Splotchy shadows

[Exon]

I have a very simple scene with an extruded spline and a light over it.. I apply a white vray material and add a camera, then turn on GI and put the preset to High..
The shadows underneath are very splotchy, how can I smooth them out?
MSO_Logo.zip

[maxedesa]

Try using Adaptive DMC for your antialiasing. You could also give your light more samples, but the DMC solution looks better.

Also, your dynamic memory limit is very low, add another 0 to that 400.

Caleb

[Svetlozar Draganov]

Hello Exon

Adaptive Subdivision Image Samper doesn't work very well with blurry values like the soft shadows in your scene. In this situation it is better to use Adaptive DMC Image Sampler.
Once you switch to DMC Sampler the result will be much better than Adaptive Subdivision , but the level of noise is still too much.
Now as maxedesa mentioned you have two options to clear the noise - either to use DMC Sampler parameters or to adjust Light Sampling parameters.
Both methods have their pros and coins:

With DMC sampler parameters - you can easily fix the quality of the entire image without wasting time to find the exact reason for the noise and to tweak it's local subdivisions - this could save you a lot of time if you are working with heavy scenes with a lot assets , but the downside is that this will cost more rendertime.

With the other approach - Light Sampling (local subdivisions) - you are able to fix directly the main reason for the noise in the image (in your case - the light), and Vray won't fire unnecessary rays which will increase rendertimes without any visible quality improvement. The coin is that you have to find out which asset is responsible for the noise - which in simple scenes is easy but in more complex scenes with a lot of lights , materials and etc is pretty challenging task.

Here I am attaching two screenshots for each method.
1.Adjust DMC Sampler (3m53s) - 1.zip

2.Adjust Light Subdivisions (46.7s) - 2.zip


You can see that with DMC Sampler (1) render time is much longer than Light Subdivisions (2) which suppose better result , but it is quite opposite - the noise level in (2) is less than (1).
The conclusion is that with (2) we force Vray to shoot more samples directly to the main causer of the noise , with (1) we tell Vray to shoot more rays to the entire image not directly to the problematic area.