Performance between VRayLight and VRayLightMaterial...

[cgrant3d]

Hey everyone, I've been wondering lately what the performance difference / quality tradeoffs exist between the actual VRay light object and the VRayLightMaterial ... So, I did a simple test whose results I find interesting...



This scene is exactly the same except the one on the left uses a VRayLightMtl on a geosphere of radius 1' and the one on the right uses a VRay Light set to sphere with the same radius. The multipliers were identical in the two images.

Here are the samples for the two images



Given the same settings I find it interesting the version using the VRayLightMtl is brighter (if it's the same surface area emitting the same multiplier light shouldn't it be the same?), has less noise (because it uses way more samples) and it rendered faster than version using a VRay Light...

Any thoughts on these results? Does the LightMtl maybe use a different falloff than the Light object does? This whole question came to me as I was trying to decide the optimum way to render recessed can lighting... Given the LightMtl is using more samples I would imagine it would not scale as well on a large scale (ie if there was a bunch of glossy reflections , displacement, etc. those extra samples would most likely make the vray light version faster? ... Maybe I'll do some more tests...

[percydaman]

im not seeing any of your images.

edit: never mind, now i see them.

[trixian]

Well, what popped into my mind first is that your test doesn't really do justice to the vray light. If you were to try to light a model with lots of fine detail, you would find that the vraymtl would not give enough fidelity, while the vray light would have no problem. The reason is the vraymtl uses the gi settings of your primary and secondary bounces, thus very optimised and for this one instance, faster. More complex materials/shaders would maybe give the imap calculation a harder time. The only issue one has to bother about with the vraylight is grain, so lighting a large, featureless room like that is bound to be fast when using the imap. Coincidentally, you could possibly achieve the same result by checking the "store with irradiance map" button on the vray light.

[cgrant3d]

...The reason is the vraymtl uses the gi settings of your primary and secondary bounces, thus very optimised and for this one instance, faster. More complex materials/shaders would maybe give the imap calculation a harder time. The only issue one has to bother about with the vraylight is grain, so lighting a large, featureless room like that is bound to be fast when using the imap. ...

Ahh... Makes sense!

[flipside]

Re: Performance between VRayLight and VRayLightMaterial...

Given the same settings I find it interesting the version using the VRayLightMtl is brighter (if it's the same surface area emitting the same multiplier light shouldn't it be the same?), has less noise (because it uses way more samples) and it rendered faster than version using a VRay Light...

The lightmtl image is less noisy not because it uses more samples. The increase of samples you see is because in that image, all light is GI light. While in the vraylight image, light starts with direct light of your vraylight, and only after that, IR map comes into play. So there is less GI to be calculated in your vraylight image. (GI only starts here where in the lightmtl image second bounce started)

This could also explain why it is darker. What are your first and second bounce GI multipliers? For this test, you should set them both to 1.0. You should also use lightcache for secondary bounces, because this computes infinite bounces, while qmc GI stops a defined number.

The noise is coming from the area shadows of the vraylight, increase subdivs to get rid of it.

Like trixian said, it depends on your scene and the detail you want. If you don't need very small detail shadows, go for the vraylightmtl. But that won't be able to create very sharp GI shadows, so if you want that, use the vraylight that casts direct light and area shadows.

It would indeed be the same if you turn on 'store with IR map' for the vraylight. This treats your lightsource as first bounce GI, just like with the vraylightmtl.

What also could make a difference is how detailed your geosphere is.