xorg-server 1.6.0 in testing

otto0 · 2009-03-04 09:08:32

byte wrote:

Use X -configure to get an initial xorg.conf.

Thanks, it works fine now. Another issue on Google Earth: when I started it, I got: "You are currently running Google Earth in OpenGL with software emulation. In this mode, Google Earth will work but it will run very slowly. ... you upgrade your graphics card driver." What is the option in /etc/X11/xorg?

EDIT: command "glxinfo | grep direct" got: "direct rendering: Yes"

Last edited by otto0 (2009-03-04 13:08:46)

congyiwu · 2009-03-04 09:14:43

bender02 wrote:

blasse wrote:
With new xserver and intel gm950 X is not resuming after hibernate/suspend or when switching to another VC and then back.
Same problem here with GM965. Any fix? I have seen some bugreports about this:
http://bugs.freedesktop.org/show_bug.cgi?id=18879
But unless I misunderstood, the patches should be already applied in current versions?

I'm having the same issue with GM965 too! Different combinations of the xserver 1.6 release candidates, intel 1.6 drivers, and libdrm packages from Debian experimental had the same issue intermittently with VT switching and suspending for at least a month on my laptop. This was before lenny got released. Not sure if it's fixed now in Debian, since I switched to arch after I got tired of all the lenny release drama, and stagnation in sid.

jt512 · 2009-03-04 09:21:55

droog wrote:

Working fine here with the new nvidia-96xx.
It seems disabling hotplugging isn't working anymore and i have to run hal/dbus now to get my keyboard and mouse working, anyone else have this problem yet?

I had this problem with the nvidia 180.29 driver—mouse and keyboard frozen at the login screen. Went back to the Archwiki Beginners Guide and tried to fix it by re-enabling hotplugging, and couldn't fix it. I had to roll back to xorg-server 1.5.

shami · 2009-03-04 09:33:45

Finally my intel card is working again.

before:
glxgears: 26 fps
RAM after boot: 185 MB

now:
glxgears: 195 fps
RAM after boot: 85 MB

only when I run glxgears that write:
get fences failed: -1
param: 6, val: 0

fijam · 2009-03-04 09:43:04

I think fencing is supported in 2.6.29 only.

whargoul · 2009-03-04 12:36:34

fijam wrote:

I think fencing is supported in 2.6.29 only.

Although slower at this point.

agapito · 2009-03-04 13:05:15

fwojciec wrote:

agapito wrote:
Anyone else is having problems with mouse buttons? Randomly not working here...
Try rebuilding your WM (or whatever application is causing problems) -- I had to rebuild dwm for mouse buttons to start working properly.

I use KDE, i tried with fluxbox rebuilded and still happens... I had to revert to stable xrog packages because this bug makes me crazier.

b9anders · 2009-03-04 13:17:15

works fine here, except for the suspend/resume thing with gm965.

Upgraded to the 2.6.3 intel driver using the testing pkgbuild and changing the version number with no difference.

andre.ramaciotti · 2009-03-04 13:32:32

It's working here but I can say nothing about performance, as I don't own an Intel GPU.

buddabrod · 2009-03-04 13:35:04

AndyRTR wrote:

Also got my nouveau dualhead desktop running again

Maybe you could add something like the nv50.patch I made to xf86-video-nouveau in order to make it usable for every nv50 user. It doesnt work without it for me but it works like a charm with that little ugly hack.

ebano · 2009-03-04 14:08:42

Since I upgraded to xorg 1.6.0 with pacman testing packages i got my intel gm965 working like a charm again.

No need for any aditional configuration. xorg is DRI2 enabled, UXA enabled, openGL apps are running almost without bug alongside compiz and performace is great!

glxgears is doing about 500fps.

My notebook is a Pentium dual core @1.6Ghz 2GB ddr2 533 Intel gm965 (X3100)

it's nice to see everything running smooth again, and even better than before

Last edited by ebano (2009-03-04 14:09:57)

buttons · 2009-03-04 15:08:52

My mini 9 with Intel 945 gets 1000 fps (in glxgears) with EXA, DRI2, and the testing video driver! o.O
I was on about 200fps before. I had given up hope, honestly.

UXA is pure crap and running a compositor kills performance, though.

Last edited by buttons (2009-03-04 15:12:14)

shining · 2009-03-04 15:14:28

glxgears -iacknowledgethatthistoolisnotabenchmark

Arkane · 2009-03-04 15:22:26

Works great with my intel (915GM I think) using UXA. My Eee can finally play DivX at full speed, and turning on compositiong doesn't reduce the frame rate at all whereas it used to divide it by 10.

I have a small suspend/resume glitch, but switching to a vc and back solves it.

eldragon · 2009-03-04 15:40:25

buttons wrote:

My mini 9 with Intel 945 gets 1000 fps (in glxgears) with EXA, DRI2, and the testing video driver! o.O
I was on about 200fps before. I had given up hope, honestly.
UXA is pure crap and running a compositor kills performance, though.

check your Xorg log, i dont think DRI2 is being enabled with EXA, which invalidates what you said. (UXA being crap and whatelse..)

i dont feel a hit when runing a compositor, there are some glitches of course... but performance/usability is on par compared to metacity

buttons · 2009-03-04 15:42:19

eldragon wrote:

check your Xorg log, i dont think DRI2 is being enabled with EXA, which invalidates what you said. (UXA being crap and whatelse..)
i dont feel a hit when runing a compositor, there are some glitches of course... but performance/usability is on par compared to metacity

Okay, DRI then. That doesn't invalidate anything. UXA is still crap.

gtkperf with EXA: 17s
gtkperf with UXA: 34s

gtkperf with EXA and compositor: 22s
gtkperf with UXA and compositor: 47s

Last edited by buttons (2009-03-04 15:45:37)

buttons · 2009-03-04 15:46:22

shining wrote:

glxgears -iacknowledgethatthistoolisnotabenchmark

No, but it is a nice indicator of relative performance on any given machine.

eldragon · 2009-03-04 15:54:58

buttons wrote:

eldragon wrote:
check your Xorg log, i dont think DRI2 is being enabled with EXA, which invalidates what you said. (UXA being crap and whatelse..)
i dont feel a hit when runing a compositor, there are some glitches of course... but performance/usability is on par compared to metacity
Okay, DRI then. That doesn't invalidate anything. UXA is still crap.
gtkperf with EXA: 17s
gtkperf with UXA: 34s
gtkperf with EXA and compositor: 22s
gtkperf with UXA and compositor: 47s

many people are claiming increased performance, behaviour, in fact, this computer works better with UXA, if it wasnt for the transparency bug, id be using it myself. the fact that you dont feel it, doesnt make it crap. and even the fact that its brand new....dont expect it to be highly optimized either. so, jumping to conclusions this fast seems obnoxious...no wait. stupid..

blasse · 2009-03-04 17:25:05

b9anders wrote:

works fine here, except for the suspend/resume thing with gm965.
Upgraded to the 2.6.3 intel driver using the testing pkgbuild and changing the version number with no difference.

Gentoo has bugreport for it, unsolved for now: http://bugs.gentoo.org/show_bug.cgi?id=261112

skottish · 2009-03-04 17:36:11

For what it's worth, I'm seeing a very slight increase in performance and about a 10MB average decrease in RAM with nVidia. I'm still using the 180.35 nvidia driver.

shining · 2009-03-04 17:38:20

buttons wrote:

shining wrote:
glxgears -iacknowledgethatthistoolisnotabenchmark
No, but it is a nice indicator of relative performance on any given machine.

In my experience, even on a given machine this indicator is totally meaningless.
I already saw after one update of fglrx driver a clear decrease of glxgears fps (from 4000 to 3000), and a clear increase of fps in a 3d game (from 40 to 60).

Now this was maybe an exception, but if people keep posting only glxgears results without any other indicators, they are still as meaningless to me.

agapito · 2009-03-04 17:45:00

agapito wrote:

JGC wrote:
agapito wrote:
Because on my comp, low glxgears frame rate = bad performance on real apps.
Ok, i went to my bios and changed memory management to DMVT and Maximum DMVT, I put UXA on my xorg.conf and performance was better, but still a little slow on composite and mupenplus give me 3 fps. I have dri2 and is cool but i have this bug (http://lists.freedesktop.org/archives/x … 44044.html) alredy solved on git. The best thing was a noticiable low memory consumption of X process. I hope 2.6.29 kernel fix these problems.
You can't have that bug, as I patched xf86-video-intel for that before releasing it to testing.
Trust me i have this bug, but only with UXA activated. I am using EXA now because is stable, but very slow.
00:02.0 VGA compatible controller: Intel Corporation 82945G/GZ Integrated Graphics Controller (rev 02)

I compiled 2.6.3 and this bug dissapears, so maybe 2.6.2 was bad patched.

mutlu_inek · 2009-03-04 17:57:48

shining wrote:

buttons wrote:
shining wrote:
glxgears -iacknowledgethatthistoolisnotabenchmark
No, but it is a nice indicator of relative performance on any given machine.
In my experience, even on a given machine this indicator is totally meaningless.
I already saw after one update of fglrx driver a clear decrease of glxgears fps (from 4000 to 3000), and a clear increase of fps in a 3d game (from 40 to 60).
Now this was maybe an exception, but if people keep posting only glxgears results without any other indicators, they are still as meaningless to me.

You are right in the sense that glxgears does not measure 3D performance per se. But it is a benchmark in the sense that anything can be used a benchmark. A 3D game can be used as a benchmark, testing performance of a certain set of functions. The same is valid for glxgears. It is valid as a test only in so far as it tests a certain set of functions which do not equal to totality of functions and therefore does not measure 3D performance per se. But it has been an acceptable estimate in many cases. The advantage of talking glxgears is that everyone has it installed and also has some sort of perception of what the numbers *might* mean. Thus, I think it is the best among the single "benchmarks" which - apart from glxgears - would mostly be 3D games.

Thus, caution with regard to the numbers as representations of performance for everyone and every case is good and necessary, but outright rejecting the validity of providing numbers does not get us anywhere, either.

I do hope, though, that a comprehensive benchmark (suite) becomes easily available to all (read: phoronix test suite needs to be a) improved and b) easily available as a package without idiotic CPAN dependencies).

blasse · 2009-03-04 18:15:08

blasse wrote:

b9anders wrote:
works fine here, except for the suspend/resume thing with gm965.
Upgraded to the 2.6.3 intel driver using the testing pkgbuild and changing the version number with no difference.
Gentoo has bugreport for it, unsolved for now: http://bugs.gentoo.org/show_bug.cgi?id=261112

Bug reported: http://bugs.archlinux.org/task/13651
automatic voting link removed - tomk

shining · 2009-03-04 18:21:01

mutlu_inek wrote:

You are right in the sense that glxgears does not measure 3D performance per se. But it is a benchmark in the sense that anything can be used a benchmark. A 3D game can be used as a benchmark, testing performance of a certain set of functions. The same is valid for glxgears. It is valid as a test only in so far as it tests a certain set of functions which do not equal to totality of functions and therefore does not measure 3D performance per se. But it has been an acceptable estimate in many cases. The advantage of talking glxgears is that everyone has it installed and also has some sort of perception of what the numbers *might* mean. Thus, I think it is the best among the single "benchmarks" which - apart from glxgears - would mostly be 3D games.

What I am questioning is the proportion of functions tested, which (I am afraid) is too low in the case of glxgears for being an acceptable estimate of the performance. And that's the problem, I have actually no idea what the numbers *might* mean.
So I am even more interested to hear subjective impression like : video playback is smooth now, compiz is much more usable, etc..

Arch Linux

#51 2009-03-04 09:08:32

Re: xorg-server 1.6.0 in testing

#52 2009-03-04 09:14:43

Re: xorg-server 1.6.0 in testing

#53 2009-03-04 09:21:55

Re: xorg-server 1.6.0 in testing

#54 2009-03-04 09:33:45

Re: xorg-server 1.6.0 in testing

#55 2009-03-04 09:43:04

Re: xorg-server 1.6.0 in testing

#56 2009-03-04 12:36:34

Re: xorg-server 1.6.0 in testing

#57 2009-03-04 13:05:15

Re: xorg-server 1.6.0 in testing

#58 2009-03-04 13:17:15

Re: xorg-server 1.6.0 in testing

#59 2009-03-04 13:32:32

Re: xorg-server 1.6.0 in testing

#60 2009-03-04 13:35:04

Re: xorg-server 1.6.0 in testing

#61 2009-03-04 14:08:42

Re: xorg-server 1.6.0 in testing

#62 2009-03-04 15:08:52

Re: xorg-server 1.6.0 in testing

#63 2009-03-04 15:14:28

Re: xorg-server 1.6.0 in testing

#64 2009-03-04 15:22:26

Re: xorg-server 1.6.0 in testing

#65 2009-03-04 15:40:25

Re: xorg-server 1.6.0 in testing

#66 2009-03-04 15:42:19

Re: xorg-server 1.6.0 in testing

#67 2009-03-04 15:46:22

Re: xorg-server 1.6.0 in testing

#68 2009-03-04 15:54:58

Re: xorg-server 1.6.0 in testing

#69 2009-03-04 17:25:05

Re: xorg-server 1.6.0 in testing

#70 2009-03-04 17:36:11

Re: xorg-server 1.6.0 in testing

#71 2009-03-04 17:38:20

Re: xorg-server 1.6.0 in testing

#72 2009-03-04 17:45:00

Re: xorg-server 1.6.0 in testing

#73 2009-03-04 17:57:48

Re: xorg-server 1.6.0 in testing

#74 2009-03-04 18:15:08

Re: xorg-server 1.6.0 in testing

#75 2009-03-04 18:21:01

Re: xorg-server 1.6.0 in testing

Board footer