You are not logged in.
Sir respectfully I don't understand what you're saying,
a) What is gromit?
b) How/where do I use "amdgpu.dcdebugmask=0x10"? Is it possible to undo that at a later stage as well?
c) What is a bisect?
I mean to say that I'd love to help, I just don't know how to.
(I'm still learning Linux)
Offline
a) The user that has posted in this thread providing kernel builds, e.g. https://bbs.archlinux.org/viewtopic.php … 1#p2210441
b) kernel configuration you can apply on your bootloader to alter kernel behaviour. You'd add that line there and it's removable at the same spot
c) the process of trying to find the exact change that caused an issue, starting from a known good and a known bad point, and iterating untill a culprit is found, gromit is doing the heavy lifting here, if you (collectively, it doesn't have to be you individually) give feedback on each kernel he's providing we will eventually land on the exact culprit. For more info on the underlying process: https://wiki.archlinux.org/title/Bisect … s_with_Git -- that's just for interest and can be somewhat involved, if you want to help, test gromits builds and give feedback on whether you still see the issue
FWIW someone on reddit suggested 58a261bfc96763a851cb48b203ed57da37e157b8 not on a system where I can trivially do a build with a revert right now, but it might help speed things up for @gromit
Last edited by V1del (2024-11-26 10:19:34)
Offline
FWIW someone on reddit suggested 58a261bfc96763a851cb48b203ed57da37e157b8 not on a system where I can trivially do a build with a revert right now, but it might help speed things up for @gromit
6.12.1.arch1 with 58a261bfc96763a851cb48b203ed57da37e157b8 reverted which required 23d16ede33a4db4973468bf6652a09da5efd1468 to be reverted first:
linux-6.12.1.arch1-1.1-x86_64.pkg.tar.zst/linux-headers-6.12.1.arch1-1.1-x86_64.pkg.tar.zst
Last edited by loqs (2024-11-26 13:31:30)
Offline
i'm not on the bisset party yet, but can confirm "amdgpu.dcdebugmask=0x10" resolved it!
i could reproduce consistently before, and it is completely gone with this option.
Operating System: Arch Linux
KDE Plasma Version: 6.2.3
KDE Frameworks Version: 6.8.0
Qt Version: 6.8.0
Kernel Version: 6.12.1-arch1-1 (64-bit)
Graphics Platform: Wayland
Processors: 12 × AMD Ryzen 5 PRO 6650U with Radeon Graphics
Memory: 30.7 GiB of RAM
Graphics Processor: AMD Radeon Graphics
glxinfo:
Vendor: AMD (0x1002)
Device: AMD Radeon Graphics (radeonsi, rembrandt, LLVM 18.1.8, DRM 3.59, 6.12.1-arch1-1) (0x1681)
Version: 24.2.7
Accelerated: yes
Video memory: 512MB
Unified memory: no
Preferred profile: core (0x1)
Max core profile version: 4.6
Max compat profile version: 4.6
Max GLES1 profile version: 1.1
Max GLES[23] profile version: 3.2
if others are curious, PSR means Panel Self Refresh. Not sure what it does... more info on https://bbs.archlinux.org/viewtopic.php?id=276352
my logs, before adding the 0x10 option above (after, this line is gone):
> kernel: [drm] PSR support 1, DC PSR ver 0, sink PSR ver 3 DPCD caps 0x30 su_y_granularity 4
Last edited by gcb (2024-11-26 14:18:43)
Offline
It's a power optimisation on laptops, you can basically tell the GPU to only render out updates to content that's actually changing while the rest remains static -- so you're using the GPU less for smaller changes which will conserve power. Disabling it means the GPU will have to repaint the entire screen for everything again, leading to an increase in power usage.
See https://gitlab.freedesktop.org/drm/amd/-/issues/3742 and https://gitlab.freedesktop.org/drm/amd/-/issues/3658 as well, apparently it might help if you disable VRR/switch to 60Hz if you currently have a higher refreshrate.
Last edited by V1del (2024-11-26 15:14:34)
Offline
I am having this issues with slow/glitches also on hyprland artifacts screw going to downgrade see if it fixes it, this is after I did an update that most likely included the kernel
Offline
6.12.1.arch1 with 58a261bfc96763a851cb48b203ed57da37e157b8 reverted which required 23d16ede33a4db4973468bf6652a09da5efd1468 to be reverted first:
linux-6.12.1.arch1-1.1-x86_64.pkg.tar.zst/linux-headers-6.12.1.arch1-1.1-x86_64.pkg.tar.zst
I can confirm this eliminates the artifacts on my Radeon 660M (Ryzen 5 6650U integrated graphics).
Offline
Did someone already find a report for this issue on the lists or in the DRM Gitlab?
Offline
See https://gitlab.freedesktop.org/drm/amd/-/issues/3742 and https://gitlab.freedesktop.org/drm/amd/-/issues/3658 as well, apparently it might help if you disable VRR/switch to 60Hz if you currently have a higher refreshrate.
Chances are it's one (or both) of these two
Offline
I'm also affected.
Linux 6.12
GNOME 47 upon Wayland
ThinkPad X13 with AMD Ryzen 6850U (RDNA2)
Turning PSR OFF with the well known boot option [m]amdgpu.dcdebugmask=0x10[/m] fixes the graphical glitches (similiar picture).
Upstream report about graphic glitches. Everyone else is complaining about frame timing or performance but our apparent issue are graphic glitches.
Last edited by hoschi (2024-11-30 18:41:04)
Offline
@hoschi have you tried with 58a261bfc96763a851cb48b203ed57da37e157b8 reverted or if that does not resolve the issue bisecting between 6.11 and 6.12? Does the patch attached to https://gitlab.freedesktop.org/drm/amd/ … te_2680869 have any effect? Is the issue still present in amdgpu-drm-next for you?
Offline
I too can confirm amdgpu.dcdebugmask=0x10 works
Offline
I had multiple freezes per day with Vega8 IG, also confirming that amdgpu.dcdebugmask=0x10 seems to solve the issue.
Last edited by lpr1 (2024-12-12 13:34:38)
Offline
Please avoid bumping the thread with me too posts.
Offline
Kernel 6.12.4.arch1-1 still has this issue, unfortunately. V1del seems to be correct, my desktop with an RX 6600 XT works just fine but my laptop with a 660M (it reports itself as a 680M) has this issue. My laptop technically has a dGPU but it has no mux chip so it's rendering on the iGPU (not sure if this is relevant information but might as well mention it just in case). 60Hz as mentioned improves the issue but does not solve it, VRR seems to have no effect.
Video of Problem - https://youtu.be/m_s9LB6e76g
Last edited by LumpyArbuckle (2024-12-15 19:56:49)
Offline
Can you try installing the kernel loqs provided so we might have an additional datapoint that that commit is the actual issue?
Offline
It appears loqs's kernel solves the problem.
Offline
Offline