Page MenuHome

Progressive Multi-jitter is causing brown spots to appear in this .blend
Closed, ResolvedPublicBUG

Description

System Information
Operating system: Windows-10-10.0.16299-SP0 64 Bits
Graphics card: GeForce GTX 1080 Ti/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 460.89
(This has been tested on multiple other PC with different builds of Windows 10, with/without nVidia GPUs, and generations of Intel chips more than ten years wide. No tests on AMD/ATI.)

Blender Version
Broken: version: 2.93.0 Alpha, branch: master, commit date: 2021-01-21 00:12, hash: rB41a945d7469f
Worked: Unsure.

Short description of error
There are brown pixels in this image. (I've circled them in the attached screenshot.) No denoiser will remove them, but changing the sampler from PMJ to CMJ or Sobol fixes it immediately.

Exact steps for others to reproduce the error
Open the attached .blend.
Add an HDRI to the node in the World thingamajig. Any work, but the problematic effect is extremely subtle and easily mistaken for normal noise unless you use the particular one from https://hdrihaven.com/hdri/?h=abandoned_slipway .. 1k works fine and makes the pixels even more obvious, but it happens with any resolution provided from the site.
Press F12.

Extra stuff
Notice that if you change the seed, you still get dots, just in different places. If you change the sampler, the dots go away entirely. Also notice, with PMJ set, how no amount of de-noising will ever remove the brown dots, and even other de-noisers fail to remove them; Blender seems convinced that they're supposed to be there.

Changing the resolution just makes more dots. Changing CPU/GPU doesn't affect the dots. Rotating the empty that the camera is parented with maintains the dots, but in different positions.

Event Timeline

Philipp Oeser (lichtwerk) changed the task status from Needs Triage to Confirmed.Jan 22 2021, 1:51 PM

I can confirm.

Not sure if this is due to not being used alongside Adaptive Sampling [for which it is mentioned to be suited for], but even when enabling Adaptive Sampling the situation does not get better.
@Stefan Werner (swerner): is this something to expect?

Will confirm for now.

Brecht Van Lommel (brecht) changed the subtype of this task from "Report" to "Bug".Jan 27 2021, 5:56 PM
Brecht Van Lommel (brecht) claimed this task.

Appears to be solved in 3.0, the PMJ pattern had significant changes.