Steps to reproduce: 1) Have dual monitors 2) Enable PP_OVERDRIVE_MASK by flipping the appropriate bit in amdgpu.ppfeaturemask. (For my configuration I used: 0xfffd7fff) 3) See artifacting. When using default featuremask, or when using a single monitor with the overdrive bit enabled behaves correctly. I heard multiple people who have the same issue, so it does not seem related to my setup. Setup: RX480 Linux 4.20-rc5 If additional information is needed please let me know.
I've just recently chatted to boukehaarsma23 about it, as it also affects me. I have uploaded a video of what this looks like which you can see here: https://www.youtube.com/watch?v=roHXWtHRhyQ Interestingly the artifact lines appear to be aligned across the screens, vsync? The main reason though I first came across this is by trying to manually downclock my GPU. With 2x 60Hz HDMI 1080p monitors attached the GPU is constantly in a higher power state (both mclkc and sclk) and wastes a considerable amount of power at idle. This is not a problem with 1 monitor attached. Also of note is that even with 2 monitors attached this does not occur on Windows where the GPU idles at 300MCLK/300SCLK and is significantly cooler. For comparison about power, according to my UPS my system idles at 112-128W when the GPU's are downclocking appropriately(1 monitor). This measurement matches with Windows where the GPU is downclocking to state 0 appropriately. When it's not downclocking appropriately (2x monitors) it idles at anywhere between 160-176Watt. Setup: 2x RX580 Gigabyte 8G (Monitors both on GPU0) Linux 4.19.4 (Arch Main) Latest stable KDE Plasma on X
Likely related https://bugs.freedesktop.org/show_bug.cgi?id=108647 https://bugs.freedesktop.org/show_bug.cgi?id=102646
Patch https://patchwork.freedesktop.org/patch/314322/ fixes this issue for me.
-- GitLab Migration Automatic Message -- This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/628.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.