Pls bring back the min shading rate (as it is still in 3dsmax too)

[stefan_laub1]

As the title said, i really do miss it. for our scenes(archviz) one could greatly speed up scenes with getting half render time in using it(set to higher value to boost gi vs cam rays), even corona has this (GI vs AA balance setting)
it would be great for pro users and really not hurt the GUI. more and more pro users seem to use C4D,

thanks!
Stefan

[viktor_angelov]

Hello stefan_laub1 sirio76 akos_kiss Thank you for your request! The Min Shading Rate feature was hidden in V-Ray for Cinema 4D and others, and as you mentioned is expected to be hidden from V-Ray for 3Ds Max, too. It's not expected to be brought back.

[sirio76]

Yes it would be very useful.

[akos_kiss]

This is a plus one for me also. These more granular settings are very useful for optimizing scenes and materials for fast and efficient rendering. It's not just great for archviz but for us other fields too. (we mostly do product and motion design).
I can remember a post from someone that, chaos will phase out these settings from 3ds max also, but i couldn't find it right now.
If there will be some other new way of optimizing the render speeds i'm all for it but, atm the shading rate overrides are the easiest way to go.

[viktor_angelov]

Hello stefan_laub1 sirio76 akos_kiss Thank you for your request! The Min Shading Rate feature was hidden in V-Ray for Cinema 4D and others, and as you mentioned is expected to be hidden from V-Ray for 3Ds Max, too. It's not expected to be brought back.

[sirio76]

Is there a specific reason why it will not be brought back? There are several scenarios where msr is extremely helpful

[viktor_angelov]

sirio76 Min Shading Rate is removed from the interface and always uses the default value of 6. This is the optimal value for V-Ray and changing it usually comes at the expense of performance.

[sirio76]

sirio76 Min Shading Rate is removed from the interface and always uses the default value of 6. This is the optimal value for V-Ray and changing it usually comes at the expense of performance.

Thanks Viktor, yes 6 works fine for an average scene, but other scenes may require quite different value and the speedup can be significant
IMO unless there are technical reasons why it can't be added easily, it would be nice to get it back since is a simple parameter and will not clog the interface, less experienced users can simply leave it at the default value.

[123456_]

Having more debuggable options is an advantage

[stefan_laub1]

thanks for the answer, as the others say, this was a very useful option, i am able to speed up scenes to 200% speed in setting min shading from 6 to like 12 or 16. a trained eye knows on which scenes that works and on which not.

i can only imagine a better tool than that replacing min shading that removing it totally: my idea would be some kind of trained AI based auto min shading rate (GI vs AA balance), that sets the sample balance based on t e gathered data it find in LC pass.
so of it learns in LC that the scene benefits from higher GI samples, it sets "min shading rate" to a higher value, if it sees most is done via cam samples, it sets it lower

just an idea, maybe you work on such a thing already

best
Stefan

that said, a good control never hurts i believe, pro users love that. the hype that everything needs to be simple might be over again ( see cool SD/CN based AI tools with trillion of settings, more than v-ray 1.09)

[sirio76]

Yes an AI solution or something like that may help to replace msr, I’m not sure if light cache can provide all the information needed though, probably it would be better to “learn” from full render tests since a complete render process can provide all the data needed to optimize the scene.
In the mean while msr will do the job

[stefan_laub1]

well lightcache as far i know can in fact get a lot of scene data and the detail and structure of it. i think adaptive lights uses this data for example very successful.

the AI training itself of course is of ready rendered scenes and the resulting speed gains with diff sample balances.
but the actual scene detection needs to be prior of render, otherwise it makes no sense as you needed to test render each scene before this can chim in, there i could imagine the LC could give enough on the scene structure as base info for the AI.
In end only Chaos/Vlado can answer such things