Material Selective Dynamic Bucket Splitting?

[zero-13]

Code being like impossible science-magic to me, I feel like this may be an extremely difficult or potential coding nightmare idea or maybe not possible at all, but how feasible would it be to be able to have Dynamic Bucket Splitting initiate upon encountering a specific material?

More often than not refractive materials like frosted glass tend to glom onto the full-size buckets never benefiting from the amazing Dynamic Splitting.

How great would it be to be able to either select or maybe weight by percentage specific scene materials that would "Form Blazing Sword!" and be the catalyst to activate the Dynamic Splitting.

And now that bucket rendering is up and running on GPU perhaps this might somehow apply to the GPU engine as well, not sure though?

[super gnu]

rather than material specific, possibly a rapid prepass in progressive to detect areas needing high sampling, then dynamic bucket size based on that.

[zero-13]

rather than material specific, possibly a rapid prepass in progressive to detect areas needing high sampling, then dynamic bucket size based on that.

Yes indeed, this sounds fantastic!

An automated, intelligently adaptive per frame solution... Yes please!

[Svetlozar Draganov]

This has been a while in our TODO list but it's not yet developed. The idea is possible, we already do something similar for adaptive lights so it theory it's possible to calculate a few passes in advance to determine what part of the image requires more render power and based on that to split buckets more or less.

[zero-13]

So awesome to know it's in the works!

I can't wait to see some initial testing when this gets to that point, I imagine it will be yet another significant speed boost to V-Ray, so great!

[Svetlozar Draganov]

I don't think it would be that significant actually, with current Dynamic Bucket Split and proper bucket size it's very rare to wait for the last bucket lot of time.
If you find yourself often into Last Bucket Stuck behavior it's a sign that the buckets are too large, try to decrease their size a bit and see if that will help.
Please note that too small bucket size also affects render times negatively.

[zero-13]

I don't think it would be that significant actually, with current Dynamic Bucket Split and proper bucket size it's very rare to wait for the last bucket lot of time.
If you find yourself often into Last Bucket Stuck behavior it's a sign that the buckets are too large, try to decrease their size a bit and see if that will help.
Please note that too small bucket size also affects render times negatively.

Thank you, I will definitely try this and test it out!

I got it in my mind to just always leave the buckets at 48 because In the past, prior to dynamic splitting, I did tons of comparative testing of the default 48 with smaller buckets, and larger buckets, and always found that the default 48 yielded the best render times... not surprising because you guys are awesome.

[Svetlozar Draganov]

Default values are not just a random values, every parameter has been tested for years on thousand of scenes from our QA engineers, that why those values works in almost every scene.