Progressive Rendering - Subdivs - Explanation in a Nutshell?

[alexyork]

Hi all,

Can anyone offer a simple breakdown of how the mix and max subdivs work in Progressive mode? Is there a general rule of thumb to increase quality at cost of speed? The way it appears to be is that the longer you let it cook the more the noise is reduced/resolved in the same way that Maxwell works etc.

And Ray Bundle size? Does this have any effect on render time not related to reducing stress over the network when using DBR?

Many thanks,

[Svetlozar Draganov]

Min and Max subdivisions in Progressive mode works the same way as in Adaptive DMC mode. Vray starts shooting rays from the camera on passes at the end of each pass the values are averaged and compared with Noise Threshold value, if the averaged result is lower than Noise Threshold Vray stops shooting rays for that pixel if not another pass is fired.
In order to get a better quality Noise Threshold value should be decreased and Max Subdivs should be increased, this of course will affect also render-times , better quality usually means longer render times. Progressive rendering in Vray is not the same as in Maxwell - in Vray there is a Min Subdivision value which ensures that the rendering for a current pixels stops if the specified quality is reached, by doing so Vray is able to concentrate cpu-power over the problematic/difficult area instead waisting time on area which already looks good enough.

Ray Bundler Size could have a very big impact in render times especially in network rendering. When rendering with Progressive Sampler in DR Vray transfer a huge information over the network which might slow down the rendering quite a bit. In that case it is recommended to increase Ray Bundle Size to a higher value - this will reduce the network usage and will improve the performance of the rendering although the refreshing of the image will not be so frequent.

[alexyork]

Min and Max subdivisions in Progressive mode works the same way as in Adaptive DMC mode. Vray starts shooting rays from the camera on passes at the end of each pass the values are averaged and compared with Noise Threshold value, if the averaged result is lower than Noise Threshold Vray stops shooting rays for that pixel if not another pass is fired.
In order to get a better quality Noise Threshold value should be decreased and Max Subdivs should be increased, this of course will affect also render-times , better quality usually means longer render times. Progressive rendering in Vray is not the same as in Maxwell - in Vray there is a Min Subdivision value which ensures that the rendering for a current pixels stops if the specified quality is reached, by doing so Vray is able to concentrate cpu-power over the problematic/difficult area instead waisting time on area which already looks good enough.

Ray Bundler Size could have a very big impact in render times especially in network rendering. When rendering with Progressive Sampler in DR Vray transfer a huge information over the network which might slow down the rendering quite a bit. In that case it is recommended to increase Ray Bundle Size to a higher value - this will reduce the network usage and will improve the performance of the rendering although the refreshing of the image will not be so frequent.

Thank you for this explanation!

Queries:

- With bundle size - if you are only rendering on 1 x machine (localhost), is it recommended to keep this value low, then? How low? Default is 128 I think. Is it best to leave this at default for single machine renders? Or 64, 32, 16 etc.?

- Mix Subdivs - default is 2 - what kind of values could we be looking at here safely? 4, 8, 16? Or should it increase more slowly, to 3, 4, 5 etc.?

The best way is to experiment with test scenes, I know, but I don't have any time for it at the moment unfortunately!

Thanks again.