The normal map represents a range from 0.0 - 1.0. So when you use a normal map set to a RAW/Linear color space it behaves the same as a glossiness or roughness map. Why does VRay default the map to 30%? Does the map represent a normal map intensity of 1.0 when set to 30% or should you pump it to 100% to represent a true 1:100 ratio instead of a 1:33.3 ratio?
Tried the Google-Fu and the forum a but couldn’t cut through the noise of using such generic language.
Awesome question, though I don’t have an answer. I’ve always wondered about this myself.
It’s lost in the way Max did bump originally (taking it exactly from 3ds DOS. it too had it set by default to 30.).
To this day, if one just adds a map to bump, with max, one has to pay special attention to keeping that influence value quite low, lest the look worsens.
To get around it, if at some speed cost (for good quality reasons, however.), one should add a colorToBump and set the height correctly in world units, and then the influence of the bump map can (and should.) be set to 100.0.
So in a PBR workflow, the idea is to not touch the material but to have the textures convey all the information, including the normal map. When Chaos enabled the PBR workflow into the V-Ray Mtl, is it a 1:1 representation of what you use or is it a 1:0.33 represtation? I think the C2B would be ok as a workaround when using random textures from around the web but there should be some documentation of normal map interpretation in a PBR workflow. Everything from Substance documentation show them exporting from a V-Ray preset to a glossy workflow instead of showing a proper PBR export. Then on the Chaos side, I’ve yet to find a PBR set up guide that says what to set the bump map to. Since the default is set to 30%, it’s not helpful when interpreting metal/roughness textures from other sources or a standard set of textures. Since PBR assumes a 1:1 ratio of raw data, it should be 100% for a bump map. I’m hoping somebody can give a definitive spec number for PBR maps.
The standard in Redshift or Arnold would be a linear representation. Is V-Ray also a linear representation or is it delta of 3 or something?
Aah I didn’t know that you had to put it at 100 for it to work correctly with world units, that explains a lot. Might want to add that information to the docs 
Does C2B interpret the normal map and convert it to a bump map( black and white 0.0 - 1.0) or does it pass the color channels into the right coordinates? If it does reduce it to the one map then an 8bit file would be 1/3 the amount of data representing your normal right? It would be 1 channel at 0 - 255 vs. 3 channels at 0 - 255.
Second question would then be. What does the real world scale represent? The Height in the Z distance (0.1cm) from the material or just the World Scale represented in 3DSMax(1.0cm = 1.0cm vs 1.0cm = 1.0m)?
This all gets into personal interpretations vs. the point of a PBR set up. Nobody has gone a measured the height of a leather bump from the base surface of the deepest part of the crack in the surface.
So back to the original question.
Does VRay interpret normal maps 1:3.33 or in 1:1?
Does 30% represent 100% of a normal or does 30% represent 30%?
Would you simply share a scene that you think isn’t working as it should, and perhaps also a scene with Arnold, showing how it should be working?
It’s got nothing to do with a third of the channels, unfortunately: it’s 30 but not bound upwards to 100, the range is -1000.0 / 1000.0
It’s something that worked back in 1995, and was dragged along.
I’ll throw something together and load it in a bit. Thanks!