2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

petelewis · 2010-06-24 08:37:38

Thanks oceans11 - and btw your English is much better than my Italian :-D

Apparently the hangcheck timer message merely indicates that the kernel noticed that the GPU has hung. This could be caused by any number of things, I guess.

There are a couple of related bug reports on freedesktop.org already:

http://bugs.freedesktop.org/show_bug.cgi?id=27187
http://bugs.freedesktop.org/show_bug.cgi?id=26345

The first seems the most likely to be related and is linked to these on arch's bug tracker:

http://bugs.archlinux.org/task/16974
http://bugs.archlinux.org/task/18356
http://bugs.archlinux.org/task/17705

Although adding xorg 1.8 to the mix on Archlinux seems to be triggering this very regularly.

amen · 2010-06-29 16:31:00

I have ASUS 1001HA netbook with:

00:02.0 VGA compatible controller: Intel Corporation Mobile 945GME Express Integrated Graphics Controller (rev 03)

Similar performance problem here. After upgrade xorg I have 40FPS in glxgears(before: 500FPS).

dominikh · 2010-06-29 23:35:28

bboston7 wrote:

I fixed this by downgrading xf86-video-intel, libgl, intel-dri, mesa, and xorg-server. I'll attach my pacman.log with the relevant information for anything I might have forgotten.

Hm, I tried that, too, but performance didn't get any better

pacman -U xf86-video-intel-2.10.0-1-i686.pkg.tar.gz libgl-7.7.1-0.1-i686.pkg.tar.xz dri2proto-2.1-2-any.pkg.tar.xz xf86-input-evdev-2.3.2-1-i686.pkg.tar.gz xorg-server-1.7.6-3-i686.pkg.tar.xz xorg-server-1.7.6-3-i686.pkg.tar.xz

Cdh · 2010-06-30 04:45:31

The 945gme IS slow and glxgears is not a benchmark.
I would suggest trying xf86-video-intel from testing. There is the 2.12 beta, which should improve 2D performance a lot:
http://www.phoronix.com/scan.php?page=a … 2_2d&num=2

gabe_ · 2010-06-30 13:06:20

Cdh wrote:

The 945gme IS slow and glxgears is not a benchmark.
I would suggest trying xf86-video-intel from testing. There is the 2.12 beta, which should improve 2D performance a lot:
http://www.phoronix.com/scan.php?page=a … 2_2d&num=2

2.12 hasn't found it's way into testing yet. The latest driver there is 2.11.901. 2.12RC1 can be found in AUR though.

EDIT: It also does not fix the 3D problems discussed in this tread... At least not on my 915GM/GMS/910GML Express Graphics Controller.

Last edited by gabe_ (2010-06-30 13:28:23)

Cdh · 2010-06-30 16:02:04

Is 2.11.901 not the beta for 2.12?

The 945gme in eee 1000h feels much better with 2D with this driver.

gabe_ · 2010-06-30 17:16:45

Upon further research, it appears as though you are correct.

If anyone needs me, I'll be over here sticking my foot in my mouth.

cb474 · 2010-07-02 01:57:28

bboston7 wrote:

I fixed this by downgrading xf86-video-intel, libgl, intel-dri, mesa, and xorg-server. I'll attach my pacman.log with the relevant information for anything I might have forgotten.

If I do this, does that mean that my xorg.conf.d settings won't work and I have to go back to using the hal/fdi/policy settings instead? Thanks.

I've been getting weird issues where the 2d image on the screen splits down the middle with the left on the right and the right on the left. And xorg just crashed when I clicked on a link in Firefox.

petelewis · 2010-07-08 08:11:56

Has anyone else tried using the latest packages but with kernel26-lts (currently 2.6.32)?

For me, though it's a little slower, I haven't had any lock-ups yet. It would be good to know if this is the same for other people.

cb474 · 2010-07-12 05:22:41

I tried used the older packages with kernel26-lts and didn't get any lockups, but I still got weird 2d screen splitting effects sometimes (strangely only when I'm using the Gnome-Terminal program). I haven't tried it yet with the latest packages.

fijam · 2010-07-14 15:53:11

Has anyone managed to fix that? The problem still persists for me with xf86-video-intel 2.12

zhuk_av · 2010-07-15 06:17:42

I'm using Arch on Acer Acpire One 150 netbook with Intel 945 GME. I have all the same issues with 3D, but no locks. In fullscreen glxgears shows 60 FPS, in windowed mode - about 20. But if you would constantly move mouse aroud while running windowed glxgears, FPS grows up to 60.

xorg.conf on my system describes only touchpad configuration.

petelewis · 2010-07-19 14:10:18

I've just had a case of the freeze I described earlier (everything seems to be still working except that the screen doesn't update other than the mouse) with kernel26-lts 2.6.32.16 and everything else from the stable Arch repos (xf86-video-intel 2.12.0, intel-dri 7.8.2). The card is a 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03).

I've actually had a very stable compositing desktop in KDE with this set up for a couple of weeks now, but I can trigger this quite reliably. Specifically, if I enable compositing in KWin and then try to change the colours in a colour scheme, clicking apply will trigger the freeze. With compositing disabled I can change colours fine.

Pete.

kokoko3k · 2010-09-24 11:36:33

Did anyone tried with the latest packages (right now we have xorg-server 1.8.1.902-1, kernel 26 2.6.35.5 and xf86-video-intel 2.12.0) ?

For performance reasons i'm still stuck with kernel 2.6.33, last time i tried 2.6.35 from testing it all was slower and i'm really tired of trying again and again

Last edited by kokoko3k (2010-09-24 11:37:34)

petelewis · 2010-09-24 12:59:42

Well, I've been using the latest versions of everything from [extra] and everything's been stable for at least a few weeks now. The original problem seems to have gone away for me at least.

The issue of freezing when colour or themes are changed in KDE and compositing is enabled still persists, and that's a known bug that's being hotly debated on the KDE bug tracker. (And there's the simple workaround of just hitting alt-shift-f12 twice).

kokoko3k · 2010-09-24 14:09:02

I just tried 2.6.35 again and the system stability seems ok.
Sauerbraten runs at right speed too (or it seems so).

The bad things:
------------------------
compiz in *direct* rendering mode has still redrawing issues (garbage on the screen), but i don't care much, i can live with indirect rendering.
What worries me is that starting from kernel 2.6.34 (and still with 2.6.35 from core), compiz runs ok only when i keep my finger on the touchpad or keep pressing and releasing any keyboard button.
As soon as i put my finger on, compiz benchmark jumps from 40..45 to 60 (vsynced)
similar issues with glxgears (about same numbers as above).

I even tried to enable testing and upgrade xorg-server, intel driver (and related packages like mesa, libgl xf86-input* and so on), but i've had no success at all.

I wonder if anyone has a solution, or if i should fill a bug report.
Could it be a kernel issue (i've had no such problems with nvidia card), an xorg issue, xinput maybe?

I'm a bit lost as time is passing by and this stupid bug persists!

Last edited by kokoko3k (2010-09-24 14:12:16)

kokoko3k · 2010-09-24 17:05:16

Found this:
https://bugzilla.redhat.com/show_bug.cgi?id=582861
appartently the bug was about low 3d performance when the cpu is idle, this was until may 2010, next it became "low 3d performance when input is idle "
Just quoting myself there:

Teapot demo from mesa exhibits the exact same behaviour as glxgears.
just launching it under linux-2.6.33 gives stable 30fps (my refresh rate
halved,60/2,vsync issue, but it's ok).
launching it under linux-2.6.35 gives 14~15fps (60/4), and guess what... just
touching the touchpad makes the magic: 30fps.

-EDIT-
Ok, let's hope for the best:
https://bugs.freedesktop.org/show_bug.cgi?id=30364

Last edited by kokoko3k (2010-09-24 17:29:42)

jass0 · 2010-09-27 22:14:40

Kernel 2.6.35.6-1 seems to have fixed the garbled fonts!

edit. nvm, it didn't, but the problem appears less frequently at least on my system...

Last edited by jass0 (2010-09-28 22:22:58)

Arch Linux

#26 2010-06-24 08:37:38

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#27 2010-06-29 16:31:00

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#28 2010-06-29 23:35:28

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#29 2010-06-30 04:45:31

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#30 2010-06-30 13:06:20

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#31 2010-06-30 16:02:04

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#32 2010-06-30 17:16:45

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#33 2010-07-02 01:57:28

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#34 2010-07-08 08:11:56

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#35 2010-07-12 05:22:41

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#36 2010-07-14 15:53:11

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#37 2010-07-15 06:17:42

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#38 2010-07-19 14:10:18

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#39 2010-09-24 11:36:33

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#40 2010-09-24 12:59:42

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#41 2010-09-24 14:09:02

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#42 2010-09-24 17:05:16

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

#43 2010-09-27 22:14:40

Re: 2.6.34 and the latest intel drivers for gm945 (Eeepc 1000h) epic fail

Board footer