Anyone had any luck bringing down 4.11 Iray rendertimes?
simpsonDDS2
Posts: 30
Obviously 4.11 uses a new iray renderer and as such things are much, much slower than 4.10.
Has anyone figured out any setting changes that mitigate this exponential increase in rendertimes?
Appreciate anyone who gives some answers, thanks!
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
4.11 is slower than 4.10 and takes more vram. So scenes that work in 4.10 may not work in 4.11. That said, 4.11 converges better and gets the denoiser. So it should take less iterations to get the same quality. And if you are fine with the denoiser "smoothing" then 4.11 is light speed compared to 4.10.
Thanks for the response. How do I switch on the denoiser?
Under: Render Settings / Filtering
Both these settings should be ON:
Post Denoiser Available
Post Denoiser Enable
I agree with Padone; 4.11 is much, much faster with denoiser and IMHO it is not bad for skins as long as you don't shorten the number of iterations to much. Very often 500 iterations is more than appropriate, and you will get a very nice preview after just 10-20 iterations.
If you set the denoiser to run after a reasonably high number of iterations, 500+, you give it a lot more info to work with and you get much sharper images. The default 8 iterations will pretty much always wash out all detail.
That's so odd as I've noticed that 4.11 seems quicker and uses less VRAM. I'll count myself as lucky!
I have the same experience.
I have seen other people saying the same thing, but in my experience its ...completly wrong. Let me explain:
I have tested doing renders with both these setups:
1) 500 iterations, denoiser starts after 8th iteration
2) 500 iterations, denoiser starts after 499th iteration
For me, it is impossible to see any difference between the two completed renders.
From what I can deduct; the denoiser is improving the latest canvas, BUT, when the rendering continues, the next canvas is NOT based on or includes the result from the denoiser, the canvas just improves with every iteration and the denoiser has a better picture to work with. Hence, it doesn't matter WHEN the denoiser kicks in, the important thing is to have enough iterations.
If the above is correct, it would mean that the denoiser would kick in numerous times during a render, but it would only be the last denois-session that would count, hence, it is a waste of resources to have the denoiser kicking in early when doing batch-renders but very resonable when doing previews.
If I am wrong I would be happy to know...!
The denoiser doesn't seem to run after every iteration. If it did it woukld drastically increase render times after it starts. I've tested various permutations of using the denoiser and that just isn't what happens. The iteration when the denoiser runs takes longer, noticeably so if the scene is complicated, and then iteration times return to normal.
I have no idea what scene you're using or how much detail it has etc, But get an image comparison program and run the two against each other. See if they truly aren't distinguishable.
I ran a render this morning that I stopped after 250 iterations because it looked like some reflection areas were not going to work well. I started again with the denoiser enabled to go at 250 iterations. It didn't kick in until 270 iterations. I'd say you need to start the denoiser earlier than you might need it towards the end of the render. In this case, maybe at 450 iterations.
I believe we did see an nVidia statement that the end result did not depend on when the denoiser was started, but obviously it is going to have an impact on the tiem taken to reach that state.
Thanks! Nice to know for sure.
Sometimes the denoiser just don't kick in regardless of the setting... It happens quit regularly when I am working back and forth but never when I am running batch-renders.
I've had this experience. I've had scenes that rendered just fine on the GPU but the denoiser wouldn't kick in at all...but another scene of a similar size it did. I have no idea why.
This is a new bug in the 4.11 release. It never happened in the beta's. I ran hundreds of renders using the denoiser without it failing once during the beta. Now it fails, forme, about one third of time.
Personally I set the iterations to 128 or 256, very rarely 512. It seems iray needs more iterations when complex reflections are into play specially for water. Then I set the denoiser to run just at the last iteration, that's 128 or 256.
I feel running the denoiser multiple times to be not useful and to slow down rendering and may also take more vram. That could also explain when it doesn't start.
Also using optix seems to work fine on my system and it does speed up by 30% or so. But in some discussions they say it may reverse to cpu depending on the card and drivers so I guess it's a bit of a gambling.
It's the same Iray in both versions, so that has to be be down to soem other factor - perhaps one version is running out of GPU memory, for example (the denoiser requires the GPU) due to some apparently trivial setting difference.
No. There are clearly differences in DS. I'm very sure I'm not running oou of VRAM, which would cause the render to drop to CPU anyway.
I had a problem with the Noise Degrain Filtering in 4.11. My renders looked aweful, so I turned it off and that cut my render times about in half. Now scenes that I rendered in 4.10 render just as quickly in 4.11. I've never used noise filtering before because I rarely have problems with noise, but I tried out the new Denoiser system and it didn't have any noticeable effect on render times.
Update on my end:
Render times are infact faster with higher quality after using Denoiser. So thanks for that solution
However, I've discovered that (alongside other people in this thread) Iray rendering doesn't use my GPU despite it getting ticked. Occasionally the GPU will decide to wake up and help out - in which case render times are much faster - but only occasionally. More often the GPU reads 0% while Daz tries to render a scene.
Be aware that the task manager only reads iray if you set the graph to CUDA instead of 3D, otherwise it reads zero but the gpu is used. You get a better reading from gpu-z. And you can always check the log anyway for cpu reversing. As for me it seems everything works fine here but may be I'm just lucky with my configuration.