There appears to be a discrepancy between how RT (GPU) renders the refraction of this bottle and how the production renders it (CPU). It’s kind of a problem, cause I work with materials in Active Shade, then render out with Production…and it’s not just a little different, it’s big. [See images below.]
Any ideas?
See my reply in the other thread.
Best regards,
Vlado
Sorry for posting in the Maya forum…my bad.
So, Here’s the story…
The bottle is a spline with a lathe modifier, with a vray mtl. The GPU is rendering what appears to be an accurate refraction. We’ve remodeled each bottle like 3 times, just to make sure the spline is nice and neat. The normals appear to be all facing the right direction (converted it to mesh and had a look). For awhile, it appeared that the model was the problem, as on the latest version everything was working nicely (we seem to have fixed with remodeling). But randomly the refraction went wonkers again, but this time on another bottle model…meaning, it was working fine on that model, and then it didn’t. “it” refers to the CPU rendering (adaptive, bf, light cache).
Your comment about the GPU trace depth isn’t really valid, since it appears to be rendering it just fine in GPU. But it did make me think that it had to do with the refraction depth in the material…screwed around with that to no avail.
IOR also came to mind, but nothing too wonky going on there…tried between 1.5 to about 4 …still no good.
When this good bottle went bad, I had just been swapping out some HDR’s in the refraction/reflection environment slots, and perhaps most importantly in the viewport (physical cam), as well as Max’s environment…it was after clearing those slots that things went south with this good bottle.
Here are some images showing this good-bottle-gone-bad in both GPU and CPU renderings.
Mind you, the reason this is even an issue is because it’s an animation and we need to farm it out for rendering (no GPU).
vray version 3.10.03
Max 2015 SP3
I’d be happy to send the file to you.
If you collapse that bottle to an editable mesh, does that fix the issue? If not, it would be best to send us the scene to support@chaosgroup.com
Best regards,
Vlado
So, the answer to your question above is no, it didn’t help (i had tried this before). But on a related note, we remodeled the spline so that there was an outside surface and an inside surface all in one spline (connected). It worked. Not that there wasn’t an inside and outside surface before, but they weren’t connected. However, why GPU and CPU were rendering the refraction of the previous model so differently (GPU correct, CPU incorrect) is still a mystery.
The problem is back again, so I’m emailing you the file. BTW, I was able to figure out what I was doing when it happened: adjusting the IOR. I’m just not sure what is causing this.
BTW, I updated to 3.2.02…still a problem.
Thanks for the scene, will look at it.
Best regards,
Vlado