Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

Per-Astra-Ad-Deum · 2024-08-11 12:59:43

Hi there,

I've got a new Laptop with a new Intel Ultra 155H CPU with Intel Arc Graphics (Meteor Lake).

Sadly I can't make use of hardware acceleration because the output glitches with heavy artifacts, to the point where one can't even make out what is being displayed.

This error occurs on any Chromium-based browsers as well as any Electron-based apps like FreeTube for example, as well as on Virt-Manager when using 3D acceleration and OpenGL Spice-Server.

Hardware acceleration does work just fine on MPV though!

Output of "vainfo":

Trying display: wayland
vainfo: VA-API version: 1.22 (libva 2.22.0)
vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 24.2.5 ()
vainfo: Supported profile and entrypoints
      VAProfileNone                   : VAEntrypointVideoProc
      VAProfileMPEG2Simple            : VAEntrypointVLD
      VAProfileMPEG2Main              : VAEntrypointVLD
      VAProfileH264Main               : VAEntrypointVLD
      VAProfileH264Main               : VAEntrypointEncSlice
      VAProfileH264High               : VAEntrypointVLD
      VAProfileH264High               : VAEntrypointEncSlice
      VAProfileJPEGBaseline           : VAEntrypointVLD
      VAProfileJPEGBaseline           : VAEntrypointEncPicture
      VAProfileH264ConstrainedBaseline: VAEntrypointVLD
      VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
      VAProfileVP8Version0_3          : VAEntrypointVLD
      VAProfileHEVCMain               : VAEntrypointVLD
      VAProfileHEVCMain               : VAEntrypointEncSlice
      VAProfileHEVCMain10             : VAEntrypointVLD
      VAProfileHEVCMain10             : VAEntrypointEncSlice
      VAProfileVP9Profile0            : VAEntrypointVLD
      VAProfileVP9Profile0            : VAEntrypointEncSlice
      VAProfileVP9Profile1            : VAEntrypointVLD
      VAProfileVP9Profile1            : VAEntrypointEncSlice
      VAProfileVP9Profile2            : VAEntrypointVLD
      VAProfileVP9Profile2            : VAEntrypointEncSlice
      VAProfileVP9Profile3            : VAEntrypointVLD
      VAProfileVP9Profile3            : VAEntrypointEncSlice
      VAProfileHEVCMain12             : VAEntrypointVLD
      VAProfileHEVCMain422_10         : VAEntrypointVLD
      VAProfileHEVCMain422_12         : VAEntrypointVLD
      VAProfileHEVCMain444            : VAEntrypointVLD
      VAProfileHEVCMain444            : VAEntrypointEncSlice
      VAProfileHEVCMain444_10         : VAEntrypointVLD
      VAProfileHEVCMain444_10         : VAEntrypointEncSlice
      VAProfileHEVCMain444_12         : VAEntrypointVLD
      VAProfileHEVCSccMain            : VAEntrypointVLD
      VAProfileHEVCSccMain            : VAEntrypointEncSlice
      VAProfileHEVCSccMain10          : VAEntrypointVLD
      VAProfileHEVCSccMain10          : VAEntrypointEncSlice
      VAProfileHEVCSccMain444         : VAEntrypointVLD
      VAProfileHEVCSccMain444         : VAEntrypointEncSlice
      VAProfileAV1Profile0            : VAEntrypointVLD
      VAProfileAV1Profile0            : VAEntrypointEncSlice
      VAProfileHEVCSccMain444_10      : VAEntrypointVLD
      VAProfileHEVCSccMain444_10      : VAEntrypointEncSlice

Output of "pacman -Qs 'libva'":

local/libva 2.22.0-1
    Video Acceleration (VA) API for Linux
local/libva-intel-driver 2.4.1-3
    VA-API implementation for Intel G45 and HD Graphics family
local/libva-utils 2.22.0-1
    Intel VA-API Media Applications and Scripts for libva

I'd be happy for any advice at this point.

kind regards

seth · 2024-08-11 13:51:13

Same issue w/ electron on X11?
Do you end up using xwayland?
https://wiki.archlinux.org/title/Chromi … nd_support
https://wiki.archlinux.org/title/Chromi … er_Wayland

And please clarify "chromium based" - does that include chromium (or maybe vivaldi) or are we talking about some MS Edge flatpak?

Per-Astra-Ad-Deum · 2024-08-11 16:11:28

seth wrote:

Same issue w/ electron on X11?
Do you end up using xwayland?
https://wiki.archlinux.org/title/Chromi … nd_support
https://wiki.archlinux.org/title/Chromi … er_Wayland

Thank you for the hints! I was able to solve the issue on Chromium thanks to the links you've provided, by enabling/forcing native Wayland mode on Chromium. Frankly the thought that it could have been Wayland-related never occurred to me.
Now hardware acceleration is working in Chromium, as confirmed by 'Intel_gpu_top'.

seth wrote:

And please clarify "chromium based" - does that include chromium (or maybe vivaldi) or are we talking about some MS Edge flatpak?

Chromium and Brave.

I haven't checked on Electron-based apps so far, could it be the same issue? And what about Virt-Manager?

best regards

seth · 2024-08-11 19:46:42

https://wiki.archlinux.org/title/Waylan … s_visually
virt-manager doesn't look like it's using electron at all, might still be an X11 only client.
Sanity check

pacman -Qs xf86-video

Per-Astra-Ad-Deum · 2024-08-11 21:02:48

According to 'xorg-xeyes' and 'xlsclients -l' virt-manager is not running on XWayland.

pacman -Qs xf86-video

returns nothing!

best regards

seth · 2024-08-11 21:36:49

When you say "virt-manager", you probably mean the guest and https://wiki.archlinux.org/title/Virt-m … celeration - not the libvirt GUI?
What is the guest and what virtual hardware does it detect?

Per-Astra-Ad-Deum · 2024-08-11 21:49:01

seth wrote:

When you say "virt-manager", you probably mean the guest and https://wiki.archlinux.org/title/Virt-m … celeration - not the libvirt GUI?
What is the guest and what virtual hardware does it detect?

Yes, the output from the guest. The "window" by which I'm looking into the Guest machine, for a lack of better description.

This affects every guest system I tried so far. Doesn't matter if Debian, Fedora or Arch. On Debian for example it detects "Red Hat, Inc. Virtio 1.0 GPU" for GPU.

These here (https://wiki.archlinux.org/title/Virt-m … celeration) are the exact settings I've been using in the past to accelerate my VMs. On the new Meteor Lake machine they produce the described faulty output.

Deactivating 3D acceleration and OpenGL fixes the issue, but of courses leaves me without acceleration.

best regards

antttezy · 2024-10-13 20:08:19

Looks like I found a temporary solution that worked for me. This issue is only observed on iGPUs with SMA memory. By adding

xe.vram_bar_size=2048

to GRUB_CMDLINE_LINUX, you can set VRAM to a fixed size (2 GiBs)
Also make sure you are using xe kernel module and not older i915

Per-Astra-Ad-Deum · 2024-10-18 21:37:12

antttezy wrote:

Looks like I found a temporary solution that worked for me. This issue is only observed on iGPUs with SMA memory. By adding
xe.vram_bar_size=2048
to GRUB_CMDLINE_LINUX, you can set VRAM to a fixed size (2 GiBs)
Also make sure you are using xe kernel module and not older i915

Sorry for the late reply. I gave up hope anyone would answer here long ago. :-D

Thank you for taking your time and dropping an answer here. Sadly, I've tried it, and it didn't fix the issue for me. Instead of showing me heavy artifacts, it just showed me nothing, a black screen.

My other issues with the iGPU aren't gone either.

Thank you nevertheless for the hint! Maybe it will lead to the real issue!

seth · 2024-10-19 13:23:30

Sorry, this dropped off the table (generally keep grinding at issues as long as they remain, post new information etc.)

I was able to solve the issue on Chromium thanks to the links you've provided, by enabling/forcing native Wayland mode on Chromium.

The problem seems somewhat related to xwayland, but that doesn't fit your virtmanager assertion, so let's see:
Is eglgears_wayland affected? eglgears_x11? glxgears?

Do the artifacts show up on screenshots? Can you share one?
Does it help to disable DRI3?
https://wiki.archlinux.org/title/Intel_ … RI3_issues

Per-Astra-Ad-Deum · 2024-10-22 14:52:16

seth wrote:

Sorry, this dropped off the table (generally keep grinding at issues as long as they remain, post new information etc.)

Thank you for getting back to me.

seth wrote:

Is eglgears_wayland affected? eglgears_x11? glxgears?

I've tried all of those three and they all seem to work pretty much flawlessly. No artifacts, no tearing, no glitches.

seth wrote:

Do the artifacts show up on screenshots? Can you share one?

There's a ton of glitches. Let me see if I can post some screenshots here.

1. First example:

I have a big issue with my virtual machines (QEMU/KVM, libvirt/virt-manager): Whenever I enable hardware acceleration with virtio, the graphics output gets scrambled.

This is how it's supposed to look (picture taken without 3D acceleration enabled):

The actual output with 3D acceleration looks like this however:

The only difference between the two screenshots is that in the second, the option "3D acceleration" (under Video Virtio) as well as the corresponding "Listen type" value "none" (under Display Spice) is set.
Every other is setting is left on default, safe for changing firmware to UEFI in the initial install dialog.

2. Second example:

Random glitches/artifacts in various programs (KDE's Dolphin, Firefox, FreeTube etc):

Look at the artifacts in the preview image and above.

3. Third example:

Again, a comparison. First image is from a game (Doom III, running natively on Linux). This is when running on my A770 (attached as eGPU via OCuLink):

This is the very same scene/save game, this time running on the iGPU. You can still make out the mangohud overlay in the top left corner and some of the HUD's elements:

This is the game's menu running in iGPU. Notice how the artifacts appear to be similar in kind to the ones occurring with libvirt/FreeTube:

seth wrote:

Does it help to disable DRI3?
https://wiki.archlinux.org/title/Intel_ … RI3_issues

Hmm. I'm willing to give it a try, but isn't that setting linked to xf86-video-intel? I don't have that package installed. Should I try nevertheless?

Thank you for your help!

seth · 2024-10-22 19:17:16

but isn't that setting linked to xf86-video-intel?

No, but the mesa/modesetting DRI3 feature is only controlled w/ an environment variable, not a server config option.

I've tried all of those three and they all seem to work pretty much flawlessly. No artifacts, no tearing, no glitches.

Ok… what about vkcube and vkcube-wayland?
And the "tunnel" demo from mesa-demos (it allows you to toggle textures…)

Per-Astra-Ad-Deum · 2024-10-22 21:51:36

seth wrote:

No, but the mesa/modesetting DRI3 feature is only controlled w/ an environment variable, not a server config option.

Alright, so here's what I did. I've added

LIBGL_DRI3_DISABLE=1

to my /etc/environment file and rebooted.

However, that did nothing to alleviate the issue. :-(

seth wrote:

Ok… what about vkcube and vkcube-wayland?

Those run without issue. Side-note: how many cubes and gears are there for testing? :-D

seth wrote:

And the "tunnel" demo from mesa-demos (it allows you to toggle textures…)

That one also seems to run fine. Ran it, played with all the commands, no issues whatsoever.

I still think it has something to do with the error it's throwing at me at every boot:

i915 0000:00:02.0: [drm] *ERROR* GT1: GSC proxy component didn't bind within the expected timeout
i915 0000:00:02.0: [drm] *ERROR* GT1: GSC proxy handler failed to init

I've reported it in this thread: https://bbs.archlinux.org/viewtopic.php?id=298867

There I've been suggested to try various bleeding-edge kernels, but none of them helped me get rid of this issue. :-(

Last edited by Per-Astra-Ad-Deum (2024-10-24 13:29:51)

seth · 2024-10-23 21:18:48

I still think it has something to do with the error it's throwing at me at every boot:

That's news, or did you mention it before and I missed it?
=> https://bbs.archlinux.org/viewtopic.php … 3#p2195183
Maybe better post your complete system journal for the boot:

sudo journalctl -b | curl -F 'file=@-' 0x0.st

Per-Astra-Ad-Deum · 2024-10-24 09:46:07

seth wrote:

That's news, or did you mention it before and I missed it?

No actually, I've mentioned this post in the one I've linked though, and you commented there as well. I should have mentioned that earlier, you think the issues are related?

seth wrote:

=> https://bbs.archlinux.org/viewtopic.php … 3#p2195183

I've given that a try and added

initcall_blacklist=simpledrm_platform_driver_init

to the kernel parameters. i915 was already present in the MODULES section on mkinitcpio.conf. Sadly, the error still persists.

seth wrote:

Maybe better post your complete system journal for the boot

Sure, here you go: http://0x0.st/XUq3.txt

seth · 2024-10-24 12:52:39

VPU errors are probably insignificant, GSC is used to update the Intel graphics system controller via the Intel Management Engine - I don't think that's causing these *extremely* selective artifacts.

https://www.reddit.com/r/linuxhardware/ … yland_kde/ alleges that the upcoming 6.12 kernel improves the situaiton.
You can also test https://aur.archlinux.org/packages/linux-drm-tip-git

Per-Astra-Ad-Deum · 2024-10-24 13:06:33

seth wrote:

VPU errors are probably insignificant, GSC is used to update the Intel graphics system controller via the Intel Management Engine - I don't think that's causing these *extremely* selective artifacts.
https://www.reddit.com/r/linuxhardware/ … yland_kde/ alleges that the upcoming 6.12 kernel improves the situaiton.
You can also test https://aur.archlinux.org/packages/linux-drm-tip-git

I've already tried the linux-drm-tip-git-kernel (see the other topic). Do you think the situation has improved and warrants a new try?

seth · 2024-10-24 13:26:32

There's a very recent batch of intel patches on the head of that git tree, so if your last test was in august, yes: try.

Sorry, I didn't get https://bbs.archlinux.org/viewtopic.php?id=298867 (your link is 404 and aktschually, you're not supposed to open new thread on the same problem but continue the old one

Per-Astra-Ad-Deum · 2024-10-24 13:31:16

seth wrote:

There's a very recent batch of intel patches on the head of that git tree, so if your last test was in august, yes: try.

Alright, will do!

seth wrote:

Sorry, I didn't get https://bbs.archlinux.org/viewtopic.php?id=298867 (your link is 404 and aktschually, you're not supposed to open new thread on the same problem but continue the old one

Thank you, I've fixed the link now. And aktschual-aktschually, the two topics being related has yet to be confirmed!

Per-Astra-Ad-Deum · 2024-10-29 15:00:09

Just a quick heads-up:

I've tried installing https://aur.archlinux.org/packages/linux-drm-tip-git twice. Both times it took several hours to compile, only to fail midways to some bullshit Python-related error. First time it complained about pyhton-yaml not being found so I've installed it, second time some other python-package apparently wasn't found, although it was installed. *sigh*

I've decided to wait until 6.12 hits core-testing and give that a chance. I'll report it the situation improves!

seth · 2024-10-29 15:45:33

python is only relevant to build the (html) documentation, you can comment the relevant bits in the PKGBUILD

Per-Astra-Ad-Deum · 2024-11-09 12:30:16

Update:

Well, this changes things I guess. This is on a different system (i7-7700K with Intel HD 630 iGPU) and different OS (OpenSuse Tumbleweed). I guess this issue isn't Meteor-Lake- or Arch-exclusive.

Don't know how to proceed from here though. Who should I report this bug to?

seth · 2024-11-09 16:22:40

This error occurs on any Chromium-based browsers as well as any Electron-based apps like FreeTube for example

Same there as well?

Have you tried the behavior w/ the LTS kernel?

Although this is a really weird bug across systems but not reported by anyone else… :\
Do you use the same $HOME on SuSE?
Do you get this w/ a fresh user account?

Per-Astra-Ad-Deum · 2024-11-09 17:42:41

seth wrote:

This error occurs on any Chromium-based browsers as well as any Electron-based apps like FreeTube for example
Same there as well?
Have you tried the behavior w/ the LTS kernel?
Although this is a really weird bug across systems but not reported by anyone else… :\
Do you use the same $HOME on SuSE?
Do you get this w/ a fresh user account?

This was a fresh installation on a friends desktop machine. Absolute clean slate. No hardware or files are configs or whatever in any way associated with my system.

I was so surprised by this that I forgot to check if this error occurs with other applications as well.

And yes, this is strange. This should affect hundreds of users, yet I see now bug reports anywhere about this?!

seth · 2024-11-09 19:27:24

seth wrote:

This error occurs on any Chromium-based browsers as well as any Electron-based apps like FreeTube for example
Same there as well?

seth wrote:

Have you tried the behavior w/ the LTS kernel?

Arch Linux

#1 2024-08-11 12:59:43

Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#2 2024-08-11 13:51:13

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#3 2024-08-11 16:11:28

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#4 2024-08-11 19:46:42

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#5 2024-08-11 21:02:48

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#6 2024-08-11 21:36:49

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#7 2024-08-11 21:49:01

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#8 2024-10-13 20:08:19

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#9 2024-10-18 21:37:12

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#10 2024-10-19 13:23:30

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#11 2024-10-22 14:52:16

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#12 2024-10-22 19:17:16

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#13 2024-10-22 21:51:36

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#14 2024-10-23 21:18:48

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#15 2024-10-24 09:46:07

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#16 2024-10-24 12:52:39

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#17 2024-10-24 13:06:33

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#18 2024-10-24 13:26:32

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#19 2024-10-24 13:31:16

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#20 2024-10-29 15:00:09

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#21 2024-10-29 15:45:33

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#22 2024-11-09 12:30:16

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#23 2024-11-09 16:22:40

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#24 2024-11-09 17:42:41

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

#25 2024-11-09 19:27:24

Re: Heavy artifacts with hardware acceleration on Meteor Lake Arc iGPU

Board footer