You are not logged in.
glxgears performance drops a lot when using DRI2
Offline
Anybody knows what this error means?
[drm:i915_gem_detect_bit_6_swizzle] *ERROR* Couldn't read from MCHBAR. Disabling tiling.
Maybe it is related to the big performance loss?
By the ways, enabling KMS in [non-testing] makes returns from suspend to ram display a black screen.
Last edited by vinvin (2009-04-14 16:27:49)
Offline
glxgears performance drops a lot when using DRI2
My does also but everything I have done so far has ran really smooth much smoother then without kms enabled.
I am getting about 18 - 22 average fps in etracer.
I am happy so far
Last edited by cu3edweb (2009-04-14 16:46:26)
Offline
This commit might make KMS more usable for those with GMA945:
http://cgit.freedesktop.org/xorg/driver … 651ce6afe2
I will try to rebuild the stack from git on the weekend.
Edit: FYI, this commit to the drm-intel branch of the kernel is supposed to speed things up for us quite a bit, too: http://git.kernel.org/?p=linux/kernel/g … cbb5de859c
Last edited by mutlu_inek (2009-04-15 17:21:24)
Offline
This commit might make KMS more usable for those with GMA945:
http://cgit.freedesktop.org/xorg/driver … 651ce6afe2I will try to rebuild the stack from git on the weekend.
Edit: FYI, this commit to the drm-intel branch of the kernel is supposed to speed things up for us quite a bit, too: http://git.kernel.org/?p=linux/kernel/g … cbb5de859c
anything on the memory leak? which seems quite more urgent than performance just asking...
Offline
You fellows are extremely happy: you've been worrying about performance. Look at where I got stuck: here
Last edited by Llama (2009-04-15 18:50:21)
Offline
mutlu_inek wrote:This commit might make KMS more usable for those with GMA945:
http://cgit.freedesktop.org/xorg/driver … 651ce6afe2I will try to rebuild the stack from git on the weekend.
Edit: FYI, this commit to the drm-intel branch of the kernel is supposed to speed things up for us quite a bit, too: http://git.kernel.org/?p=linux/kernel/g … cbb5de859c
anything on the memory leak? which seems quite more urgent than performance just asking...
The memory leak has been marked as fixed ( http://bugs.freedesktop.org/show_bug.cgi?id=20704 ). It was a xorg bug which seems to be solved on 1.6.1 (just updated on [extra]).
My blog: blog.marcdeop.com
Jabber ID: damnshock@jabber.org
Offline
eldragon wrote:mutlu_inek wrote:This commit might make KMS more usable for those with GMA945:
http://cgit.freedesktop.org/xorg/driver … 651ce6afe2I will try to rebuild the stack from git on the weekend.
Edit: FYI, this commit to the drm-intel branch of the kernel is supposed to speed things up for us quite a bit, too: http://git.kernel.org/?p=linux/kernel/g … cbb5de859c
anything on the memory leak? which seems quite more urgent than performance just asking...
The memory leak has been marked as fixed ( http://bugs.freedesktop.org/show_bug.cgi?id=20704 ). It was a xorg bug which seems to be solved on 1.6.1 (just updated on [extra]).
ohh thanks, i just booted into KMS after updating xorg to see if it was fixed....now i see xorg sits tight at 50ish megs
now the last bug that annoys me....the transparency bug
Offline
Anyone already tried the new intel-2.7.0 release?
Last edited by GogglesGuy (2009-04-16 02:02:56)
Offline
Anyone already tried the new intel-2.7.0 release?
just built it.... had a couple of xorg restarts out of the blue while loading, yet now everything works fine.........feels faster. transparency bug still there though (KMS + UXA)
EDIT:
Mythtv freezes everything when run on fullscreen.
some font glitches are gone
performance on par with previous version (2.6.99something)
Last edited by eldragon (2009-04-16 03:26:24)
Offline
Anyone already tried the new intel-2.7.0 release?
Running the RC3 here, no KMS nought. Maybe this weekend I'll find time to give it a try. Anyway, performance is acceptable with this release (still *not* good)
My blog: blog.marcdeop.com
Jabber ID: damnshock@jabber.org
Offline
xorg-server-1.6.1 really fixed the memory leak. Finally KDE4 is enjoyable again. The memory of X is still increasing over time but very slowly. I will keep it running for a couple of hours and see if anything happens.
Offline
I'm running 2.7.0 right now. I needed to disable tiling to avoid corruption.
I've just come from 2.6.3 and performance may have improved a bit. I was having issues with youtube (X -> 100% CPU), and that may have improved a bit - not really sure yet, or how related that is to xorg-server.
Offline
well, random lockups with 2.7 and KMS
im back to EXA, which boosts performance....feels nice.
Offline
Well xf86-video-intel 2.7.0 didn't improve any 3D performance on my i915. UXA with KMS works, but 3D is awful - 1-2fps at etracer. xf86-video-intel-legacy is a bit better (at least menus work smoothly) but game still unplayable.
and game was playable before on about 30fps in some combination of xorg-server and drivers (don't remember which).
Offline
I'm running 2.7.0 right now. I needed to disable tiling to avoid corruption.
I've just come from 2.6.3 and performance may have improved a bit. I was having issues with youtube (X -> 100% CPU), and that may have improved a bit - not really sure yet, or how related that is to xorg-server.
Same thing here. Disabled tiling otherwise screen contained garbage. (on a GM965 )
On a positive note, don't see any video tearing anymore!!
Last edited by GogglesGuy (2009-04-16 22:19:58)
Offline
Anyone already tried the new intel-2.7.0 release?
I tried it today, installing the xf86-video-intel-2.7.0 package from testing and activating KMS (the 'late' activation). No problems playing videos, but the menus are slightly garbled and the performance is worser than that obtained with xf86-video-intel-legacy.
the output of lspci|grep VGA is
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
Here are some benchmarks:
xf86-video-intel-2.7.0 + KMS
glxgears ~320 FPS, gtkperf -a ~26 seconds, assaultcube 8-9 FPS, torcs 7-8 FPS, tremulous ~8 FPS.
xf86-video-intel-legacy-2.3.2 (no console framebuffer, i.e. no vga= option in grub menu)
glxgears ~780 FPS, gtkperf -a ~13 seconds, assaultcube ~45 FPS, torcs 7-8 FPS, tremulous garbled, unplayable.
Notes: assaultcube is custom compiled, torcs and tremulous come from the community repos. Torcs window lacks the decoration with intel-legacy, tremulous show garbled characters and mouse pointer, I'm even unable to start a game.
So I'll stick with the intel-legacy driver for now, but I'll try the new intel driver in testing when it'll be updated.
EDIT: I forgot to mention that tried xf86-video-intel-2.7.0 + KMS with and without the nopat option in grub, I didn't notice any difference.
Last edited by gborzi (2009-04-16 23:57:27)
Giuseppe Borzi' - Registered Linux user #34028
Offline
GogglesGuy wrote:Anyone already tried the new intel-2.7.0 release?
I tried it today, installing the xf86-video-intel-2.7.0 package from testing and activating KMS (the 'late' activation). No problems playing videos, but the menus are slightly garbled and the performance is worser than that obtained with xf86-video-intel-legacy.
the output of lspci|grep VGA is00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
Here are some benchmarks:
xf86-video-intel-2.7.0 + KMS
glxgears ~320 FPS, gtkperf -a ~26 seconds, assaultcube 8-9 FPS, torcs 7-8 FPS, tremulous ~8 FPS.
xf86-video-intel-legacy-2.3.2 (no console framebuffer, i.e. no vga= option in grub menu)
glxgears ~780 FPS, gtkperf -a ~13 seconds, assaultcube ~45 FPS, torcs 7-8 FPS, tremulous garbled, unplayable.
Notes: assaultcube is custom compiled, torcs and tremulous come from the community repos. Torcs window lacks the decoration with intel-legacy, tremulous show garbled characters and mouse pointer, I'm even unable to start a game.
So I'll stick with the intel-legacy driver for now, but I'll try the new intel driver in testing when it'll be updated.EDIT: I forgot to mention that tried xf86-video-intel-2.7.0 + KMS with and without the nopat option in grub, I didn't notice any difference.
ive got the exact same chip, my 3d performance is slightly better on glxgears (900fps), cube (i asume assaultcube = cube) 50fps, rest i wouldnt know
chromium at about 50fps, etracer at about 20fps, urban terror (800x600, about 2Xfps)
compiz works better too (slightly)
Offline
I'm getting some random freezes for the first time since 2.5.x, but the kernel modesetting is working great on dual monitor (GM45).
The other issue (that's yesterdays log):
$ zcat messages.1.gz | grep -c "freeing invalid memtype"
117672
I assume this is what I'm waiting for:
http://marc.info/?l=linux-kernel&m=123923026401648&w=2
Offline
I tried it today, installing the xf86-video-intel-2.7.0 package from testing
I would like to try it too, but i can't find it in testing. pacman -Sl testing doesn't displays any X related package.
How did you guys get this driver working in arch?
Thanks
Offline
http://www.archlinux.org/packages/?q=xf86-video-intel
Wait for your mirror to sync, but at the moment there is only a x86_64 version.
Last edited by john_schaf (2009-04-17 09:59:22)
Offline
Ah yes ok sorry. Now I know where to look at least. I'm with i686. Thanks
Offline
ive got the exact same chip, my 3d performance is slightly better on glxgears (900fps), cube (i asume assaultcube = cube) 50fps, rest i wouldnt know
chromium at about 50fps, etracer at about 20fps, urban terror (800x600, about 2Xfps)
compiz works better too (slightly)
Do you use XAA acceleration or the default EXA, like I do?
Giuseppe Borzi' - Registered Linux user #34028
Offline
eldragon wrote:ive got the exact same chip, my 3d performance is slightly better on glxgears (900fps), cube (i asume assaultcube = cube) 50fps, rest i wouldnt know
chromium at about 50fps, etracer at about 20fps, urban terror (800x600, about 2Xfps)
compiz works better too (slightly)Do you use XAA acceleration or the default EXA, like I do?
im using EXA, but UXA+KMS drops similar results (except for glxgears which drops to 400fps, who cares...) my device section here:
Section "Device"
Identifier "Card0"
Driver "intel"
VendorName "Intel Corporation"
BoardName "Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller"
BusID "PCI:0:2:0"
Option "AccelMethod" "EXA"
Option "MigrationHeuristic" "greedy"
Option "DRI" "True"
EndSection
im not sure if the DRI option is needed. and im not using kms+uxa cause of the random lockups and transparency bug.
Offline
I don't think DRI is really needed, I tried using the "bare" ubuntu xorg.conf and got similar results.
Section "Device"
Identifier "Configured Video Device"
Driver "intel"
EndSection
Take a look at /var/log/Xorg.0.log to see which option is really used.
Giuseppe Borzi' - Registered Linux user #34028
Offline