VRAY5 GPU = render time variances

[S2Architecture]

I have one file that i created. Using GPU rendering it takes 18 minutes. I save a duplicate of the file, swap two materials, go to render and it take hours to render. render settings are identical. materials are not complex. one thing i did notice different from one file to the other (and no idea why) is the cars are a different color. I am using a multi sub texture to drive color of paint and it is set to render ID so i just assume the computer assigned a different render ID when i submitted the job for rendering.

why is there such a difference in render times. both files will render in 1,5 hrs using standard VRAY 5 CPU rendering.

My hardware is dual xeon processors with 64 threads, 128GB ram, and an RTX 3090.

Let me know what I could be missing. thanks.

[S2Architecture]

forgot to mention, the quick one finished with less than 6000 passes. the longer one is at 150000 passes so far and still going.and the 6000 pass one looks better so far,

[hermit.crab]

Does the render time difference occur when rendering an identical file (no color changes)? Does the slowdown occur when rendering locally? Also, when comparing CPU and GPU performance you would need to compare strictly by noise threshold; make sure the samples are high for both engines and compare with an identical noise threshold. GPUs are generally faster, so the time differences may not necessarily be abnormal in your specific scenario. Do send a scene via our contact form if your concern persists. Mention this forum thread in the message along with the results from your comparison.

[S2Architecture]

I created one file. set up the cameras. did a save as, swapped 2 materials. i did another save as and swapped two materials. so now i have 3 scenes that are the same in regards to content and camera position and sun angle. the noise threshold is set to .007 for all scenes, passes set to 5000, time set to 0. i rendered the first one in 18 minutes, it showed just under 6000 passes completed/, started the second one, is was showing 200000 passes after 2 hours, so i just stopped it and saved the images and am using those, rendered the third one and it finished in 18 minutes with just under 6000 passes. i got what i needed so no big deal. but it doesn't make any sense as to why one of the three files would render so differently. appearance wise was the same other than the time and amount of passes.
i rendered the scenes in CPU mode earlier in the week with an alternative camera angle. all three rendered in the same amount of time. approx 1hr 38 minutes if i remember correctly. i installed the RTX 3090 and wanted to try out the GPU rendering. that is when i noticed the discrepancy. I was just curious if this was common and to make a mental note to watch for it. i want to leverage the GPU rendering more in the future and am in the testing phase now. I can continue to render in CPU mode if need be, just waiting for my render farm to be upgraded to the new VRAY5. then time is not an issue anymore either way.

[hermit.crab]

Yes, it seems possible that your GPU rendering test has not yet reached the desired threshold. Try pumping up the samples and add a VRaySampleRate element to track the result data.

[S2Architecture]

how is it possible one file reaches its threshold, but the other two don't? they are all the same files with the exception of two materials.6000 vs 200000 passes.

[hermit.crab]

Well, the sampler limit may be reached before the noise threshold. Add a VRay Sampler Rate element and check the Sampler Rate page in the documentation to see if that is so.

[S2Architecture]

here are the two to compare. this is a difference. but there is no reason for this. the render settings are identical.

[hermit.crab]

Do send a scene via our contact form so we can have a look. Mention this thread in the message.

[S2Architecture]

i can not obtain permissions to send the file, sorry. i know this doesnt help the situation but my hands are tied.