0
$\begingroup$

I rendered this basic scene with glass BSDF object, plain texture into Principled BSDF on floor, no skybox, sun light at 9 strength, non-moving camera. On rendered video, you can see that strange horizontal flickering effect on the glass in frames that should be identical. While rendering each frame, those darker horizontal areas appear around 400/1024 samples.

Here is a GIF from the same frame rendered twice (uncompressed, no denoising, all settings identical). I highlighted the different height of where the darker area starts with the red arrow. While carefully confronting, more than one band can be seen. enter image description here

GPU Rendering on RTX A4000, i9-11900k, Blender 3.3.1, Optix, 1024 samples, Win 10 64-bit. Flickering stays both with and without denoising. It also stays both with default and increased light paths parameters. It stays with caustics enabled and totally disabled.

Any suggestions on how to get rid of it? What could I be doing wrong?

$\endgroup$
2
  • $\begingroup$ Hello and welcome. While files, images, and external videos or links may be helpful additions they should not be the only way to obtain information about your issue. Don't make understanding your question rely on downloading a file, watching a video or visiting an external site. Use the builtin tools to upload images or gifs, along with thoroughly explaining the problem in written form so it can be indexed and searched for thus helping future visitors with similar issues. $\endgroup$ Apr 12, 2023 at 13:58
  • $\begingroup$ Ty for your advice. Gonna add a still frame to the question as soon as possible $\endgroup$ Apr 12, 2023 at 14:03

1 Answer 1

0
$\begingroup$

Ok, so using CUDA instead of Optix basically solves the problem. Here is the same frame rendered with CUDA instead (all settings stay the same): CUDA rendering

For some reason Optix has trouble at rendering Glass BSDF; but must be said that apart from the bands, glass reflections rendered by Optix are way more convincing and generally brighter (less light reaches the table past the glass, as you can see the shadow with Optix being darker than CUDA one). I hope Optix technology quickly evolves correcting such issues, since it is already performing better than CUDA in other situations.

I wonder if such Optix issues with glass are already known.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .