Which Glossy shading model is the fastest?

Title says it all.

As-in GCX or Beckmen, or whatevers.

more like when and why is it appropriate to use any of these ?

happy cl

Sharp should be the fastest, that is why cycles automaticaly uses sharp when the roughness is zero. I guess that beckmann is little bit faster than the GGX and Ashikhmin-Shirley but I really doubt that the difference is big enough to even care.

Yes thats what I think as well: differency is not big enoug to even care. Here I rounded off to the seconds. Multiscatter is the slowest.
When Ashikmin Shirley came out, I remember that a lot of people found Ashikmin Shirley the best shader, but what I see here is that it stands out in being the darkest one.
@RickyBlender , mhh appropiate. I think each time they try to find the best approximation for a glossy shader. I think it’s best to use your eye, and if you have reference, compare. Also you see the difference more the more rough you put.
I also see that GGX Multiscatter is most saturated in colors.
BTW it wonder how I should measure it exactly … until I am happy with the noise? I gave it just 320 samples now, and 2 diffuse, 3 glossy bounces with reflective caustics on.

well would like to know what was the goals to included these different options
right now not certain which one to use and why or when
I know eye balling is always there but you should always think and plan as well as you can before loosing hours of rendering if possible

and wiki does not really help either

I do like that the new one gives a clearer image at high roughness

thanks
happy cl

I think new one is preferable if you are using high roughness otherwise the others should work well
but again still wondering why we have so many !

thanks
happy cl

History reason. First the beckman has been implemented, then came the GGX (which should be better), then there was the Ashikhmin Shirley and now we will have the GGX multiscatter which should be best of all in terms of realism.

Another comparison test that might help. On forums I see that most people stick with GGX ( GGX multiscatter was not there yet). There was one that recommended GGX for metals, and Beckmann for plastics, but I think that is his personal preference. So I stick with GGX or GGX multiscatter as Ace says, it’s most realistic. But I could use ashikmin shirley maybe when I plan to make a nice metallic motorbike. Don’t know, have to try it with a few samples.

would be nice to see this with some architectural light fixture using the new IES patch
using some 4 feet fluo with 3 or 4 lamps inside

the light distribution is not the same at all
so which one is the most realist ?

thanks
happy bl

IES patch… Is that already in the daily builds ? I didn’t read about IES patches.
Most realistic is the latest one. As mentioned in this thread; beckmann came first, then GGX, then ashikmin shirley, then GGX multisc.
Ashikmin Shirley is realistic, but there is energy loss ( darker ) said Ace.

Where is that IES patch?

The IES path is not in master at the moment, but there is a chance of Lukas getting around to it since he’s been focusing on finishing some of his projects.

Also, at least you can count on the fact that the impact Multiscatter GGX has (when it comes to render times) is not near as high as what’s reported in the paper (due to numerous optimizations made by Lukas and Brecht), so we’re technically using a faster version of the shader.

just realize that this SVN is old an does not have the new Gloss option in it
but I already ask Lukas if he could make an update built for win to test the new gloss CGX multi with IES lamps

hope he can make a new built

thanks
happy cl

I usually just stick with GGX, but in some cases I will for simulating brushed but highly polished metal (like chrome tubing) use Beckman Anisoptropic. I don’t know if this is “correct”, but it can give me a significant anisotropic tail (we have no control over GGX tail) stretch without the need for excessive blur on the reflections. The way roughness works for the various types seems to differ, not sure if buggy or just a side effect on how the maths is performed.

I have no idea about the multiscatter thing, haven’t tried it.