Wired issue

[biochemical_animations]

When rendering to production rt cuda on my WS, it takes 2 minutes. When I send the render through backburner, the frames render longer on my WS. I see in the message window noise 0.1 achieved in about 2 minutes and all gpus were working, then 1 gpu works for about 6 minutes on the part where it says 12% complete, then 22% complete then goes thru the changing rays per pixel stuff.

I have 14 lights in this segment on uniform probabilistic lighting.

Has anyone seen this, know what it is, or know what I can try?

Using vray rt cuda 3.5.04

[biochemical_animations]

Maybe it's the denoiser elements?

[biochemical_animations]

Running some tests today to see if adding in the denoiser elements is slower then just rendering to a lower noise number?

Does rendering the denoiser elements only use 1-2 GPUs? When there are more GPUs on the system?

EDIT: Nope, setting the noise lower results in longer render times and more noisy image then the denoiser method...I'll keep testing.

[biochemical_animations]

Some tests:
Denoiser render element off.
Render using BF+LC noise at 0.1 time was 4:53
Render using BF+BF noise at 0.1 time was 3:16
Render again using BF+LC noise at 0.1 time was 3:32

...puzzling. When I did my sequence render yesterday, I was expecting faster render times because when I was making the scene I was getting good render times...but when I saw much longer times, I thought I should test today.

[biochemical_animations]

Ok, going with BF+BF noise at 0.1, enabling denoiser elements, takes around 3 minutes to get to 0.1 noise as stated in the message window and all GPUs are used, then after this only one GPU is used for the 12%...22%...Number of rays per pixel changed to 109, total samples etc...then one or two GPUs are used not fully, but they bounce around for the remainder of the render which finishes at 8:56.

So 3 minutes to noise 0.1 and 6 minutes for the rest...is the rest the denoiser elements?

I'm just trying to understand this...thanks.

And are the denoise elements not multi-GPU optimized yet?

[biochemical_animations]

Some more tests:
Denoiser render element off.
14 lights in scene
BF+BF noise at 0.1 uniform probabilistic (8 lights used) time was 2:49
BF+BF noise at 0.1 full lights evaluation time was 2:49 (identical image to uniform prob above)
BF+BF noise at 0.1 adaptive lighting (8 value set at spinner) time was 4:49 (image was more dark shaded then above images) (message window stated it was using probabilistic lighting (14 lights found, 8 used) eventhough adaptive lighting was selected in global switches.

[biochemical_animations]

Are the denoised render elements not multi-gpu?

[piotrus3333]

looks like you're rendering animation and denoising after every frame. - It would be better to run Denoiser after you have all the frames in multichannel exrs. works better as it takes previous and next frame into consideration. plus you can test both denoising types and different settings to get optimal results.

as far as utilizing all the gpus you can try running two or more instances of vDenoise tool (get it if you are not using it). have not tried it though.

adaptive lights need LC to function.

[biochemical_animations]

I'm rendering to "only generate render elements" then use stand alone denoiser afterwards on the sequence.