[fglrx] very poor 3D performance of fglrx

Bug #202712 reported by Olivier
2
Affects Status Importance Assigned to Milestone
linux-restricted-modules-2.6.24 (Ubuntu)
Invalid
Undecided
Unassigned

Bug Description

I am using the restricted driver for my ATI X600 mobility and the 3D performance is really poor under hardy.

Direct Rendering is activated. See glxinfo for details.

As an example, I get 60 fps with glxgears while using the radeon driver, I get 1600 fps...

The only error I found in the logs is when I type dmesg

[10829.940186] [fglrx:firegl_lock] *ERROR* Process 10776 is using illegal context 0x00000007

I will be happy to help track that issue

Revision history for this message
Olivier (olivier-lacroix) wrote :
Revision history for this message
Olivier (olivier-lacroix) wrote :
Revision history for this message
Olivier (olivier-lacroix) wrote :
Revision history for this message
lod (altoas) wrote :

i had the same issue, and it was fixed when I installed clean alfa (not trough update-manager). perhaps there's other way to solve the problem for you...

Revision history for this message
Olivier (olivier-lacroix) wrote :

Thanks for your answer. However, I would be very glad to get another method to make it work without the hassle of reinstalling everything. I mean, there are no big errors here, why are the performances that low ?

Revision history for this message
Bryce Harrington (bryce) wrote :

I see you tested with "EXA" at some point - did that make any difference?
(Fwiw, I think "greedy" heuristic is irrelevant for fglrx)

In general it sounds like your system is getting software rendering rather than hardware. Not clear from the logs why that might be. Please post a log with EXA turned on for comparison.

Changed in linux-restricted-modules-2.6.24:
status: New → Incomplete
Revision history for this message
Bryce Harrington (bryce) wrote :

Maybe also look at bug 206337, which also reports performance issues.

Revision history for this message
Olivier (olivier-lacroix) wrote :

Thanks Bryce for looking into that issue.

I never tried using fglrx with EXA. the commented part is just a left-over of the time when I was using the radeon driver.

Here is the log with EXA enabled. it seems that XAA is still used. Weird. The same performance issue happens.

I also think that software rendering is used. glxgears eats 80% of the CPU when running. If this is true, why does glxinfo report direct rendering as enabled?

Looking at bug 206337, and using vmstat, I do not experience the same symptoms.

Revision history for this message
Bryce Harrington (bryce) wrote :

Okay, so that answers the question about if -fglrx supports EXA. Guess not.

Another thing to try I found in googling around a bit, please try adding:

Section "ServerFlags"
Option "AIGLX" "off"
EndSection

to your xorg.conf

Apparently AIGLX, which isn't supported on fglrx, kicks into software rendering mode when it finds itself active on a system that doesn't support what it needs for HW rendering.

Revision history for this message
Bryce Harrington (bryce) wrote :

[This is an automated message]

As of Intrepid (8.10), we have a dedicated package 'fglrx-installer' for fglrx bugs, which now includes a process for upstreaming bugs to AMD.

  http://bugs.launchpad.net/ubuntu/+source/fglrx-installer

To transition your bug into the new fglrx-package, we need your help. Please do the following:

 a. Verify the bug occurs in Intrepid.
     (Intrepid ISOs: http://cdimage.ubuntu.com)
 b. If you haven't already, please include in the bug:
     * Your /var/log/Xorg.0.log
     * The output of `lspci -vvnn`
     * Steps to reproduce the issue
 c. Click 'Also affects distribution'
 d. Set 'Source Package Name' to 'fglrx-installer'
 e. Click Continue

Thank you. This will assist us in reviewing and upstreaming your fglrx bug, as appropriate.

[We'll expire the fglrx bugs in l-r-m-* in a month or so.]

Revision history for this message
Olivier (olivier-lacroix) wrote :

I have switched to Archlinux; and as a result, won't be able to do any further test.

feel free to close this bug.

Revision history for this message
Jonathan Thomas (echidnaman) wrote :

Closing then.

Changed in linux-restricted-modules-2.6.24:
status: Incomplete → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.