Inheritance - split from separate AO discussion

[samuel_the_hamster]

dave got another question when i select "inherit for parent" in an L16 it goes to L8 even tho the primary input is L16 why

[Moderator note : Thread split from separate discussion on AO)

[davescm]

Parent refers to inheriting from the graph settings. You want inherit from input (I can't remember exact wording as I am not at my PC this afternoon).

Dave

[samuel_the_hamster]

no problem. also for being able to see a node info you need to compute thubnail right if no its shows 0 ms and nothing

[samuel_the_hamster]

also you cant change output resolution right base color is stuck in C8 for example

[davescm]

An output node takes its bit depth from the node connected to its input. So to get C16 the preceding node(s) needs to be C16

Dave

[samuel_the_hamster]

do you normally increase the res of the result? like C32 and if u dont why not?

[samuel_the_hamster]

i just set the base material to C32 so some outputs C32

[davescm · Accepted solution]

Whether there is a benefit in increasing a node's bit depth individually (as opposed to a chain of nodes) depends on the content and what the node does. For example a shape node might as well be L8 if it is outputting just black and white, but there is benefit in making a following BlurHQ node L16 or higher as that will make any gradations smoother. 
I tend to set a graph up to output metallic and roughness at L8, BaseColor C8 or C16 depending on content, Height minimum of L16. Normal C8 or C16 depending on content.

Whilst you would think it would be beneficial to work at 8k L32/C32 throughout, in fact that would be very inefficient in terms of calculation time and memory use within the graph and is likely to have little benefit come render time.

There are some good tips here:

https://helpx.adobe.com/substance-3d-designer/best-practices/performance-optimization-guidelines.htm...

Dave

[samuel_the_hamster]

dave, your explanations are the best, tysm

[samuel_the_hamster]

I understand that using 8 or 16 is ok 32 is way too much in majority of cases right 

[davescm]

I try and stick with the lowest bit depth that will give me the result quality I want. That way I am not using unnecessary GPU memory when eventually rendering.

Dave