Photometrics Do Slow Down Render
SickleYield
Posts: 7,644
I've been experimenting with taking various different precreated scenes, deleting all their lights so they use only the default HDR, then gradually re-adding lights and keeping track of time while I fiddle with their settings. My results were somewhat unexpected.
1. Adding more photometrics to a scene does not make it faster than just an HDR provided the HDR is reasonably bright.
2. Adding more photometrics to a scene doesn't slow it down a lot provided they are all point geometry.
3. Adding more photometrics to a scene definitely slows it down if they are any other geometry type, including disc, rectangle and sphere. If these geometries are used, the scene gets slower with every photometric that is added even if only by a couple of seconds.
4. Adding more photometrics to a scene is still faster than adding an equal number of mesh lights.
5. Photometrics are still recognized even when you add a very large number; there may be an upper limit but it is higher than 50.
6. Exception: Adding point photometric lights to a very dark scene makes it go faster compared to just an HDR; the only time when HDR-only is not faster is when a scene is very dark.
So it appears that what I've been told, and what I have regrettably also repeated, is incorrect. The idea that a brighter-lit scene renders faster in Iray, and more photometrics actually make a scene speed up, appears to be repeatably wrong.
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
Thanks for the testing/info, and for the info in the Iray lights thread in the commons. I suspected that was the case, because it seems that every light added would increase the calculations required.
I wonder though if some of the chatter about this isn't referring to the speed with which the render "clears up", since a poorly lit scene will take a long time for the render/image to resolve, and a well lit scene will finish faster?
Brighter = faster...yes.
More light (not necessarily brighter) = faster...yes.
More lights = faster...no. More physical lamps (whether they are photometric or mesh) will eventually, depending on the number of them, slow down a render.
As to a point with geometry...generally the point with geometry is faster than a similar mesh light, except for single poly rectangles. The least number of polys you can make a sphere is 8. A spherical point is somewhat faster than that...not a whole lot, but some. When you start getting into more 'spherical' mesh spheres the difference grows.
HDRI trumps all.
Thanks for these results.
Is it possible to quantify your render times? Are point lights say, 10% faster than an equal number of mesh lights? As a professional photographer I've long viewed any point light source as the ultimate evil and used soft boxes or reflectors almost exclusively.
Not necesarily, speed depends on several factors, you can have a scene with a mesh light of high intensity put at 1 mile away, It'd render quick and you'll get a dark render. Start getting closer to the light source and start placing objects (liquids or crystals) between the camera and the light and you'll experience a slow down in render time.
IRay works by tracing rays of light which bounce in the scene affected by the object materials in the scene (laws of physics). A mirror surface requires the few calculations, a simple Lambert's law calculation, a dispersive material like water or crystal requires far more complex math.
More light means more rays traced by IRay.
Very early on when 4.8 was still in beta someone at Daz mentioned that (to the effect ) "more lights makes it render faster." This is only true in the sense that the more path tracings there are, the more samples there will be to any given pixel, and therefore the faster the pixel convergence -- pixel convergence is a "the more the better" sort of thing. But there is a tradeoff in that additional light paths must be calculated for each light source. As stated by others above, it is more accurate to say that the more *light* there is, the faster the render. (This is a generality, as there are many types and materials that faster render ot slower, given the specifics. Materials with a high frequency noise will tend to take longer to reach convergence. In these cases, pixel convergence is harder to estimate, because nearby pixels have such a high variation from one another.)
As with emissive sources, rendering takes longer when using a geometry with a photometric source, especially if that geometry has curves. Point emitters have an immeasurably small irradiance profile. Unfortunately, point emitters tend to provide the least attractive shadows, to the extent that they often make the scene look unrealistic; so we allow some increase in render time in order to achieve a better result.
Along with understanding how path samples contribute to pixel convergence, it's important to remember that the tone mapping settings don't have much effect in the outcome. A dark scene brightened by adjusting tone mapping doesn't provide much of a benefit. It's better to light to a workable brightness, and then adjust tone mapping for the desired appearance and mood. Iray uses the tone mapping settings to internally set a nominal luminance level, which it uses as part of its path tracing and pixel conversion estimates.
More light will not always help if it's localised or strongly directional, leaving areas that are lit only occasionally by a bounced ray and that therefore take a long time to converge. A well-lit, evenly lit scene tends to converge quite quickly - but also tends to be less interesting to look at. As with many things, there's a trade-off.