While doing some testing I found out photongi to be a little too strong with lighting.
here's sun and sky in luxcore without photongi:
for comparison here's cycles:
now of course it's different and very hard to match them cause sun and sky in cycles works terribly, but overall lighting distribution seems more or less the same, nothing weird is happening...
now here's photongi with default values in blender:
you can see how is the scene flooded with light with objects on the ceiling almost looking like some fake AO is going on.
And here's photongi with photon depth set to 3 (default in blender is 8):
now the lighting is much closer to the non-pgi render while performance boost is still present
So my question is, what is happening with more photongi bounces/depth that gives this strong lighting?
PhotonGI strength
Forum rules
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
Re: PhotonGI strength
I have the feeling it may be more related to the bias of PGI when working with glossy surfaces (try to di the same test with only matte) ... and, you know, next week, when back in town.
Re: PhotonGI strength
I did few tests with production scenes and it does look like a lot of reflections go wrong often when using pgi...
I'll try to make a good example case once you're back.
I'll try to make a good example case once you're back.
Re: PhotonGI strength
Sorry but I don't really trust your current test scene.
Re: PhotonGI strength
So it appears this is happening only on disney shader.
glossy disney I'm not surre though which result is actually correct cause glossy shader has it's own problems (the edge darkening)
there's also some opencl artifacting due to ngon triangles on the wall(CAD tesselated geometry), unrelated to this issue but it would be nice if opencl rendered things the same as cpu so that we don't get nasty surprises on final renders.
glossy disney I'm not surre though which result is actually correct cause glossy shader has it's own problems (the edge darkening)
there's also some opencl artifacting due to ngon triangles on the wall(CAD tesselated geometry), unrelated to this issue but it would be nice if opencl rendered things the same as cpu so that we don't get nasty surprises on final renders.