Decrease SSS noise?

[bentway23]

This is a very stupid question--how can I get rid of my SSS noise (in any circumstance, CPU renderer, mainly for the skin material but this applies to all I've tried)? I know other renderers you can dial in SSS samples but I haven't found that option in VRay (probably looking right past it--I also can't find the milk in the fridge). I also haven't found a way to increase light samples. The only way I've found is to jack the min subdivs in the sampler section WAY up--taking it from 1 to 20 got rid of a lot of the noise (but it's still quite noisy) while increasing the render time of my test area from 15 seconds to more than 2 minutes. This is just a dumb simple test scene with a slab of skin and a light rectangle underneath.

I know this is very remedial--to date I've just sort of hacked through things and buried them under film grain, but it's time to put on my big boy rendering pants.

[Dubbie99]

having this same problem right now.

Nothing seems to reduce the SSS noise.

Using the ALShader right now.

[david.anastacio]

Muhammed_Hamed sounds familiar this topic

Dubbie99 in V-Ray 5 you can still find the SSS Subdivs in the shader by they are hidden, but they appear not to do anything anymore. The only way to clean is to throw more AA samples at it.
You can although unhide the Min. Shading Rate parameter from V-Ray 4,(ROP -> Edit Parameter Interface.... -> show hidden -> unhide the parameter), this was a great way to clean smooth surfaces and speed up your renders previously (The Chaos team is gonna bring this parameter back after some talks).

Here's a comparission using default settings in the ROP, and only changing the Min. Shading Rate.
The shader has a map in the diffuse and the SSS Density scale was increased to 100, the reflection has a high roughness (0.9). Motion blur in ON and the camera is rotation arount the model 360º over 100 frames.



200% zoom to the head region. LEft side uses 24 Min. Shading Rate, right side uses the default 6. There are more examples where the difference can be even more pronounced.



The noise clears away a bit, at a 100% this should be enough. The best part is the render time gets lower (if you are not dealing with a lot of motion blur).

Min. Shading rate 6: 46.7s
Min. Shading rate 12: 37.5s
Min. Shading rate 24: 37.3s

With more motion blur and lower roughness the render time change, where now the increassed shading rate is slightly slower than the default 6, but it's still better quality overall the render. To get the same render quality using a Min. Shading Rate of 6 we need to use higher max AA samples or a lower Noise Threshold, that will cost us much more time. In scenes that have noise mostly in areas of secondary rays, this a great way to dirrect more samples at secondary rays.


Regarding these changes in the overall sampling, making the default settings the best they can is great, the more sceneraios the default settings can emcompass the better. But removing features that help more advanced users to fine tune their renders that's not so good. Throwing more AA samples at problems is not the best and most efficient way usualy, having per-shader or at least an overall shading multiplier would be great for cases like this.

Best,
David.

[bentway23]

This is great (using the Vray 4 hidden parameter trick). My quick tests aren't that interesting but suffice it to say I had to jack up samples and render time considerably to get comparable results to making the minimum shading samples 128. Thanks for the tip!

[Dubbie99]

Thanks David! Very helpful. I didn't think of checking for hidden params.

[GoodbyeKansas]

Would be great with some speedups of the new SSS in VRay material. It's very slow for us, and even slower with motion blur

-- Erik

[david.anastacio]

Would be great with some speedups of the new SSS in VRay material. It's very slow for us, and even slower with motion blur

-- Erik

I also bumped into the same, the render speed is not so good. I also had different sampling when rendering through the standalone copared to the UI render which was very odd.
Some more issue that I reported are: the SSS scale parameter is inverted, the inputs for scale and amount are missing and you have to increase the IPR GI bouces to 16 to get the correct look, all these are reported and they are working on it.

Best,
David.

[GoodbyeKansas]

david.anastacio Thanks for the update. Then it's all already reported to support and no need for me to do it again.

-- Erik