Intel x3100 enormous regression in ubuntu 9.10 beta

Bug #451523 reported by GT
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
xorg (Ubuntu)
Invalid
Undecided
Unassigned

Bug Description

Binary package hint: xorg

On my Laptop, I am using Ubuntu 9.04 and the 3D performance of the intel x3100 is quite low (compared to older Ubuntu versions).
Therefore, I checked the 3D performance with the Ubuntu 9.10 beta Live CD (I expected the 3D performance to be better in the new version).

However, apparently the 3D performance of the INTEL X3100 dropped enormously from 9.04 to 9.10. As indicator I was using glxgears, which was
always over 4000 frames in 5sec in 9.04 and now is
never over 2000 frames in 5sec in 9.10.

Apparently there is an extreme performance regression?! Or is there any reason why the measurement using glxgears does not make sense?

Thank you
Aldi

ProblemType: Bug
Architecture: i386
Date: Wed Oct 14 18:19:08 2009
DistroRelease: Ubuntu 9.10
LiveMediaBuild: Ubuntu 9.10 "Karmic Koala" - Beta i386 (20090929.2)
MachineType: Acer, inc. TravelMate 6292
Package: xorg 1:7.4+3ubuntu5
PccardctlIdent:
 Socket 0:
   no product info available
PccardctlStatus:
 Socket 0:
   no card
ProcCmdLine: BOOT_IMAGE=/casper/vmlinuz file=/cdrom/preseed/hostname.seed boot=casper initrd=/casper/initrd.lz quiet splash -- debian-installer/language=de console-setup/layoutcode?=de
ProcEnviron:
 LANG=de_DE.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.31-11.36-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.6.0~git20090817.7c422387-0ubuntu6
 libdrm2 2.4.13-1ubuntu1
 xserver-xorg-video-intel 2:2.8.1-1ubuntu2
 xserver-xorg-video-ati 1:6.12.99+git20090825.fc74e119-0ubuntu2
SourcePackage: xorg
Uname: Linux 2.6.31-11-generic i686
XorgConf: Error: [Errno 2] No such file or directory: '/etc/X11/xorg.conf'
XorgLogOld: Error: [Errno 2] No such file or directory: '/var/log/Xorg.0.log.old'
dmi.bios.date: 03/03/2008
dmi.bios.vendor: Acer
dmi.bios.version: v1.3601
dmi.board.name: Victoria
dmi.board.vendor: Acer, Inc.
dmi.board.version: Not Applicable
dmi.chassis.type: 1
dmi.chassis.vendor: Acer, Inc.
dmi.chassis.version: N/A
dmi.modalias: dmi:bvnAcer:bvrv1.3601:bd03/03/2008:svnAcer,inc.:pnTravelMate6292:pvrNotApplicable:rvnAcer,Inc.:rnVictoria:rvrNotApplicable:cvnAcer,Inc.:ct1:cvrN/A:
dmi.product.name: TravelMate 6292
dmi.product.version: Not Applicable
dmi.sys.vendor: Acer, inc.
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.31-11-generic

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

glxgears is not a valid benchmark.

Changed in xorg (Ubuntu):
status: New → Invalid
Revision history for this message
ErMejo (andrea-lombardoni) wrote :

See this link for more information http://wiki.cchtml.com/index.php/Glxgears_is_not_a_Benchmark

Basically glxgears is not a good benchmark.

In order to compare the graphic performance of two releases, you need at least to compare results from something such as http://www.phoronix-test-suite.com/

Revision history for this message
Chris D (cdekter) wrote :

gxlgears may well not be a good benchmark, but the performance HAS dropped significantly. Just compare 9.10 to 8.10, they are lightyears apart in responsiveness. I'm going to run the prescribed Phoronix benchmarks and report back here - it is unacceptable for this kind of performance issue to go ignored. Karmic is AFAIC unusable because of this.

Revision history for this message
Mike Richards (mrmikerich) wrote :

I'm seeing the same thing here. There is a severe video slowdown with a fresh 9.10 install on a T61 with Intel X3100 graphics. On 9.04 I consistently get 4000+ fps; with 9.10 it's just over 2000 fps. glxgears may not be a proper benchmark, but clearly there is something wrong here. There are real-world consequences to this bug, not just glxgears; for example, running a virtual XP under kvm/qemu is *extremely* jerky in 9.10 -- to the point of being unusable -- whereas it is completely smooth on my 9.04 system.

You don't really see the slowdown too much if you're just running standard desktop apps, which may be why more people haven't reported this, but this bug is hardly "invalid".

I'll run the Phoronix tests on 9.04 and 9.10 and report back with results. Can this bug be changed from "invalid", or should I submit a new report?

Revision history for this message
Mike Richards (mrmikerich) wrote :

Er, nevermind that comment of mine... Turns out it the general problem was with pulseaudio, not the Intel driver. Aside from the odd glxgears framerate slowdown, everything is fine, maybe even a bit faster than in 9.04.

Revision history for this message
Chris D (cdekter) wrote :

What did you do to resolve the problem re. pulseaudio?

Revision history for this message
Mike Richards (mrmikerich) wrote :

It was app-specific in my case. For example, kvm/qemu was configured to use something other than pulseaudio, and was fighting with pulseaudio, causing the video output to be jerky as it used 100% cpu trying to play sound. Once I reconfigured the errant apps to talk to pulseaudio instead of whatever they were trying to talk to, things were smooth.

In all fairness, the problem wasn't even pulseaudio itself, but rather misconfigured apps. But I never had problems before pulseaudio, and there was recently a Slashdot article bashing pulseaudio, so that's where the blame went...

I never did get the Phoronix test running; it bombed out with some kind of gtk error. But I convinced myself that it wasn't the Intel driver at fault by comparing extremetuxracer framerates. It was significantly faster in 9.10 -- double the framerate at 800x600 (60fps vs 30fps), and about 20% faster at 1680x1050 (30fps vs 25fps).

If you want a test that doesn't depend on sound or require installing extra software, look at your screensaver demos. They don't interface with the soundcard at all -- just 2D and 3D graphics. If the screensavers look like they are running at the same speed under both your systems, then it's probably not the video driver at fault. That should have been my tip-off to look elsewhere, but I ignored this data point and still blamed the video driver.

One other thing that may or may not apply in your case: While trying to solve this problem, I read about some people who had trouble if they had an Intel and Radeon card in the same system (Intel on motherboard and Radeon card). If they installed the radeon proprietary driver, then I think the Intel card would get slow or broken. It didn't apply to me though, so I didn't read it too closely.

Revision history for this message
Chris D (cdekter) wrote :

I just ran Karmic on the system I had problems with, and it ran just fine. There is a little bit of sluggishness in some Compiz animations, mainly around events that involve mapping a window (open, unminimize, etc). Otherwise, it seems quite good. I have no idea how this is possible, the only thing I can think of is that the least time I installed the RC and then updated after the release, and this time I used a clean install of the final release version.

Revision history for this message
Uri Shabtay (uri.shabtay) wrote :

there is still a lot of work to be done within the responsiveness of the effects. it is horrifying.
enclosed herewith is the AWN issue (https://bugs.launchpad.net/awn/+bug/478790) -- which is important in realising what influences the Intel drivers..

i truly hope this will be resolved in Lucid..

cheers for now

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.