01:43kisak: The new AMD cards have a problem rendering the cursor with X and the normal hardware plane? weird. ( https://www.google.com/search?q=7900+swcursor )
02:59agd5f: kisak, you need this patch if you are using X: https://cgit.freedesktop.org/xorg/driver/xf86-video-amdgpu/commit/?id=9c959fac3af28d191105f63236096ad456dca614
03:20kisak: thanks for the hint. I don't have the hardware to be affected, just happened across it in a random forum and wasn't looking in the right places for any relevant info to see if it was previously evaluated.
05:18gekret005: i happen to have such hardware and am glad i know where the fix is :)
05:18gekret005: i just disabled hardware cursor
18:24gitargy: Hello, recently purchased a ASRock Challenger D Radeon RX 6600 XT. I am using it on debian 11 with kernal 6.0. While it techinacally works the performance on benchmarks is significantly lower than expected. I have exceeded my ability to diagnose the problem myself. Can yall help me out?
18:27gitargy: paste.debian.net/1266992
18:27gitargy: output of dmsg
18:29gitargy: I downloaded dimgrey_cavefish_sos.bin from the linux firmware git repo, but this error has persisted regardless
18:32Remco: Did you also rebuild the initramfs?
18:32gitargy: I have not done this.
18:34gitargy: What is it and how do it rebuild it?
18:36Remco: https://wiki.debian.org/Firmware
18:36Remco: If you put it in the right place, you only need the last command to update initramfs
18:37Remco: As for what that is, see https://wiki.debian.org/initramfs
18:38gitargy: $ sudo update-initramfs -c -k all
18:38gitargy: run this?
18:38Remco: That's the one yes
18:38Remco: And then after a reboot it should magically work
18:39Remco: You do need to keep it up to date yourself, so keep that in mind
18:40gitargy: might have a problem
18:41gitargy: paste.debian.net/plain/1266993
18:41KitsuWhooa: you typed alDl instead of all
18:42Remco: Yep
18:42gitargy: ok tyty
18:43KitsuWhooa: Also, maybe it's a better idea to install firmware-amd-graphics
18:43Remco: I checked, it's not in there
18:44KitsuWhooa: https://packages.debian.org/sid/firmware-amd-graphics
18:44KitsuWhooa: I assume you need more than just the _sos one
18:44gitargy: firmware-amd-graphics is already the newest version (20210315-3).
18:44KitsuWhooa: You should be able to manually download one from a newer version and install it
18:46gitargy: do i download it and run dpkg -i?
18:46KitsuWhooa: I think so, yeah
18:47gitargy: from the bookworm repo?
18:47KitsuWhooa: https://packages.debian.org/bookworm/firmware-amd-graphics says it has it in the bookworm repo, so yes
18:48gitargy: ok do that then reboot. Right?
18:48KitsuWhooa: yes
19:07gitargy: howdy, I'm still getting the same dmesg error
19:07gitargy: performance is still very sub-optimal
19:08_ds_: I'd grab that from testing/unstable rather than backports.
19:09_ds_: (firmware-amd-graphics)
19:09_ds_: Or for that matter from stable.
19:09gitargy: which thing?
19:09gitargy: firmware and graphics
19:10gitargy: *frimware-amd-graphics
19:10_ds_: backports has 20210818-1~bpo11+1; testing & unstable have 20221214-3
19:11KitsuWhooa: stable doesn't have the firmware blobs for this card
19:11gitargy: bookworm or bulleye repo?
19:12KitsuWhooa: gitargy: which version did you install?
19:13gitargy: amdgpu-install_5.4.50401-1_all.deb
19:13_ds_:doesn't see any package of that name
19:13KitsuWhooa: You said you installed a deb for
19:13KitsuWhooa: firmware-amd-graphics
19:14KitsuWhooa: which version was that?
19:14_ds_: Scroll up a bit and you'll see 20210315-3, which is what's in stable.
19:14gitargy: apt shows as (20210315-3)
19:15KitsuWhooa: And I told them to install the deb from bookworm
19:16_ds_: ⇒ https://packages.debian.org/bookworm/all/firmware-amd-graphics/download (pick a mirror)
19:16gitargy: do I need to add the bookworm repo to sources?
19:16KitsuWhooa: http://ftp.us.debian.org/debian/pool/non-free/f/firmware-nonfree/firmware-amd-graphics_20221214-3_all.deb
19:17gitargy: nvm I autocompleated the wrong thing
19:17gitargy: my bad m8
19:17gitargy: tierd AF
19:18gitargy: we gonna try this again
19:25gitargy: Kitsuwhooa worked like a charm, benchmarks went up by a factor of 10. thanks a bunch m8
19:27KitsuWhooa: nice
20:06haagch: huh, I managed to make kwin_wayland sad without a mst adapter. so the mst adapter was not the sole cause
20:08haagch: i have 2 displayport and 1 hdmi monitor attached. I swapped my two displayport monitors, basically reconnecting each to the cable the other one was connected (while tty switched to an x server, not sure if that plays a role). then i switched back to kwin_wayland and everything was fine, it even automatically switched the desktops
20:10haagch: then I set the freesync mode on the left displayport monitor from standard to advanced, effectively reconnecting it. then the kwin display settings became very confused and showed both displayport monitors with the same name and it kept detecting the hdmi monitor connecting and disconnecting. when switching to x11, everything was fine, switching back to wayland, reconnection hell again
21:49gekret005: haagch: I figured out the cause of the bad performance it seems that dxvk isn't free
21:50gekret005: also I had friends that didn't bother to check and see how they were actually running their games
21:50gekret005: leading to wrong expectations
21:53gekret005: async hides real gpu bottlenecks
21:54haagch: isn't free?
21:56gekret005: it adds gpu load
21:56gekret005: at least I think unless there's a way to itemize performance items on the gpu
21:58haagch: not the expert here but it probably can be worse or better, entirely depending on the actual code that is being run
21:59haagch: iirc back when proton was released, valve said after their optimizations, beat saber run slightly better on linux with proton
22:01haagch: and people have been reporting better performance with dxvk than native d3d11 in various games on windows
22:11HdkR: Definitely heavily depends, especially if you're comparing against an entirely different driver stack on the Windows side.
22:33gekret005: ill have to test more games but the ones i play the most perform worse it seems which made me pull hairs out till i started asking others to test