IRAY Light Speed -- Additive to accelerate?
Scavenger
Posts: 2,674
I'm working on an indoor scene. It had been lit by 2 4 poly emmissive surfaces, a "florecent light" (the top item in the pic).
A polygon on each side, I used the REAL LIGHTS flourencent shader on it.
It went REAL slow... after 12 hours, still 0% convergence.
I decided, since I wasn't in a rush, to experiment some with other light types.
I've made a photometric light tube (the bottom item). It's settings pretty much replicate the florencent testings in a black box test. (it's a point light in column mode that's the length of the column it's sitting in, which is a primative with iray's light frosted glass on it).
I WAS going to replace the two emiisives with two of the tubes..see how it's speed is.
Then I thought, what if I keep both..would the presense of the emissives with the photomotreics speed things, or do they add more complications, so slow things?
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
© 2026 Daz Productions Inc. All Rights Reserved.
Comments
I just ran a quick test...
Point light as a tube: Total Rendering Time: 35.21 seconds
Tube (4 poly) mesh light: Total Rendering Time: 53.17 seconds
Both, together: Total Rendering Time: 1 minutes 18.72 seconds
Also, times were longer before making sure everything in the scene had Iray materials applied.
I tried a siligty different tact (because I decided a different process might be good)...
using the photometric tubes, withougt the glass turned on, and an HDRI...light looks to be about the same in total.. 12 hours later, I'm at 3%
One thing that is clear from all this, I have no idea what I'm actually doing :)
A couple of things...image size matters...huge sizes will take longer, my renders were 512 square. The timings/combinations will scale and hold true. So the general point of the photometric being the shortest will hold true.
The more fixtures/sources the longer it will take. The more light, the faster Iray will converge. Both of those while true aren't quite cut and dry...2x the number of sources doesn't mean twice as long and 2x the amount of light isn't going be 1/2 as long. There are also points where more doesn't do much or will hurt much more than it helps.
another point to consider. if you are using geometry for an emitter , ie a cylinder as a 'flourescent tube' , model the cylinder with as few polys as you can. the denser the mesh, the longer the renderer will take to calculate the lighting. also, use a proper IES setting. To see the actual zone of lighting produced, there is a free 'IES Viewer' available for download.This will actually give you an image that shows exactly the light pattern each IES setting will produce.
Well, this IS huge (3200x2071) --- I render with an eye toward printing... AND it has a number of mirrors in it..I have it limited to 10 bounces..7 left a visible black reflection.
If I recall, the light speeds are Photometric, HDRI, emissive. ?
3Anson: don't understand a bunch of the words you use, but I recognize them, so I'll look that up. I'm trying to use photometrics and hdris. instead of emitter's as much as possible, using emittiers for just things like screens and special effects.
pic at 4%
If that's your full scene (not a crop), you don't need most of what you've described. A HDR light source along with a Photometric spot or two (with geometry) will get the same results.
Closeups with hair will always take a lot longer.
thats a tiny crop of the picture.
19 hours, 9%..think I'll call it good when it hits 10%
Set Iray to "Sun&Sky"
It will still use your scene mesh lights but it will calculate things a little differently and speed up the render meaning more samples in less time.
I dont know what all its doing under the hood differently then the other modes but having used LuxRender in the past I would guess its using a different algorithm to calculate light bounce.
I have an interior scene I use to test new character skins and the room is just a primative cube with a backdrop and a single primative plane for a mesh light. Using the "Scene Only" mode, it take a lot longer to get a decent amont of samples. In Sun&Sky mode I can hit 1000 samples in 2 min or less
Right now it's set to Dome & Scene....woudn't "sun&sky" not use the hdri image or the lights in the scene? (at least, that's what I've assumed)
I dont use HDRI myself. It will use the lights in the scene. My test scene is only lit by a single mesh light.
I have been doing some lighting test in a closed room with several options discussed here. Lights were always set at 6500K temperature and 500000 lumens.
1: Dome and Scene, Architectural OFF, Caustics OFF, Point light square shaped 100x100cm: 17min35secs, 3190 iterations.
2: Sun and Sky, Architectural OFF, Caustics OFF, 100x100 plane primitive set to emissive: 16mis24secs, 3183 iterations.
3: Sun and Sky, Architectural ON, Caustics OFF, 100x100 plane primitive set to emissive: 14min39secs, 935 iterations.
4: Sun and Sky, Architectural ON, Caustics ON, 100x100 plane primitive set to emissive: 16min47secs, 442 iterations.
5: Dome and Scene, Architectural ON, Caustics OFF, Point light cylinder shaped 100x5cm: Stopped it at 60% convergence after more than 30 minutes.
6: Dome and Scene, Architectural ON, Caustics OFF, Point light square shaped 100x100 with IES area light attached: 15min54secs, 927 iterations.
7: Dome and Scene, Architectural ON, Caustics OFF, Point light without geometry (point shaped) with IES area light attached: 4min46secs, 298 iterations.
8: Sun and Sky, Architectural ON, Caustics OFF, 1x1cm cube primitive with IES made emissive with IES area light profile attached: 6min22secs, 364 iterations.
I couldn't get the IES profile to work in a 100x100 primitive plane for a direct comparison to case 6 because the ies was oriented towards an edge of the plane, and turning two sided light ON left a shadow in the middle.
Naturally, some solutions are not usable if the light source is to be seen in the render; but a point light with an IES profile seems the fastest and the source is invisible; so you can place it inside any lamp or whatever. Of course, using the IES profiles does not show the same lighting that the other options (although it shows the same level of overall brightness in the room), but there are plenty to choose from and surely you can find one suitabe to any scene.
Hope this helps!
Edit: Forgot to test! A point light without geometry, without IES profile 28min8secs, 1688 iterations.
Edit2: In all cases except the fifth one, the reached 95% convergence using default render settings.
Rafmer, I'm not sure I follow your post. You have several different scenarios but they all result in different iterations so I don't see how they really correlate. For example Scenearios 2 and 3 are similar except use of architectural sampler and seems to indicate in your test scene that turning architecural sampling on is a bigt time savings. But on the other hand the number of sampes are way fewer. So were you testing to reach a certain level of convergence and some reached it with fewer iterations needed?
In the Create Primitive dialog box, try setting the Primary Axis to Z positive or Z negative.
Lol sorry, forgot to mention it. All scenarios but the fifth one reached 95% convergence.
Nice! Thanks for the tip.