Every interaction was painfully slow – even dragging windows around. The only solution was to downgrade them back to the previous versions. AFAICS nothing odd appeared on journalctl.
I am currently using the vulkan renderer, I will try again later with ngl to see if things improve, but I wanted to check if this is a known issue and if there is any workaround.
Thanks for the pointer @glb , seems like it is indeed related (I have a amdgpu as well). At least in my case, it was not caused or worsened by Chromium, even on an empty desktop every operation would lag horribly.
Good to know. Haven’t updated yet so still on 24.3.2. Performance has been stellar.
Edit:
Just updated. Can’t say that I’m affected. Performance seems to be the same, granted I have only tested one game with Proton and general Plasma usage (Wayland). Using a RX 6750XT with amdgpu.
I experienced the same problem: upgrade to mesa 24.3.3 caused extreme slowdown under GNOME. I use a laptop with AMD and (discrete) Nvidia GPUs. In fact, checking with nvidia-smi, switching to 24.3.3 causes all processes to be sent to the discrete GPU as opposed to only a select few, e.g., games.
After checking online, it seems the problem lies with Mutter. The combination of the latest mesa drivers and Mutter may somehow causes Mutter to select the discrete GPU as the primary GPU.
Thankfully, you can tell Mutter the GPU to be used a primary GPU with a udev rule. I followed these instructions and everything seems to work fine so far. Or you can downgrade to older mesa drivers and wait for an update.
Reporting the same thing here. Fedora 41 on Gnome 47. AMD(Ryzen 7) and discrete NVIDIA gpu(NVIDIA GeForce GTX 1650). Tried switching the renderer but nothing seems to work. Ended up downgrading to 24.2.4 for now.
Of course always check in advance if there is already a related ticket and maybe add your details to the existing one: the more is consolidated in one ticket (data, people who experience it, …), the higher the likelihood is that resources are allocated to its solution.
If you already know of a bug ticket upstream, mention it as well of course (the same about topics here in ask.fp).
Sure you don’t have these from rpmfusion? (which is a different bug tracker → see my last post). You can verify the repo with dnf info mesa* --installed
If I get the problem right, it might make sense to create a bug ticket in both (and link both to each other), at least in case you cannot identify if it is one of the rpmfusion or one of the default packages that causes the issue. The mentioned packages are interdependent and the update will be performed by dnf only once all dependencies are available (which is why the updates of the default repo usually wait some time until the rpmfusion packages are updated as well, and then all mentioned packages are updated at once - the rpmfusion packages usually take a few days longer → outcome: you cannot test them individually in an easy manner).
Sounds good. I indeed could imagine that something GNOME-related is either the cause of the issue or contains the code that provokes the issue: I also have the same mesa packages on an amdgpu but with KDE, and I don’t have any issues (alternative: it is related to differences caused by different amdgpu hardware, but I think that’s less likely)
Actually this is more than a GNOME problem. I’m having issue with SDDM alone. After upgrading to 24.3.3-2, SDDM crashes on boot. Here’s the related log from journalctl -b -1:
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.126] Loading module '/usr/lib64/libweston-14/gl-renderer.so'\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.130] Using rendering device: /dev/dri/renderD128\n[18:10:49.130] EGL version: 1.5\n[18:10:49.130] EGL vendor: Mesa Project\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.130] EGL client APIs: OpenGL OpenGL_ES \n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.130] EGL features:\n EGL Wayland extension: yes\n context priority: yes\n buffer age: yes\n >
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.131] failed to create context\n[18:10:49.132] EGL error state: EGL_BAD_ALLOC (0x3003)\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.132] EGL error state: EGL_SUCCESS (0x3000)\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.132] failed to initialize egl\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "[18:10:49.132] fatal: failed to create compositor backend\n"
Jan 16 18:10:49 luke-desktop sddm-helper-start-wayland[1814]: "Internal warning: debug scope 'drm-backend' has not been destroyed.\n"
I have temporarily rollbacked to 24.2.4-1.
edit: I have a 7800XT dGPU and AMD iGPU. The monitors are attached to the dGPU ports.
I have currently no external monitor - you might try to see if it is the external screen that provokes the issue? I have a AMD Ryzen 7 PRO 6850U & Radeon 680M on F41 KDE Spin, default packages except the two freeworld mesa packages from rpmfusion. So far, no issues.
This issue is not in my area of expertise, but can affected people provide an output of sudo journalctl --boot=0 --no-hostname -k (please put it in a code box, or use an external link) during an affected boot? Maybe with some details of times you experience issues and what you did at the very time? Also, immediately after boot and again after you experienced an issue, what is the respective output of cat /proc/sys/kernel/tainted (might save time and indicates where/when/what to check in the logs)? I would skim it to see if something catches my eye off the cuff (that might be also useful information for the bug reports anyway)
Maybe with some details of times you experience issues and what you did at the very time?
Basically SDDM never started (and C-M-F1 stayed dark). I switched to tty using C-M-F3 and used command line to downgrade and inspect logs.
Should I open a separate thread to discuss my issue? What I experienced (SDDM crashing) might be unrelated to the performance degradation OP experienced.
If you have the same mesa packages installed like the OP, and if your issue occurred also after the very update the same way as at the machine of the OP, and if the mitigation that helps the OP also helps you: stay here. If any of these points does not apply to you, I suggest to open a separated topic and assume that you have a different issue, although you are encouraged to mention in your new topic any similarities to other issues/topics (including to this one). However, a confirmation here would be nice so that we know if that can also occur on KDE or not. Otherwise, I would go back to the assumption that this might be somehow linked (caused or provoked) by anything from GNOME
Thanks