I was wondering if there were any performance difference between either method. I'm going to be using the Oculus Quest 2 to run the project I'm working on, but there doesn't seem to be any straightforward answer on whether you should be sticking with a PBR workflow or take a more traditional route.
Is there any documentation that goes further in depth into the actual use of individual textures in VR? I seem to be having a hard time finding more about that other than Oculus's vague "art direction" page.
I would say that it's more or less same. This textures has same bit rate, but when you pack them into aRM or SG.. texture you will have better result with RM pipeline, cos of channel blending with dxt compression. That's true for ue4, but can be different in other engines.
you need less total texture channels to get the same information if you use metallic/roughness - even if you add a specific dielectric reflectance/specular occlusion map
you need 10 for spec/gloss
you need 8 for standard metal/rough or 9 if you add a specular reflectance map
Both are using PBR models. I don't know the calculation specificities but i'd guess they are pretty similar in cost.
Lighter shading models would be Lambert, Blinn...or unlit, with/without baked lighting.
You have to perform tests, it all depends of your scene structure, size, assets, polygons...
All other things being equal there's basically no difference in terms of how the information is unpacked and passed to the clever bits.
You can pack metal/rough into less textures and save yourself the cost of a sample but the actual calculations are basically identical