GPU CUDA Max Subdivs passes

[dariofx]

Maya 2017 U4 & Vray for Maya 3.6 stable nightly 27496

NVIDIA GP100 (enabled) + Quadro M6000 (not enabled)

Image Format: exr (multichannel)

Production engine: CUDA
GPU Resize Testures: Full size textures
GPU Tex Format: 8-bit

AA Filter type: Lanczos
Size: 2.0

Max Subdivs: 100
Noise threshold: 0.001
Max. render time (min): 0.0
Ray bundle size: 128

GI: On
Primary: Brute force
Secondary: Light Cache

I'm trying to wrap my head around the Max subdivs and how many passes it will do when rendering an image that is 6000 X 6000 pixels and the above GPU CUDA render settings.

Basically my question is, if I'm going to be doing similar renderings at 6kx6k pixel dimensions, with Max subdivs of 100 and a noise threshold of 0.001, about many passes will it do if it uses the full 100?

How is this figured out?

Thank you for your help.

[savage309]

There are three limits for RT CUDA rendering - time, noise and max subdivs. You can use those together and V-Ray rendering will stop whichever comes first.
Time is pretty straight forward - V-Ray will render maximum as much time as you have specified.

Max noise defines a noise level, after which a pixel is stopped being traced. This way V-Ray can focus more time rendering the pixels, which are actually noisy. You put that lower - the image is rendered cleaner (and longer).
Max subdivs defines how many rays top should be traced through each pixel. This value (as all subdiv values) is squared, which means that Max subdivs 100 = maximum 10000 rays traced per pixel (in 3ds Max, this is named Max rays per pixel).

Max noise and max subdivs though are kinda linked - it is not a good idea to put max subdivs to a very big number, assuming that max noise will manage to figure out when all the pixels are clean. Some pixels are harder to guess that they are clean and they might be oversampled a lot. Max subdivs between 50 and 100 should be more than enough for most scense.

It is hard to say how many rays V-Ray will trace for any of those settings, because it depends on the scene. If it is a simple scene, which samples are less noisy, V-Ray will figure that out and will stop when the noise level threshold is reached. For complex scenes, it will take longer.

Best,
Blago.

[dariofx]

Thank you for the explanation Blago!