Vray 3.4 and 3.5 comparison tests/renders (adaptive lights performance test)

[eligiusz]

Hi,
I've done some tests on an animation camera.

First the 3.4 version, Brute Force and Light cache,render time: 52m


Second the 3.5 version, without any changes in the settings, render time: 57m


Third the 3.5 version with adaptive lights turned on, render time: 32m


Now the Irradience map in multiframe incremental mode with lightcache in flythorugh mode.

GI calculation in 3.4 render times


GI calculation in 3.5 with adaptive lights render times


and the rendered frames, 3.4 render time: 15m27s


3.5 render time: 16m

[eligiusz]

For the BF+LC version here is VrayRawLighting crop with adaptive lights off


and adaptive lights on


For the IRR+LC vesrion herer is VrayRawLighting crop in 3.4


and adaptive lights on in 3.5

[algato]

Hello eligiusz,

can you please open the Maxscript listener and just post the results for

show renderers.current.twoLevel_adaptiveMethod
show renderers.current.progressive_dynNoiseThreshold
show renderers.current.dmc_useLocalSubdivs

Kind regards
algato

[algato]

Hello Vlado,
it would be very helpful to have a script, which shows the current settings and then you could switch to the new or, when rerendering an old Project, the old behaviour.

Kind regards
algato

[vlado]

Well, if you happened to have the two projects in the VFB history you can load and compare V-Ray settings from there.

Best regards,
Vlado

[eligiusz]

1
30
true

and what is the purpose to this ?

im wondering why there is a slight difference in light intensity between 3.4 and 3.5 (interior lights) and also strange noise behaviour with adaptive lights, there is like regional more noise

[kosso_olli]

and also strange noise behaviour with adaptive lights, there is like regional more noise

I am not entirely sure, but the areas with more noise seem to be the black metal parts. The RGB result mostly consists of the reflection or specular component. Thus, the diffuse component which only contributes very little to the RGB result, gets sampled much less.

[eligiusz]

no, its not only black parts, its less visible because there is more sampling in those areas but its definately visible even in the final render,download the renders and compare it in PS for example.

[william.morris]

Hello,

Can I ask you which is your prefered method for animations? Brute Fore + LC or Ir radiance + LC?

thank You

[eligiusz]

Well for me prefered would be BF+LC but I dont have a lot of computing power so I still use IRR+LC. I just started testing this method BF+LC because although it adds to render time, it saves me time in compositing work when there are moving objects in the scene.

I just tested it for an exterior shot and the render times are not that bad if You decide to use denoiser. I have a rendered sequence if You want to take a look at it.

[william.morris]

Yes that would be great to see please. Also to have an indication of your frame times compared to IRR+LC? Clr threshold level?

I am in the same position, not got a lot of processing power but can see huge advantages of using Brute force + LC for animations with moving objects. I hate having to pre-calculate everything using irr+LC. Done a small test but found my render times have doubled in some instances using Brute force.

Thank You!

[eligiusz]

Here is the raw render BF+LC


and final denoised image


Render times for BF+LC are around 28-32 min
Render times for precalculated IRR+LC are around 20-25 min

color threshold is 0,012
noise threshhold 0,12
dome light subdivs 32
I use local subdivs and have subdivs multiplier at 1,4

[^Lele^]

Render times for BF+LC are around 28-32 min
Render times for precalculated IRR+LC are around 20-25 min

plus the irmap precalculation, for that sequence, right?

[eligiusz]

thats right, but with per 30th frame its not that much time. Anyway with denoiser render times are perfectly fine for BF so I will be using it for sure now, at least for exteriors. I think it might be even faster than the IRR rendering in two passes for moving objects.

[^Lele^]

ah-ah, it was 1 in 30, that's what i didn't understand. Thanks (and great to hear you're willing to give it a go!).