avatar Jul 16 '22 06:07 edwin-yan

Based on my quick tests, dual 3090 using RIFE NCCN is about 1.5x-1.6x faster than single 3090 using RIFE CUDA. However, both 3090 have almost 100% usage on RIFE NCCN while the usage is only around 20%-30% using RIFE CUDA. Each of my overclocked card pulls over 450W on 100% usage. I am not sure if I should just stick with single card using RIFE CUDA...

edwin-yan avatar Jul 16 '22 06:07 edwin-yan

Same problem with 2x MSI GTX 1660 Super X Gaming 6GB only one card used. I can run on 1'st or 2'nd card, but not both. This is with latest v1.40 When will this feature will be working, or bug fixed?

akierum avatar May 07 '23 12:05 akierum