HOME | DD

#dazstudio #iray #lighting #tutorial #sickleyield
Published: 2016-10-12 05:15:13 +0000 UTC; Views: 4216; Favourites: 29; Downloads: 77
Redirect to original
Description
The purpose of this scene is to test1. Time vs. number of photometric lights
2. Whether photometric lights still have an added effect with more than 7 in a scene (someone has anecdotally claimed they do not)
It contains free items from the Daz Studio Starter Essentials only (the Alchemy Chasm, the default HDR, and photometric lights generated within DS). There are three groups you can hide or show to add or subtract their lights: A single photometric light, a group of 7 photometric lights, and a group of 50 photometric lights.
I'm rendering with 2x GTX 980, 2x GTX 740, and a Core i7 CPU.
My render results with the test scene are:
Single photometric light: 13 sec
7 photometric lights: 50 sec
50 photometric lights: 1 min 53 sec
Oddly enough, this appears to refute the claim that more photometrics are faster (something I've heard repeated as firm fact since Iray came out). I can also conclusively refute the idea that 7+ photometrics have no effect. The scene with 50 lights is noticeably brighter (which is why I've included it as the preview image for the test scene).
I redid the render with an identical group of 7 photometric lights set to a much lower light level (also included in the test scene) to control for whether higher or lower light level really has an effect.
7 photometric lights: 50 sec
7 DIM photometric lights: 13 sec
Confusingly, this appears to suggest the opposite of the doctrine that Iray prefers more light in the scene. When I turned off the HDR (by setting it to Scene Only in Render Settings) it went to 7 sec, even faster!
A quick test of my own wherein I set the top of the roof to a separate material group and hid it made all scenes faster, but their ratios remained the same.
Next I will try again with the same lighting but more open geometry and see if that's making a difference. Iray does not like partly or entirely enclosed geometry in a scene, maybe that's interfering somehow.
Related content
Comments: 24
BryonSmothers [2017-05-30 15:38:02 +0000 UTC]
While this is nice. I want to talk about that rig you have 2x GTX 980, 2x GTX 740 so 4 graphics cards for rendering?
👍: 0 ⏩: 1
SickleYield In reply to BryonSmothers [2017-05-30 17:08:53 +0000 UTC]
At the time of that post I did.
Now I have one GTX 1080, one GTX 980, and one GT 740 for running the monitor (unchecked in the render settings and not used in Iray). Iray doesn't use multiple cards very well so your render speed depends more on the speed of the fastest card. The 1080 is physically huge inside the case, so I decided it was better to get rid of one of the 740's to accommodate it. My rendering is much faster now as a result.
👍: 0 ⏩: 1
BryonSmothers In reply to SickleYield [2017-05-30 17:29:28 +0000 UTC]
I was wondering about that I render only in iray. I have a gtx 980ti. I would like a 1080 but only if it will make a large difference in render time.
👍: 0 ⏩: 0
Tarbicus [2016-10-21 22:06:47 +0000 UTC]
Some tips in the iray Prgrammers Manual. Worth a read.
www.migenius.com/doc/realityse…
👍: 0 ⏩: 0
SlimMckenzie [2016-10-13 11:42:56 +0000 UTC]
Thank you for taking the time to work (and share) this info with us here on DA. Your assistance is well appreciated!
👍: 0 ⏩: 0
Tarbicus [2016-10-12 21:44:25 +0000 UTC]
I suspect the reason iray *might* render faster with more light is because it has less need to filter the fireflies that increase necessary render iterations. You should also consider trying the Architectural shader and see if that helps, as I think it's designed for enclosed spaces.
👍: 0 ⏩: 1
SickleYield In reply to Tarbicus [2016-10-12 21:59:19 +0000 UTC]
The architectural option is NEVER worth it. It slows down every situation I've ever tried it in.
👍: 0 ⏩: 2
Tarbicus In reply to SickleYield [2016-10-12 22:24:28 +0000 UTC]
You might also want to try switching on the Architectural Sampler in Optimisation.
👍: 0 ⏩: 2
Dick--Justice In reply to Tarbicus [2016-10-13 20:38:42 +0000 UTC]
Architectural sampler utterly kills performance most of the time. Seems it also has some sort of effect on VRAM, because busier scenes for me tend to drop back to CPU rendering with the sampler enabled. Plus, it strengthens specularity in my scenes beyond what's shown in the viewport preview, making it difficult to anticipate the final render.
👍: 0 ⏩: 1
Tarbicus In reply to Dick--Justice [2016-10-13 22:31:58 +0000 UTC]
The Architectural sampler is designed for interior scenes and reduces noise. The NVidia iray blog tells us "...don’t be disappointed if the rendering after some dozen frames still looks more noisy than the “classic” version, most of the time the architectural sampler will catch up later-on." The Architectural shader has different settings to the Uber shader and its default is high glossiness, so you need to change that anyway, like you would with most shaders. Mec4D did some tests and found an interior using Architectural sampling finished a simple room in 8600 iterations, while the default settings got to 10,000 iterations but was only at 86% when she halted the render.
👍: 0 ⏩: 0
SickleYield In reply to Tarbicus [2016-10-13 02:07:24 +0000 UTC]
I tried the architectural sampler with the 10 mesh lights scene (previously 40 sec) and got 2 min 15 sec; I tried it with the 7 photometric scene (previously 50 sec) and got.. 51 sec. Huh. Oddly it didn't slow down the photometric scene much, but in no cases did it make things faster or more attractive (the scenes look the same as well).
I may have been confused about what you said earlier. What is an "architectural shader" in this context? I thought you were just misstyping "shader" for "sampler."
EDIT COMMENT: I'm very tired and it appears I actually forgot to turn on the sampler for the 7 photometric scene. With that actually on, the time was 2 min 1 sec (more than double the no-sampler time). I'm looking at the two renders side by side and see no difference in level of noise or fireflies.
👍: 0 ⏩: 1
Tarbicus In reply to SickleYield [2016-10-13 10:18:01 +0000 UTC]
Righto. Maybe I should give it a quick whirl when I get a chance.
👍: 0 ⏩: 0
Tarbicus In reply to SickleYield [2016-10-12 22:18:14 +0000 UTC]
And caustics were slow in the earlier days of iray, but I've had some really good results lately in enclosed spaces, even with SSS figures in the room (caustics are utterly awesome for frosted glass and the like). I'm pretty convinced the renders were a tad faster in some cases. it might even be that the Archi shader seems slow to begin with but doesn't produce as many fireflies, so the render is actually faster in the long run. I don't know, but I think you should give it a whirl.
👍: 0 ⏩: 0
fallenboy84 [2016-10-12 20:36:36 +0000 UTC]
Iray is a complete mystery to me, I have no idea how people get such amazing results out of it because I follow the instructions and get random results if it even finishes rendering at all.
👍: 0 ⏩: 0
DonKevinMartin [2016-10-12 16:25:10 +0000 UTC]
Thanks for sharing your test results! Iray remains such a mystery!
👍: 0 ⏩: 0
vwrangler [2016-10-12 05:52:23 +0000 UTC]
So basically, you want to put seven or so LOW level photometrics in a scene?
That's ... confusing.
👍: 0 ⏩: 1
SickleYield In reply to vwrangler [2016-10-12 05:55:39 +0000 UTC]
I tried it with an i13 scene that was more open and it behaved more characteristically. Now I'm thinking it's a result of the enclosed geometry. Tomorrow I will try it with Barefoot Dancer (more open) and see if it's that. Iray doesn't like a mostly-enclosed scene.
👍: 0 ⏩: 1
vwrangler In reply to SickleYield [2016-10-12 06:30:15 +0000 UTC]
Ah, OK, that makes more sense. Reality/Lux does the same thing; more open spaces render faster -- sometimes MUCH faster.
👍: 0 ⏩: 0
LmAnt [2016-10-12 05:41:20 +0000 UTC]
hehe... now I am back at the confused-status again
Thank you anyway for sharing. It's kind of a relieve that not only the nooby-me but also the experts are surprised by what this render-engine does.
However, and I am NO expert at all when it comes to this render/light stuff! - but could it be possible that the different opinions people have about the influence of light on the rendertime is also influenced by the configuration of their machines?
And maybe by the complexity of the scenes?
👍: 0 ⏩: 1
SickleYield In reply to LmAnt [2016-10-12 05:53:01 +0000 UTC]
I tried it with the scene from i13's Eclectic Style and it behaved more conventionally. Tomorrow I may retry it with a different scene geometry - I'm wondering if the level of enclosure is making it behave uncharacteristically.
👍: 0 ⏩: 1
LmAnt In reply to SickleYield [2016-10-12 05:58:27 +0000 UTC]
I am not sure what "level of enclosures" means (I have two issues to live with: The language and the technique ) ..
however, I think it would be a cool thing if different people with differently configurated machines would do the same tests like you.
I would happily volunteer for that, cause I am a big fan of stuff like that, but my machine is at the lowest imaginable end of the performance scale, so it would take me days to make the test.
👍: 0 ⏩: 1
SickleYield In reply to LmAnt [2016-10-12 06:02:21 +0000 UTC]
Enclosures = more walls around the scene.
3delight did not care if you were rendering under a skydome or indoors.
Iray does not like fully enclosed rooms and it HATES non-HDR physical skydomes. Render times get bigger and bigger the more your camera is inside a closed box.
The ideal Iray scene is a room with only one or two walls. If at all possible you want to take the roof away and as many walls as you can, and depend on your HDR for reflections in surfaces. NEVER use a skydome, instead use the HDR from the Environment tab for fill lighting and backdrops.
👍: 0 ⏩: 1
LmAnt In reply to SickleYield [2016-10-12 09:58:59 +0000 UTC]
Oh..
and Oh!
..and Oh!
Lots of news for me!
And lots to consider before setting up my next scene!
Thank you very much for that!
👍: 0 ⏩: 0