[arch-general] intel gpu hang

Ralf Mardorf silver.bullet at zoho.com
Mon Jan 11 20:22:33 UTC 2021

On Mon, 11 Jan 2021 01:58:57 +0000, Gustavo De Nardin wrote:
>Reading the Arch wiki on Intel graphics
>(https://wiki.archlinux.org/index.php/Intel_Graphics#Installation) I
>noticed the note about Debian & Ubuntu, Fedora, KDE recommending the
>modesetting driver instead of the xf86-video-intel one. So I tried the
>modesetting driver, and have had no more issues so far. So, I suggest
>trying that (if you're not already using it, being on Fedora). I
>haven't noticed any performance loss.

After removing the xf86-video-intel package and changing the driver in
/etc/X11/xorg.conf from "intel" to "modesetting" I got rid of most
issues. Just one time Firefox suffered from a short, but still too long
performance issue (maybe a random thing), but usually even Firefox seems
to perform better, apart from this the performance of all other apps
seems to be faster, which not necessarily is an important win, it's
just good that it seems to be no loss, performance with 4.19 kernels
and the "Intel" driver is very good, too, maybe scarcely perceptible
slower. FWIW this machine does use SSDs only.

Now at least

$ uname -rm
5.10.6-arch1-1 x86_64

is usable.

Virtualbox does work, but there's one pitfall, Google-Chrome is that
broken now, that I needed to switch to tty2 and to reboot. I did not
test impact on pro-audio and MIDI performance. I even did not boot with
"threadirqs". I've done no 3D test or video acceleration test and still
did not test dual head yet.

The session where I needed to restart + this session, both sessions
together run for around 2 hours, so more good, bad or ugly might happen.

On Mon, 11 Jan 2021 16:40:10 +0000, u34--- via arch-general wrote:
>The nature of, supposedly light, memory failure is that they start
>with few memory cells. On the other hand, if I remember correctly, you
>switched hardware. Presumably with new, fresh from the facory, memory

This machine is a few years old. However, when using different kernels
or migrating to older packages, issues disappear. The issues would not
disappear if RAM would be broken.

Fazit: It seems to be that migrating from the "intel" to "modesetting"
driver solved most, if not all issues, but unable to use google-chrome
is a serious new issue.

More information about the arch-general mailing list