maya shaders vs mr shaders

[bgawboy]

What is the context of your testing?
In other words, what kind of lights? Falloff settings? Are they area lights?
FG only, no GI?
Are you using objects as virtual lights, like HDRI skydomes or IBL (environment shader)?


I am a firm believer of setting up shaders in sets of Kits, where the values that are transported around the scene are well matched with each other. This could be aimed at being physically accurate, or just aimed at being artistically intuitive. It also has to make some assumptions on how you use and set up the options of FG, GI, shadows, etc.

In that respect, the Maya shaders are meant to work with each other. Most of basic shaders are designed to work as a set of material/surface shader, shadow shader and photon shader.

This means that one has to be careful, for example, if you start with a Maya Blinn and then override the material/surface shader with say a DGS. There is still a lingering shadow and photon shader in there. The way to prove this is to export to an mi file and look at the material(Shading Group).

Now, the methodology using DGS with GI and FG is that you shouldn't use shadow shaders. The same effect is handled by refracting photons, or refracting secondary FG rays. Also, the photon shader should match the material/surface shader so that the transparency and D, G, and S coefficients match.

Now if you assign a DGS shader to an object directly, it will create a 'clean' material (Shading Group) without shadow and photon shader, but you may also have to remember to hook up that same DGS material shader to the photon slot of the material (in the mental ray tab of the Shading Group). (The DGS doubles as a material/surface shader and a photon shader, so that you can reuse it exactly for this purpose.)

The Maya base shaders tend to cover a lot of functionality. They are what we call monolithic shaders, rather than component shaders which are typically combined into shader networks and made into Phenomena. Many folks tend to like the way these shaders present themselves in the UI, as it makes for a shorter learning curve when first using them.

It does however put a lot of potentially unused functionality into the shading calculation, and this is why some people find smaller component shaders faster for how they want certain materials to look. Once you start making bigger shader networks, the Maya shaders should come back to show better performance.

Also, in monolithic shaders, it may take some evolution to optimize for typical use cases so that the performance can match simple shaders.

[Chadrik]

well put, bart.

i've always thought of the mr shaders as example shaders for shader writers. in my opinion, the shaders are not very artist friendly, and reveal a very programmer oriented design. just take a look at their basic spotlight shader -- to set the cone angle you must input the cosine of 1/2 of angle. not terribly intuitive. the value is input in this way so that the shader can avoid having to recompute this figure for every call, making it slightly faster, and infinitely more difficult to use. if you are a single artist striving for fast, photorealistic results, the mr shaders could be the solution for you, but it is difficult to build a pipeline for a large production around them because they are so unfriendly. and as bart points out, by the time you've constructed a phenomenon sophisticated enough to be artist-friendly then your speed margin has disappeared. i'm very curious if many studios are attempting to leave behind the maya shaders in favor of all MR shaders and custom shaders -- it would be a monumental feat to eliminate the maya shaders entirely.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.