Iray rendering questions:
areg5
Posts: 617
First question: I have the max samples and time maxed out. So, the rendering typically stops at 15000 samples (sometimes 3 1/2 hr). Is there a significant difference in quality if I max the samples at say 10000 or lower?
Next question: I have an i7 4790 with 16 gig Ram and an nvida 4 gig 970 gtx. If a rendering takes 3 hrs, how much faster (if any) would it be if I added a second 970 card? Also, could my 750 watt power supply run 2 cards?
Post edited by areg5 on
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
Does the scene fit into the current card's memory? Having to use such high values does suggest, if the scene isn't stopping for convergence, that you could do with revising it - adding more light, for example, and then adjusting the exposure.
I'm not sure what you're asking. The file size is about 85 MB. I'm doing 1280X720 image size, HD. The lighting seems ok, this one is using an overhead mesh light and took about 3 hours, and din't completely comverge.
Download GPUz; once installed, you can select the graphics card, it will tell you what resources are being used. The best way of telling if its being used or not.
I downloaded it. What do I do with the results? The card is being used.
You need to look at the Sensors tab in that app. It has real-time running bar graphs for how much memory is being used, GPU load, core clock speed, etc. If you have insufficient memory for the scene you will see the memory used max out then be dumped as it falls back to CPU only (and corresponding drops in GPU load and clock speed).
I am willing to bet you could cut your render time, perhaps significantly, by getting rid of the "overhead mesh light". And that's not just my bias showing.
Do mesh lights slow it down?
Potentially. There's a lot of factors at work there.
I use them spareingly when I use them at all.
Thanks for the advice! I'll stick with spots and points.
Ok. Definitely seems like mesh lights slow things down. Much faster with photometric spots. Something else: it seems like the more lights you add, or turn up their intensity, it all renders faster, but you might not see a big difference in lighting of the figures. I may be off base here, but that's what it seems like. With 3Delight, as soon as you turn up the spots the image is brighter. In Iray, when you turn up the spots the main effect seems to be faster rendering. I was wondering if anyone else noticed that.
An easy way to tell if it is using GPU or ZPU. In studio, go to help menu > troubleshooting > log file. CTRL + A, backspace, CTRL +S. Then start render. If it has to fallback to CPU for render due to memory, it will say it in the log. I forget the exact phrase it uses, but it's fairly obvious. Not sure what text program you use, but I use notepad++ and it will refresh it automatically for me, you might have to push F5 once in a while to get it to refresh, not too sure though.
Also there is this in store, http://www.daz3d.com/iray-memory-assistant ; very helpful to estimate.
I definitely have the memory. GPU is maxing out.
For Iray lights, intensity isn't the right parameter. For brighter lights, add more lumens. For more light, add more lights or more lumens to the existing lights. Iray will render faster with more light. If the scene is too bright, you can keep the lights but change the exposure in the tone mapping tab.
Yeah, I know about the lumens thing. That's what I meant about intensity.
Yep, that shows you are using GPU.
As I come from a Reality/Luxrender background I naturally kept on using mesh lights, but I found DAZ pm lights preferable after some experimentation. However I use the mesh options in them as I still dislike point lights in PRB rendering because of their harsh unnatural shadows and tendency towards shading artifacts on organic surfaces (to be fair, harsh HDRI can do this too)—they aren't in tune with the PBR paradigm. I also generally attach an ies profile to them to give them a predictable "shape". There are some HDRI mesh light products available though that can work quite well (e.g. Dimension Theory's iRadiance). I normally rely on HDRI environment lights though, even for interior scenes, as long as I can eliminate walls and ceilings. As long as they are true HRDI (some simply do not have sufficient dynamic range) I find those the most efficient way of lighting. Still, I am keeping my mind open as to lighting options.
What you also need to bear in mind when using DAZ pm lights is that the default tone mapping settings are for bright sunlight and the pm lights emulate artificial lights. Unless artificial lights are arc lamps or those incredibly powerful daylight movie lights (and then the lumens will be very high), then you need to increase the exposure to the sort of parameters that would be used for standard artificial lighting with a traditional camera. E.g. F1.4 and 1/60th or higher in terms of exposure.
While I prefer a reasonable facsimile of physical plausibility, Tone Mapping is slow. Sometimes the best thing you can do is turn the lights up.
Or render to EXR and tone map externally.
That's it exactly. There was a photography page where the photographer described his studio setup. His thesis was that fill light should be basically ambient (like a cloudy day) with no shadows or specular reflections. To do this in a white-walled studio, he mounted six powerful spots and pointed them at the ceiling. The result was bounce-only light which he used as the basis for his photos (setting his exposure using this as a baseline, pretty much always f/8). For fun, I tried to mimic that in Iray (I think the lights were 400W each). The result was indeed a very even fill. But it took a ridiculously long time.
We can do real-world lighting with parameters that simulate real-world lights in Iray. But we can buy _much_ brighter bulbs than real photographers.