Intel graphics performance regression in 2.6.32-19 lucid kernel update (was: Firefox Slows Down Compiz)

I feel this is affecting me as well, but I think it did not earlier, just a week or two ago in lucid. I wonder if some of the kernel intel changes have caused this major performance regression or what. Too bad the older kernels are not available anymore!

Is your lucid a very new installation or have you been using lucid for some time already? I wonder if eg. http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34-rc3-lucid/ should be tested.

For the record, I have GM965 ie. GMA X3100.

I managed to salvage 2.6.32-16 kernel from beta1 alternate amd64 CD. Using that kernel the problems seems to be away.

I cannot find 2.6.32-17 or 2.6.32-18 so cannot test those. I think that at least 2.6.32-17 was ok, at least, possibly -18 as well. I don't currently remember when I upgraded/rebooted to which kernel.

there is an Intel graphics performance regression every release... please please please don't drop security support for hardy

After a few days of using -16 kernel and 2.6.34 kernel and installing all the lucid updates, I cannot currently reproduce the problem when going back to -19. Putting the bug back to NEW. Previously it was quite immediately seen as slower behavior in Firefox and Ubuntu Software Center.

Meanwhile I have tweaked a couple of stuff like removed hal/hal-info packages which are not needed anymore on lucid, and I'm not sure if those, in addition to the updates, have anything to do with the problem disappearing.

It's not to say that Intel wouldn't be relatively slow in lucid (and before) - it's not blazingly fast and I'm not sure if there is some specific part to blame or not. For example compiz cube rotation is easily a bit jerky.

(another thing to note is that Ubuntu Software Center seems somewhat sluggish even in this "normal" situation)

I have noticed this as well on two of the three Intel-graphics laptops that I have to test with. On both an IBM ThinkPad X40 (Intel 855gm, I believe) and an HP Compaq 2710p tablet (Intel GMA x3100, apparently) the Ubuntu Netbook Remix UI is unusably sluggish with 2.6.32-19 and is fine with 2.6.32-18.

I also have an Intel Atom/GMA950 HP Mini. It is fine with both kernels. This problem also existed in an earlier Alpha kernel and was fixed in time for beta 1. Please let me know if I can provide any more information.

Doh. This is so darn complicated. I now measured similar performance of gtkperf with 2.6.32-16 and 2.6.32.20. However, performance drops after some usage.

Could you, experiencing slowdowns, add CPU clock speed panel application to GNOME panel? I just noticed that the ondemand speed selection somehow kept my CPU at lower speed than it should even when under 100% CPU load! Now everything is rocket speed when I forced this up to.2.5GHz.

It might not be a intel graphics issue then, at least in my case, but a processor speed issue.

Relevant bug reports bug #519142 and bug #555315. I wonder if this is related to the fact that I've set the governor manually with the GNOME applet somehow, similar to bug #519142, which causes the bug to appear. If you can also confirm this in addition to the problem being with the governor, it'd be quite beneficial to know what goes wrong when the governor has been once set manually, even after setting it back to ondemand.

As for the kernel differences, they might be indeed explained by the fact that the governor might set the clock speed higher after boot, and decrease it later on, so depending on how soon you do comparison after a reboot, you might think it was the change of the kernel causing the speedup.

I have a problem with Compiz as well. If it is enabled it just makes everything go so slow the desktop is unusable. As soon as you disable Compiz the system works just fine.

All was working perfectly on 2.6.32-16, then 2.6.32-17 introduced the issue. 2.6.32-18 the issue was resolved. But the -19 and -20 versions are still having Compiz issues as per slow and unusable desktop.

The machine is question is running 64 bit version that was originally installed from a Alpha 2 iso. The laptop itself is not super fast and is powered by a Celeron 540 and Intel X3100 graphics with 2Gb DDR2 RAM. But was running perfectly on the initial Alpha 2 install.

I have just updated to the 2.6.32-21 tonight and the slow performance is still there with this update.

Hi Richard,

Good analyzing, the latest kernel working properly without the slow troubles is 2.6.32-18.

I have also tested like you the latest 2.6.32-21, the bug occur :-(

So, after the 2.6.32-18, the bug is here and not fixed ...

A++

Also, I have omitted to say I'm running 32 bits version, so, it not only occur on the 64 bits ...

A++

Hello,

I wonder if this is related to what I have seen on my Dell Vostro laptop with Intel GM965 graphics card. Since
kernel 2.6.32-19, whenever I startup my laptop, Xorg is stuck on 50% of cpu regardless of what I am doing,
I have also seen this on kernel versions -20 and -21. I have tried restarting gdm but this does not cure
the problem. This wasn't a problem on the pre -19 kernels.

Funny thing is, I have another Dell laptop (latitude D630) with a similar Intel GM965 graphics card and that is
not affected by this problem at all.

Let me know if you need further information about my installation.

Best regards,
  Pankaj

Hi Virgil,

If you could also please test the latest upstream kernel available that would be great. It will allow additional upstream developers to examine the issue. Refer to https://wiki.ubuntu.com/KernelMainlineBuilds . Once you've tested the upstream kernel, please remove the 'needs-upstream-testing' tag. This can be done by clicking on the yellow pencil icon next to the tag located at the bottom of the bug description and deleting the 'needs-upstream-testing' text. Please let us know your results.

Thanks in advance.

    [This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

Still occurs using mainline kernel 2.6.34rc5. It only seems to slow compiz when firefox is up.

One vote for me, compiz makes the system unusably slow. MacBook Santa Rosa with Intel 965G, too bad :/

Any progress here? We have no hardware acceleration...

Rendering is taking several seconds per frame here, with GMA965, even with the plain Gnome desktop (no Firefox required) and basic desktop effects enabled. Without desktop effects the desktop and OpenGL both work fine. Linux Valen 2.6.32-22-generic #33-Ubuntu SMP Wed Apr 28 13:28:05 UTC 2010 x86_64 GNU/Linux

So it looks like the problem appeared in the .18->.19 release, and there aren't many patches in there that touched the i915 driver. I'll investigate and perhaps build a test kernel that backs out the patch(es).

There was a patch that went in to disable FBC for 915GM and 945GM devices, to prevent some suspend/resum problems.

That was for launchpad bugs 492392 and 539609.

I can't see any way -18 could be ok and -19 bring up this issue at all at least from the drm side. Virgil, can you please try a few mainline kernels without updating any userspace components in between and see if they can reproduce it?

This one should not exhibit the slowdown problem if it is really a problem in the kernel drm:
http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.33.1-lucid/

And additionally If this one works, it would point the finger at a ubuntu specific patch:
http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.32.12.3-lucid/

Thank you for the help!

Using Mainline 2.6.33.1 as linked above.

In this version performance seems to be fairly low, but firefox makes no difference to performance. Screenshots of what I mean included.

I can verify similar behavior and performance now on Lucid standard 2.6.32-21. It is the same as the mainline.

Ok,

Here some new informations :

- My graphic card reported by lspci is : 00:02.0 VGA compatible controller: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c)
- I have installed Lucid Final on another disk.
- I have applied all updates available.
- The kernel running is 'linux-image-2.6.32-22-generic-pae'.
- I have set the theme to 'Dust'.
- I have activated the tweak 'have_icons' :
    - /desktop/gnome/interface/menus_have_icons
    - /desktop/gnome/interface/buttons_have_icons
- I have activated the tweak 'button_layout' :
    - /apps/metacity/general/button_layout
    - I set it to : menu:minimize,maximize,close

After that I go to System -> Preferences -> Appearance -> Visual effects, and select 'Extra' (it was already set to 'Normal').
At this stage all things seems working properly, compiz working with defaults effects from the 'Dust' theme, etc ...

Now, I restore all my compiz parameters I have saved under Karmic (see file in attachment').

Bingo !!! Compiz gona slow down and render the desktop unusable (seem to refresh screen every 15-20 secs).

It's sure the problem come from the kernel, because under my Karmic, I have also backported kernel from lucid, and I'm actually forced to stick with the 'linux-image-2.6.32-18-generic-pae' because if I try every versions >= '-19', then same problem occur.

I have not already tested 'linux-image-2.6.32-18-generic-pae' under Lucid to see if it fix the problem, but I'm pretty confident it act like the same as under my Karmic environment with kernel backported ...

Hope it help ...

BINGO !!!

Since my post #23, here is what I have made :

- I have installed 'linux-image-2.6.32-18-generic-pae'.

Then rebooted to this kernel and load my compiz preferences from the file in post #23.

ALL THINGS WORK FINE RIGHT NOW !!!

For all developers in doubt, I have made a screencast you can see here : http://bigbob.fun.free.fr/Ubuntu-2.6.32-19-regression.m4v

So, please, can someone say what can cause this trouble since 2.6.32-19 ??? !!!

A++

For Steve Conklin, Robert Hooker and Virgil Brummond

Another investigation :

I have reversed the 'UBUNTU-SAUCE-drm-i915-Disable-FBC-on-915GM-and-945GM.patch' and compiled 'linux-image-2.6.32-19-generic-pae' without it, same problem, so it come from another patch :-(

Hope it help ...

A++

Thank you for the additional information. we'll see if we can find this.

Is there anything new on this?

After I upgrade my Karmic to Lucid using apt-get dist-upgrade, I got an unusable compiz environment. Disabling compiz will get me a quite usable, but still some application getting slow and leave artifact in screen. This is using 2.6.32-22.33 kernel.

I reboot back the kernel 2.6.31-20.58 (thank god I have not uninstall this kernel yet). I reactivate the compiz, and it's working great as before upgrade. I'm using this kernel while writing this comment. No configuration changed in xorg.conf.

I'm using intel(0): Chipset: "965GM" with UXA AccelMethod (default) selected by xserver-xorg-video-intel 2:2.9.1-3ubuntu5

Any other information that I can provide to help fixing this bug in the latest kernel?

2.6.31-21 works for me too, I'm using it right now. Still no luck with the latest kernel.

It's been two months, are we getting a fix on this or should I just switch to Windows?

For what it's worth, I haven't had this problem anymore since I replaced the faulty charger for my Dell (I was using the machine with AC power, battery attached but the battery didn't charge). I think the AC/battery problem mixed up CPU speed a bit like described in bug 555315. I think it wasn't ever an actual Intel/graphics problem directly for me, but a CPU slowdown.

Trying to set the CPU speed manually with GNOME CPU speed applet just caused more problems. If you have used it, it's probably better try to reset whatever it did. If you feel like kicking the ondemand scheduler itself, try:
echo 20 > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/up_threshold
echo 20 > /sys/devices/system/cpu/cpu1/cpufreq/ondemand/up_threshold

A different bug, but regarding compiz and maybe what Colin and Stavros are after, one slow compiz problem was earlier solved for me by not having "unredirect_fullscreen_windows" selected in compiz (compizconfig-settings-manager application), but it was not the actual slowdown problem of this bug. Please redirect compiz problems to other bugs.

All in all, it's not very clear what this bug is about: compiz, intel, CPU scheduler, faulty hardware... for me it was CPU speed problem with faulty charger, but let's stick to the original reporter's point of view where possible - if he can't confirm the CPU being (all the time) low clocked at the time the problem occurs, which can be seen with powertop application or GNOME applet, then this bug report is about something other than what I experienced and described and I'm in the wrong bug report anyway. Likewise, if you are seeing eg. "one frame per second" kind of _really_ slow compiz problem, it's not what Virgil seems to be experiencing and probably worth of another bug report about compiz settings. You also shouldn't be asking when _this_ bug is going to be fixed, as your problem is not covered in this actual bug report (similar to my problem possibly not being covered).

Hmm, it doesn't seem that it's a CPU problem, as my CPU *does* remain underclocked with compiz running, but it's not being used anyway (it's at about 10% usage). It seems to be a graphics problem, I'm getting the 1 fps you're talking about for some reason, but my CPU runs fine in other applications. I can see in the panel applet that the clock scales just fine when under stress.

Since everything works fine with the old kernel, this points to a kernel problem (at least not a hardware problem, anyway). I am very frustrated because I have Compiz set up with some shortcuts very useful to my workflow, and my productivity suffers from this bug. On top of that, it doesn't seem to be getting any attention...

My problem does indeed not seem to be the same as this poster's, I will try looking for another bug, thanks for your feedback.

It might even be some sort of xul issue as the frame rate drops noticeably only when FF is open.

I have disabled "Sync to VBlank" as per Timo's suggestion and that fixed it, many thanks!

(A small note that Stavros' comment was to bug #340578 getting fixed he was experiencing and I was also experiencing a year ago, not actually this bug. I wrote to him: "I remembered a wrong setting that fixed my "1 fps" problem earlier. It was disabling the sync to vblank in compizconfig-settings-manager")

Virgil: Just to make sure, observe the CPU clock speed when eg. doing a lot of scrolling. If it doesn't increase from the minimum MHz, it might be the same problem as I was having. Otherwise I will stop commenting on this bug report :)

I have tried several kernels, all with the same performance issues. It is not a CPU issue, setting the frequency makes no difference at all.

In the end, the 2.6.31-11-rt kernel works like a charm. Compiz is incredibly fast.

00:02.0 VGA compatible controller: Intel Corporation 82Q33 Express Integrated Graphics Controller (rev 02)

I installed 10.4 at launch and had great performance. However in late August or early September the symptoms described here appeared. I first noticed an increase in fan speed, but his week my PC has turned almost useless. The system monitor typically shows Compiz using 40 - 60% of CPU.

I am running Linux 2.6.32-24-generic in a default configuration with automatic updates. The motherboard has GMA X4500 graphics with the G41 Express chipset.

I have not found any other bug or discussion related to this latest problem.

Virgil Brummond, this bug was reported a while ago and there hasn't been any activity in it recently. We were wondering if this is still an issue? If so, could you please test for this with the latest development release of Ubuntu? ISO images are available from http://cdimage.ubuntu.com/daily-live/current/ .

If it remains an issue, could you please run the following command in the development release from a Terminal (Applications->Accessories->Terminal), as it will automatically gather and attach updated debug information to this report:

apport-collect -p linux <replace-with-bug-number>

Also, could you please test the latest upstream kernel available following https://wiki.ubuntu.com/KernelMainlineBuilds ? It will allow additional upstream developers to examine the issue. Please do not test the daily kernel folder, but the one all the way at the bottom. Once you've tested the upstream kernel, please comment on which kernel version specifically you tested. If this bug is fixed in the mainline kernel, please add the following tags:
kernel-fixed-upstream
kernel-fixed-upstream-VERSION-NUMBER

where VERSION-NUMBER is the version number of the kernel you tested. For example:
kernel-fixed-upstream-v3.11.1

This can be done by clicking on the yellow circle with a black pencil icon next to the word Tags located at the bottom of the bug description. As well, please remove the tag:
needs-upstream-testing

If the mainline kernel does not fix this bug, please add the following tags:
kernel-bug-exists-upstream
kernel-bug-exists-upstream-VERSION-NUMBER

As well, please remove the tag:
needs-upstream-testing

Once testing of the upstream kernel is complete, please mark this bug's Status as Confirmed. Please let us know your results. Thank you for your understanding.

Closing this bug with Won't fix as this kernel / release is no longer supported.
Please feel free to open a new bug report if you're still experiencing this on a newer release (Bionic 18.04.3 / Disco 19.04)
Thanks!