KMS & ATI & kernel 2.6.31.2 - slow performance

TigTex · 2009-10-06 16:15:24

Today I made as usual a pacman -Syu. It updated my kernel to version 2.6.31.2. So far so good. I rebooted my system (it's a mediacenter with an ati radeon 9250se, ati opensource drivers) and I noticed that when it loads UDEV modules it changes my screen resolution (it enables kms...). I can read it on dmesg.
So.. I have kms enabled by default. OK. Tried to open xbmc and nothing appens. A black screen and full cpu usage... hummm. tried with lxde. It opens just fine.
I decided to run glxgears to check video performance... and imagine: 160fps (10 times lower than usual). Games run now at 1 or 2 fps. DRI is enabled. XBMC (after deleting the .xbmc folder) runs SLOW SLOW SLOW, 3-4fps just moving the mouse...

I need to disable KMS because it made my mediacenter useless (unable to play video...). Or at least I need to get my performance back!!

Any ideas?

Everything was OK before updating. (2.6.31.1 -> 2.6.31.2)

EDIT: the temporary solution is to add nomodeset to your menu.lst (kernel line)

Last edited by TigTex (2009-10-06 17:41:51)

wonder · 2009-10-06 16:25:02

i think the ati video driver need to be updated too.

TigTex · 2009-10-06 16:34:04

humm. Look's like a long time solution... I need a fast one Probably I have to recompile the kernel without kms...

Beini · 2009-10-06 17:00:39

Isn't there a setting, something like modeset=0 you can set to the kernel line in menu.lst or to modprobe.conf?

Last edited by Beini (2009-10-06 17:01:30)

vit · 2009-10-06 17:00:46

You should try new driver (I think git version is the best). It will be very fast when finished

TigTex · 2009-10-06 17:04:34

I will try both things you said. Lets see...

Beini · 2009-10-06 17:07:59

TigTex wrote:

I will try both things you said. Lets see...

http://bugs.archlinux.org/task/16149

Tells you how to disable it.

Army · 2009-10-06 17:08:43

I'm quite sure that in the mailing lists someone said that there will be newer versions of the ati drivers in testing. So I'd say, either wait, or use the git snapshot versions in the AUR

wonder · 2009-10-06 17:22:21

http://mailman.archlinux.org/pipermail/ … 13773.html

TigTex · 2009-10-06 17:26:38

Beini wrote:

TigTex wrote:
I will try both things you said. Lets see...
http://bugs.archlinux.org/task/16149
Tells you how to disable it.

Yep, "nomodeset" in menu.lst solved the problem for now. Thanks :D But this will happen to more people with ati gpu's...

Do you think that a simple linux pc dedicated for xbmc (mediacenter) would be better with or without KMS? I already read a lot about KMS but the main conclusion that I take is that KMS makes the GPU slower

Last edited by TigTex (2009-10-06 17:39:56)

mutlu_inek · 2009-10-06 19:44:14

FYI: The new ati video driver was just released to [extra].

Army · 2009-10-06 20:43:04

wonder wrote:

http://mailman.archlinux.org/pipermail/ … 13773.html

With this my system doesn't boot anymore, it stucks at udev. It only works if I leave MODULES in the mkinitcpio.conf empty. Strange..

wonder · 2009-10-06 20:45:02

@army better bring it to arch-general ml and give information about your system/setup.

Army · 2009-10-06 22:54:28

done, let's see what we get here

Mad Max · 2009-10-07 00:41:59

I confirm the problem, - I use current kernel26 2.6.31.2-1 with up to date xf86-video-ati-git on my ATI mobility X2300, and my compositing in KDE is gone fishing. The glxgears FPS drop even underlines the dramatical effect: ~8900 frames/5 sec (@ previous kernel; same driver git) -> down to 1945 frames/5 sec (current kernel ^ ; still same driver git).

To wait for a new driver just, when kernel gets it minor update, is NOT the user friendly solution! Epic linux usability fail on desktop, since similar situation occurs periodically frequent, just every time in some other place.

Last edited by Mad Max (2009-10-07 02:01:57)

ngoonee · 2009-10-07 01:20:30

Mad Max wrote:

I confirm the problem, - I use current kernel26 2.6.31.2-1 with up to date xf86-video-ati-git on my ATI mobility X2300, and my compositing in KDE is gone fishing, Glxgears FPS drop even underlines the dramatical effect: ~8900 frames/5 sec (@ previous kernel; same driver git) -> down to 1945 frames/5 sec (current kernel ^ ; still same driver git).
Wait for a new driver just, when kernel gets it minor update, is NOT the user friendly solution! Epic linux usability fail on desktop, similar situation occurs periodically frequent, just every time in some other place.

Sigh, I don't think Arch Linux has ever been advertised as 'the user friendly solution'.

Things break. Relax, find out why, fix it if you can, downgrade if you can't, but remember the bug-report.

Mad Max · 2009-10-07 01:32:24

ngoonee wrote:

Sigh, I don't think Arch Linux has ever been advertised as 'the user friendly solution'.
Things break. Relax, find out why, fix it if you can, downgrade if you can't, but remember the bug-report.

What is it then? Masochist friendly? I claim to know my system (at least the stuff I need to care of), so once pacman -Syu, and you can watch your system's pieces fly apart, like those pigs do...

To tell the truth, I hate it most: always to worry, if you gonna survive next kernel update...

Last edited by Mad Max (2009-10-07 02:03:55)

mutlu_inek · 2009-10-07 03:08:09

Mad Max wrote:

What is it then? Masochist friendly? I claim to know my system (at least the stuff I need to care of), so once pacman -Syu, and you can watch your system's pieces fly apart, like those pigs do...
To tell the truth, I hate it most: always to worry, if you gonna survive next kernel update...

Seriously? Then don't update. Especially not to a kernel in [testing]. You could wait before updating even after the kernel comes to [core] and see whether other users report problems. Especially so if you know that basic things are moving in the graphics subsystem. Or, just use a distribution which tests stuff and releases it in a state that works for most users. You seem to misunderstand how Arch works.

MP2E · 2009-10-07 03:27:57

mutlu_inek wrote:

Mad Max wrote:
What is it then? Masochist friendly? I claim to know my system (at least the stuff I need to care of), so once pacman -Syu, and you can watch your system's pieces fly apart, like those pigs do...
To tell the truth, I hate it most: always to worry, if you gonna survive next kernel update...
Seriously? Then don't update. Especially not to a kernel in [testing]. You could wait before updating even after the kernel comes to [core] and see whether other users report problems. Especially so if you know that basic things are moving in the graphics subsystem. Or, just use a distribution which tests stuff and releases it in a state that works for most users. You seem to misunderstand how Arch works.

I agree. If you're talking about being "afraid" of switching kernels, and complaining about User-Friendly ness when not only are you using a distribution that is NOT centered around these, but you are using the [testing] repo and freaking out when you find bugs, then there is something wrong. As someone in another thread put: "Arch is only as stable as you are". If you want Arch to be stable, do not use the testing repo and wait for others to test out the packages when they arrive in [core] (even waiting a day or 2 would suffice).

ngoonee · 2009-10-07 04:41:26

Mad Max wrote:

ngoonee wrote:
Sigh, I don't think Arch Linux has ever been advertised as 'the user friendly solution'.
Things break. Relax, find out why, fix it if you can, downgrade if you can't, but remember the bug-report.
What is it then? Masochist friendly? I claim to know my system (at least the stuff I need to care of), so once pacman -Syu, and you can watch your system's pieces fly apart, like those pigs do...
To tell the truth, I hate it most: always to worry, if you gonna survive next kernel update...

Claims are easy. Proof isn't.

Please read http://wiki.archlinux.org/index.php/The … er-centric. If you haven't before, that would explain your attitude and expectations, and hopefully those will change. If you have... please re-read until you get it.

Army · 2009-10-07 08:51:18

Mad Max wrote:

always to worry, if you gonna survive next kernel update...

There is a kernel26-lts (long term support), maybe that's perfect for you

Army · 2009-10-07 22:24:30

Army wrote:

wonder wrote:
http://mailman.archlinux.org/pipermail/ … 13773.html
With this my system doesn't boot anymore, it stucks at udev. It only works if I leave MODULES in the mkinitcpio.conf empty. Strange..

Got it, I had to put intel_agp in there as well

AndyRTR · 2009-10-08 04:21:24

somebody should write a wiki page like we have for nouveau and intel drivers...

Mad Max · 2009-10-08 11:34:15

mutlu_inek wrote:

Seriously? Then don't update. Especially not to a kernel in [testing]. You could wait before updating even after the kernel comes to [core] and see whether other users report problems. Especially so if you know that basic things are moving in the graphics subsystem. Or, just use a distribution which tests stuff and releases it in a state that works for most users. You seem to misunderstand how Arch works.

MP2E wrote:

I agree. If you're talking about being "afraid" of switching kernels, and complaining about User-Friendly ness when not only are you using a distribution that is NOT centered around these, but you are using the [testing] repo and freaking out when you find bugs, then there is something wrong. As someone in another thread put: "Arch is only as stable as you are". If you want Arch to be stable, do not use the testing repo and wait for others to test out the packages when they arrive in [core] (even waiting a day or 2 would suffice).

1. Don't tell me what to do, and I am not "afraid", just annoyed.
2. I'm using [testing] because of some updated packages I need, so what? Following your logic, I shouldn't use it, because it's [testing]. And when I need some packages available only there atm? Compile and MAKEPKG myself? NO thanks, my time is precious.
I can only repeat, it is NOT a major kernel update, it is still 2.6.31 but 2.6.31.2-1. This should not break my 3-D acceleration, doesn't matter, what your usability point of view is. It can even happen, when the kernel reaches [core] too, if those drivers aren't ready yet.

ngoonee wrote:

Claims are easy. Proof isn't.
Please read http://wiki.archlinux.org/index.php/The … r-centric. If you haven't before, that would explain your attitude and expectations, and hopefully those will change. If you have... please re-read until you get it.

Please don't come with your RTFM. We are not in school here, so spare me your "re-read until you get it". What makes you think, I didn't understand "The Arch Way"? What proofs do you need? I am just angry about a concrete problem: my 3D acceleration is gone with a minor kernel update.

Concerning my statement of "not surviving next kernel update" - I have had a lot of such "surprising" experiences in the past, when after kernel update your X-server fails to start, or some hotkey buttons on my laptop failed to work as they did before, or suspend/hibernate issues, or many other issues collected over several years. This happens on every Linux distro, even Ubuntu, so please don't come with "you don't understand, how it works". Main reason for such issues is, because it's developed by community and sometimes community doesn't give a damn fuck, how far other projects are, - but hey, they've got their release.

SCNR

Last edited by Mad Max (2009-10-08 11:36:26)

ngoonee · 2009-10-08 14:41:45

Mad Max wrote:

1. Don't tell me what to do, and I am not "afraid", just annoyed.

Most would read the word 'worry' to be closer to afraid than annoyed. Just an FYI.

Mad Max wrote:

2. I'm using [testing] because of some updated packages I need, so what? Following your logic, I shouldn't use it, because it's [testing]. And when I need some packages available only there atm? Compile and MAKEPKG myself? NO thanks, my time is precious.
I can only repeat, it is NOT a major kernel update, it is still 2.6.31 but 2.6.31.2-1. This should not break my 3-D acceleration, doesn't matter, what your usability point of view is. It can even happen, when the kernel reaches [core] too, if those drivers aren't ready yet.

If you're using [testing], then don't complain when something breaks. If you need some packages, then either:-
a) do your own makepkg (not attractive due to time constraints, if you're that busy)
b) only install some packages from [testing] (doesn't always work)
c) use [testing] and accept that things will break sometimes. People make mistakes. The POINT of testing is for interested users to be able to test things out and point out 'hey this is broken'. Preferably on the bug tracker or ML, not with some random forum rant.

Mad Max wrote:

Please don't come with your RTFM. We are not in school here, so spare me your "re-read until you get it". What makes you think, I didn't understand "The Arch Way"? What proofs do you need? I am just angry about a concrete problem: my 3D acceleration is gone with a minor kernel update.

Well, forgive me for assuming that the very fact that you're pissed that something has broken when you're using [testing] means you don't understand what was in the wiki. Either that or you don't care. Either way, take some responsibility for your system and choices, you're not paying any money for the priviledge of using this distro.

Mad Max wrote:

Concerning my statement of "not surviving next kernel update" - I have had a lot of such "surprising" experiences in the past, when after kernel update your X-server fails to start, or some hotkey buttons on my laptop failed to work as they did before, or suspend/hibernate issues, or many other issues collected over several years. This happens on every Linux distro, even Ubuntu, so please don't come with "you don't understand, how it works". Main reason for such issues is, because it's developed by community and sometimes community doesn't give a damn fuck, how far other projects are, - but hey, they've got their release.

Updating critical component of your system could lead to breakage. Surprising, isn't it? Why don't these volunteer devs make doubly sure that MY system works fine with my specific combination of X, Y, and Z packages. I'm surely paying them more than enough for the time and effort spent. I demand my rights!

Mad Max wrote:

SCNR

Neither could I =p

Arch Linux

#1 2009-10-06 16:15:24

KMS & ATI & kernel 2.6.31.2 - slow performance

#2 2009-10-06 16:25:02

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#3 2009-10-06 16:34:04

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#4 2009-10-06 17:00:39

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#5 2009-10-06 17:00:46

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#6 2009-10-06 17:04:34

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#7 2009-10-06 17:07:59

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#8 2009-10-06 17:08:43

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#9 2009-10-06 17:22:21

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#10 2009-10-06 17:26:38

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#11 2009-10-06 19:44:14

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#12 2009-10-06 20:43:04

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#13 2009-10-06 20:45:02

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#14 2009-10-06 22:54:28

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#15 2009-10-07 00:41:59

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#16 2009-10-07 01:20:30

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#17 2009-10-07 01:32:24

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#18 2009-10-07 03:08:09

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#19 2009-10-07 03:27:57

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#20 2009-10-07 04:41:26

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#21 2009-10-07 08:51:18

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#22 2009-10-07 22:24:30

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#23 2009-10-08 04:21:24

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#24 2009-10-08 11:34:15

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

#25 2009-10-08 14:41:45

Re: KMS & ATI & kernel 2.6.31.2 - slow performance

Board footer