Closed MasterMind2k closed 1 year ago
No idea where you got that from, but it shouldn't be necessary.
I used that featuremask to get rid of a latency issue in the system console. I have not tested whether it is still necessary, or whether it was only necessary due to a quirk of my own hardware. It might also be made unnecessary by an update to the kernel driver.
Doesn't look like it's still necessary, I removed it and there is no latency. I had the latency before though. I think this issue could be closed in general?
I'd prefer to get a couple more user reports.
Through weekend I'll re-insert my card into machine and verify if it works on nixos-unstable.
If I understand release cycles of NixOS, in april (23.04) this will be on stable channel?
23.05 is the target, so May.
For simplicity, here's all that's needed to get the 7900 XTX working as of now:
inputs.nixpkgs.url
points to github:NixOS/nixpkgs/nixos-unstable
services.xserver.videoDrivers
to [ "modesetting" ]
(important. if it's set to amdgpu
you will get a black screen as soon as X is started. wayland seems to work fine, but I haven't tested it much)I've been running this on linuxPackages_testing
(6.2.0-rc6) for around 12 hours with no issues. If the above doesn't work for you, try also changing your kernel to linuxPackages_testing
.
With the current nixos-unstable and linuxPackages_testing
it's all peachy, even with amdgpu
. With 6.2-rc7 instead of the current linuxPackages_testing
which is 6.2-rc6 I get EGL_BAD_ALLOC
as soon as I log in to plasma (wayland).
The current version of Mesa in unstable (22.3) still has quite a few bugs with the 7900 XTX, particularly related to crashes. Looking at the Mesa repository it seems like there have been multiple fixes submitted recently, but unfortunately I couldn't get Mesa 23 (which includes these fixes) to work reliably either.
Running with card. Without setting "modesetting" as video driver for xserver I don't see the cursor. Running xserver on modesetting I do notice HiDPI behaviour is a little bit off.
Other than that, it seems to work. Will write down if I get any crashed during the week.
@MasterMind2k the invisible cursor issue is fixed in the xorg amdgpu driver repo but they've not cut a release for it.
Just want to report that other than the kernel console and USB-C port oddities, I have had no issues running my 7900 XTX on NixOS now that LLVM15 and the updated Mesa builds are in Nixpkgs. Though I am running Wayland with Sway, so not sure about Xorg.
OpenGL and Vulkan programs both work perfectly, and Blender properly detects it as a HIP device (using blender-hip
package). The only problems I'm having appear to be issues upstream (both in mesa
and amdvlk
). Steam games (both native and Proton) seem to work fine as well.
Just wanted to add my 2-cents in case people are still waiting for further reports to close this ticket.
@MasterMind2k the invisible cursor issue is fixed in the xorg amdgpu driver repo but they've not cut a release for it.
Is there a ticket I could track? I would like to test it as soon as it hits nixos-unstable.
Please just use the modesetting
driver.
Please just use the
modesetting
driver.
I am and using the hardware already without bigger problems. But HiDPI setting is a little bit different than on amdgpu and would verify it when the cursor on amdgpu is fixed. Some apps in FHS has small font with modesetting driver, that with amdgpu do not.
From my point of view, this ticket can be closed before amdgpu driver fix is released.
That cannot possibly be affected by the DDX, assuming the other settings are unchanged.
You're not running Plasma Wayland, are you?
They haven't released for a year: https://github.com/freedesktop/xorg-xf86-video-amdgpu/tags
That cannot possibly be affected by the DDX, assuming the other settings are unchanged.
I am just reporting what I saw. Granted, it might be something else. But that was the only change :S.
They haven't released for a year: https://github.com/freedesktop/xorg-xf86-video-amdgpu/tags
I understand now :). I assumed they had shorter cycles. As said earlier, I just wish to track to check if HiDPI problem is related to DDX driver at a later date. Not to block this issue to be resolved, from my point of view, problem is resolved.
I had this HiDPI issue since moved to AMD Ryzen 9 7900X and Radon RX 7900 XTX hardware and was curious why using amdgpu xserver driver suddenly fixed it.
Please just use the modesetting driver.
In that case, should the default order of drivers be changed to reflect that?
It should probably just default to [ "modesetting" "fbdev" ]
or possibly even [ "modesetting" ]
at this point.
Is there a ticket I could track? I would like to test it as soon as it hits nixos-unstable.
https://gitlab.freedesktop.org/xorg/driver/xf86-video-amdgpu/-/issues/61
I can confirm the 7900xtx is working fine with linux 6.2 and the modesetting video driver.
I think we'd better mark this as fixed. Of course, discussion can continue here and/or new issue can be linked with more up to date description.
Well, on NixOS 22.11 these GPUs probably don't work nice, but that would be expected due to being newer than 22.11.
xorg-video-amdgpu version 23.0.0 just got released with the mentioned patch.
If I see correctly here, it is still at 21.0.0?
xorg-video-amdgpu version 23.0.0 just got released with the mentioned patch.
If I see correctly here, it is still at 21.0.0?
Am I right, when i say: When I do an overlay with pulling in xorg-video-amdgpu version 23.0.0 i dont need „modesetting“ anymore ? Or has this nothing to do with it ?
As stated multiple times before, please use modesetting
.
As stated multiple times before, please use
modesetting
.
Please let us understand why? It seems the xserver amdgpu driver has released a version that fixed the issue with hidden cursor.
Is a policy we should all use modesetting from now on? I am fine if default is set to modesetting and is then my decision to swap to amdgpu xorg driver when it is fixed. Is amdgpu xorg driver going to be removed? If so, why?
@MasterMind2k the invisible cursor issue is fixed in the xorg amdgpu driver repo but they've not cut a release for it.
@edel-km, this comment says so. But I am unable to verify.
modesetting
is a generic driver that uses device independent kernel APIs. It's much more actively maintained, and those APIs are also used by all Wayland compositors, Android and pretty much everyone else. The "X11 driver" model is effectively dead, and all the remaining drivers are on life support.
modesetting
is a generic driver that uses device independent kernel APIs. It's much more actively maintained, and those APIs are also used by all Wayland compositors, Android and pretty much everyone else. The "X11 driver" model is effectively dead, and all the remaining drivers are on life support.
Thank you. This makes it crystal clear!
It's probably also worth explicitly mentioning that using the modesetting
driver still provides full hardware acceleration for graphics, video decode and basically everything else.
It's probably also worth explicitly mentioning that using the
modesetting
driver still provides full hardware acceleration for graphics, video decode and basically everything else.
Maybe this is a stupid question, but why would we even want to use amdgpu instead of modesetting then?
You generally don't want to use it.
Then we should also go about updating the wiki page where it is explicitly recommended to use amdgpu
That is woefully outdated, as you can see by the fact that it refers to NixOS 20.09.
Well, I suppose we could update it either way: https://github.com/NixOS/nixpkgs/pull/218433
As seen in the release announcement and specifically pointed out in the related Phoronix article, the new release only has a couple of actually meaningful changes. Certainly looks a lot like a minimal life support release given the long time since the previous one.
This issue has been mentioned on NixOS Discourse. There might be relevant details there:
https://discourse.nixos.org/t/setting-up-wayland-gnome-in-nixos-with-amd-rx-7900-xt/24463/4
Describe the bug
Booting into system with Radeon RX 7900 XTX does not show cursor nor has hardware acceleration.
Steps To Reproduce
Steps to reproduce the behavior:
Expected behavior
A shown hardware cursor and hardware acceleration.
Screenshots
N/A
Additional context
Issue for tracking addition of new support.
Notify maintainers
@K900 @LunNova (gathered from linux-firmware git blame)
Metadata