[UXA] Anyone getting good perfomance on Intel 945gm?

agapito · 2009-04-30 16:47:36

rc4 is launched can you update it?

bangkok_manouel · 2009-04-30 16:49:56

icenine457 wrote:

I just updated my PKGBUILD and removed the arch patches, as they conflict with the rc3 patchset.

FTR, -rc4 was released earlier today and it fixes the tiling problem.

icenine457 · 2009-04-30 17:12:28

Jeez, I just started recompiling . Ill update and see if I get any failures.

icenine457 · 2009-04-30 17:58:30

When compiling the latest rc, the following errors were thrown:

drivers/gpu/drm/i915/i915_opregion.c: In function 'intel_opregion_free':
drivers/gpu/drm/i915/i915_opregion.c:422: error: implicit declaration of function 'acpi_video_exit'
make[4]: *** [drivers/gpu/drm/i915/i915_opregion.o] Error 1

Any ideas? Otherwise I'm going to keep the rc3 up for now.

Last edited by icenine457 (2009-04-30 18:06:45)

agapito · 2009-04-30 18:08:15

I compiled rc4 fine and performance still SUCKS.

With 2.7 driver from testing a little better, but still bad performance.
With 2.7.99 driver from aur composite doesn't work.

This card on XP or Vista works fine, with linux is a pain.

ngoonee · 2009-05-01 00:19:36

SomeGuyDude, you may want to change your config from make config to make menuconfig or makegconfig if you're using gnome (I ilke gconfig, somehow).

SomeGuyDude · 2009-05-01 00:34:06

Openbox, actually. But I'll take a look.

bangkok_manouel · 2009-05-01 06:32:10

icenine457 wrote:

When compiling the latest rc, the following errors were thrown:
drivers/gpu/drm/i915/i915_opregion.c: In function 'intel_opregion_free':
drivers/gpu/drm/i915/i915_opregion.c:422: error: implicit declaration of function 'acpi_video_exit'
make[4]: *** [drivers/gpu/drm/i915/i915_opregion.o] Error 1
Any ideas? Otherwise I'm going to keep the rc3 up for now.

you need to chose acpi_video to be built-in to get this to compile properly (or i915 as a module).

baze · 2009-05-01 08:49:07

just took the -ninekernel stuff from aur, changed -rc3 to -rc4 and built that kernel and opengl is finally back to nice performance with it! no xorg.conf, kms, dri2, uxa everything is running and is running smoothly
looks like the 2.6.30 release will finally bring this whole performance regression phase to an end.

btw, i'm running xf86-video-intel-git, latest master.

Last edited by baze (2009-05-01 08:53:39)

keratos · 2009-05-03 22:21:51

why do people insist on using GLXgears.

use phoronix-test-suite ... in the repos !

I mean if nexuiz plays at 750fps then your computer should perhaps be more entertained attempting to calculae the largest prime number.

wnwek · 2009-05-04 03:07:29

I have an Intel 945GM chipset and GNOME with Compiz-Fusion on my Dell Latitude 520 runs quite okay on my laptop, though I have a nagging feeling, my xorg.conf is not exactly most efficiently written. Especially because it has multiple "Device" and "Screen" sections.

And it has no keyboard and mouse config statements. It seems to be all running on auto-configure, and I know I enabled hot-plugging, but I can't remember which statements in my xorg.conf does that.

Can anybody tell me why this seems to be working? And how it can be made better?

Edit: Also, my screen blanks out sometimes, but it comes back on immediately, and my CPU temperatures seem to be running at around 65 degrees C (147 F) constantly, even though there doesn't seem to be any significant consumption of CPU cycles by any process.

I was earlier on Ubuntu where my core temperatures would run to about 55 degrees C at the most, and 65 seems to be really high to me. (I don't want to go back, so I'd love to have somebody give me some pointers.)

Section "Device"
    Identifier    "Builtin Default intel Device 0"
    Driver    "intel"
EndSection
Section "Screen"
    Identifier    "Builtin Default intel Screen 0"
    Device    "Builtin Default intel Device 0"
EndSection
Section "Device"
    Identifier    "Builtin Default i810 Device 0"
    Driver    "i810"
    Option      "DRI"                    "True"
    Option      "NoDDC"                  "True"
    Option      "XAANoOffscreenPixmaps"  "True"
    Option      "EnablePageFlip"         "True"
    Option      "RenderAccel"            "True"
    Option      "AccelMethod"            "XAA"
EndSection
Section "Screen"
    Identifier    "Builtin Default i810 Screen 0"
    Device    "Builtin Default i810 Device 0"
EndSection
Section "Device"
    Identifier    "Builtin Default vesa Device 0"
    Driver    "vesa"
EndSection
Section "Screen"
    Identifier    "Builtin Default vesa Screen 0"
    Device    "Builtin Default vesa Device 0"
EndSection
Section "Device"
    Identifier    "Builtin Default fbdev Device 0"
    Driver    "fbdev"
EndSection
Section "Screen"
    Identifier    "Builtin Default fbdev Screen 0"
    Device    "Builtin Default fbdev Device 0"
EndSection
Section "ServerLayout"
    Identifier    "Builtin Default Layout"
    Screen    "Builtin Default intel Screen 0"
    Screen    "Builtin Default i810 Screen 0"
    Screen    "Builtin Default vesa Screen 0"
    Screen    "Builtin Default fbdev Screen 0"
EndSection
Section "Extensions"
    Option "Composite" "Enable"
EndSection

Last edited by wnwek (2009-05-04 05:47:52)

tjwoosta · 2009-05-04 05:21:40

keratos wrote:

why do people insist on using GLXgears.
use phoronix-test-suite ... in the repos !
I mean if nexuiz plays at 750fps then your computer should perhaps be more entertained attempting to calculae the largest prime number.

i was intrigued by your post and decided to install the phoronix-test-suite

i must admit its great, although ive only completed the render-bench test so far

which tests would you reccomend ?

Liquen · 2009-05-04 06:40:03

wnwek wrote:

(...)

Your xorg.conf looks messed up. It's specifying both the new intel and the old i810 drivers at same time. Just wipe it out, take a look at the Xorg wiki page and try one of the methods to generate a new xorg.conf (I used X -configure and it worked fine; here is mine for anyone interested). You can also try without xorg.conf.

By the way, I'm still using kernel 2.6.28, it has a better video performance than .29 on my system. Still waiting for .30 to solve all my problems.

About your temperature problem: Did you enable CPU Frequency Scaling? Try to follow this steps if you didn't, it really helps to cool things down. EDIT: you may want to see if it worked by using the CPU Frequency Scaling Monitor Gnome applet.

Last edited by Liquen (2009-05-04 07:02:15)

wnwek · 2009-05-04 07:06:03

Actually, I did follow the wiki. But let me try again. I'll post the results.

Question: Now that I know that it's messed up, why is it working? How does xorg.conf work? I mean in case of multiple definitions, which is the one that is used?

Liquen · 2009-05-04 07:10:03

wnwek wrote:

Actually, I did follow the wiki. But let me try again. I'll post the results.
Question: Now that I know that it's messed up, why is it working? How does xorg.conf work? I mean in case of multiple definitions, which is the one that is used?

You can try inspecting /var/log/Xorg.0.log to see which one is being used.

(By the way, maybe it's not messed up as I told you, I'm a Xorg newbie. But I never saw a xorg.conf like that...)

Last edited by Liquen (2009-05-04 07:11:18)

keratos · 2009-05-04 07:11:43

tjwoosta wrote:

keratos wrote:
why do people insist on using GLXgears.
use phoronix-test-suite ... in the repos !
I mean if nexuiz plays at 750fps then your computer should perhaps be more entertained attempting to calculae the largest prime number.
i was intrigued by your post and decided to install the phoronix-test-suite
i must admit its great, although ive only completed the render-bench test so far
which tests would you reccomend ?

the suite name is x-testing , it has a number of tests within. have a play running the suite or individual tests. you will probably get a more realistic fps. it is both ridiculous and erroneous to use glxgears as a benchmark for X and OpenGL performance.

glxgears yields but a single performance figure for a single attribute related to the performance of the graphics subsystem. I suppose it is analagous to measuring the performance of your PC by just examining the performance of memory. one would have to of course measure not just the memory, but the CPU, the graphics performance, the HDD performance, the systembus/FSB, and so on.

I wish they would take glxgears out of the repos !

Last edited by keratos (2009-05-04 07:18:44)

Dr4go · 2009-05-04 07:42:45

@ Liquen: You are right. His config is totally messed up.

@ wnwek: If I were you, I would kill the xorg-server, then (as root) do a "Xorg -configure" and move it from "/root/xorg.conf.new" to "/etc/X11/xorg.conf". After that you can / should edit it a bit:
- for example if you want to use evdev you can / or you should move out everything regarding InputDevices. If you don't want to use evdev you need to specify AutoAddDevices in the serverflags section and change it to false.
- second: why do you use xaa in your xorg.conf? You should use exa or uxa as acceleration. If you don't use kms, simply use exa. And also use exa if you don't know what kms is. Add the following to the "intel" driver section:

Option        "AccelMethod"            "exa"
Option        "MigrationHeuristic"        "greedy"
Option        "DRI"                "true"
Option        "ExaNoComposite"        "false"
Option        "ExaOptimizeMigration"    "true"

And add a new section like this:

Section "DRI"
    Mode    0666
EndSection

You should be good to go... but remember: xaa, exa and even dri (version 1) only works 'til version 2.7.lower numbers (not with >=2.7.99.x)! That means, that the hints and code above are only useful up to this version. After that (i.e. when v2.8 gets released) you really should switch to kms / uxa.

keratos · 2009-05-04 08:02:20

Dr4go wrote:

{cut}.... After that (i.e. when v2/.8 gets released) you really should switch to kms / uxa.

Why?

Where is the evidence to support the need for migration.

I have run phoronix-test-suite against X performance and observe the following:
2.6.29 + intel + XAA = crashes!
2.6.29 + intel + EXA = 130ps
2.6.29 + intel + UXA + kms = 115fps
2.6.29 + legacy_intel + XAA = 110fps
2.6.29 + legacy_intel + EXA = 135fps
2.6.29 + legacy_intel + UXA + kms = 115fps
2.6.27 + intel + XAA = crashes
2.6.27 + intel + EXA = 117fps
2.6.27 + intel + UXA + kms = crashes/not compatible

I did not try legacy with 2.6.27 as I require 2.6.29 for my H/W

so based on the above and assuming we accept phoronix-test-suite as a viable alternative to glxgears (!!) then the conclusion I draw is that 2.6.29 + EXA + legacy driver is by far the better outperformer. GEM/UXA does isnt cutting it compared to the matured legacy driver (which was always damm good!)

wnwek · 2009-05-04 08:52:11

Thanks a lot for your help, Liquen and Dr4go

Okay, I did an X -configure, and I have a more orderly xorg.conf now. There was this brief moment of madness when I added those extra lines in Section "Device" which Dr4go recommended I remove. But it's working okay now. No arbit flickers on my screen now.

Dr4go, I'll add the lines into my xorg.conf, and write what happened.

Liquen, you were right I think. My processor was being over-clocked, though I don't understand how that happened. I installed cpufrequtils and brought it within the limits. Although I am relieved that the core temps are saner values now, I am sad that it's not as zippy as before Oh well, better a slower computer (but it's still much faster than Ubuntu) than a fried CPU.

wnwek · 2009-05-04 08:55:34

Dr4go, here is the latest version of my xorg.conf. I'd be grateful if you could look it over and see if anything is still messed up.

Section "ServerLayout"
    Identifier     "X.org Configured"
    Screen      0  "Screen0" 0 0
    InputDevice    "Mouse0" "CorePointer"
    InputDevice    "Keyboard0" "CoreKeyboard"
EndSection

Section "ServerFlags"
    Option "AutoAddDevices" "False"
EndSection

Section "Files"
    ModulePath   "/usr/lib/xorg/modules"
    FontPath     "/usr/share/fonts/misc"
    FontPath     "/usr/share/fonts/100dpi:unscaled"
    FontPath     "/usr/share/fonts/75dpi:unscaled"
    FontPath     "/usr/share/fonts/TTF"
    FontPath     "/usr/share/fonts/Type1"
EndSection

Section "Module"
    Load  "extmod"
    Load  "glx"
    Load  "dbe"
    Load  "record"
    Load  "dri"
    Load  "dri2"
EndSection

Section "InputDevice"
    Identifier  "Keyboard0"
    Driver      "kbd"
EndSection

Section "InputDevice"
    Identifier  "Mouse0"
    Driver      "mouse"
    Option        "Protocol" "auto"
    Option        "Device" "/dev/input/mice"
    Option        "ZAxisMapping" "4 5 6 7"
EndSection

Section "Monitor"
    Identifier   "Monitor0"
    VendorName   "Monitor Vendor"
    ModelName    "Monitor Model"
EndSection

Section "Device"
        ### Available Driver options are:-
        ### Values: <i>: integer, <f>: float, <bool>: "True"/"False",
        ### <string>: "String", <freq>: "<f> Hz/kHz/MHz"
        ### [arg]: arg optional
        #Option     "NoAccel"                # [<bool>]
        #Option     "SWcursor"               # [<bool>]
        #Option     "ColorKey"               # <i>
        #Option     "CacheLines"             # <i>
        #Option     "Dac6Bit"                # [<bool>]
        #Option     "DRI"                    # [<bool>]
        #Option     "NoDDC"                  # [<bool>]
        #Option     "ShowCache"              # [<bool>]
        #Option     "XvMCSurfaces"           # <i>
        #Option     "PageFlip"               # [<bool>]
    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"    
    Option      "ExaNoComposite"        "false"
    Option      "ExaOptimizeMigration"    "true"
EndSection

Section "DRI"
    Mode    0666
EndSection

Section "Extensions"
    Option "Composite" "Enable"
EndSection

Section "Screen"
    Identifier "Screen0"
    Device     "Card0"
    Monitor    "Monitor0"
    SubSection "Display"
        Viewport   0 0
        Depth     1
    EndSubSection
    SubSection "Display"
        Viewport   0 0
        Depth     4
    EndSubSection
    SubSection "Display"
        Viewport   0 0
        Depth     8
    EndSubSection
    SubSection "Display"
        Viewport   0 0
        Depth     15
    EndSubSection
    SubSection "Display"
        Viewport   0 0
        Depth     16
    EndSubSection
    SubSection "Display"
        Viewport   0 0
        Depth     24
    EndSubSection
EndSection

wnwek · 2009-05-04 09:00:36

Reporting results on using the new xorg.conf, and it seems to be working quite okay. No random lines, no random blackouts. Thanks again Dr4go, and Liquen.

eldragon · 2009-05-04 14:01:53

DRI AccelMethod ExaNoComposite are by default those values expressed there, so they are actually not needed.

ExaOptimizeMigration, i could not find it in the manpages anywhere, is it a valid option? check xorg.0.log for more details

having a minimal xorg file is desirable. i would comment out default options. and of course, remove invalid ones (but check your logs first, might only be undocumented )

wnwek · 2009-05-04 14:30:34

@eldragon, thanks for your comments. I checked xorg.0.log - no errors except something cryptic about exaCopyDirty:

exaCopyDirty: Pending damage region empty!

I am not sure whether that has something to do with ExaOptimizeMigration, but that's the only mention of exa anywhere in my log.

Two questions:

1. xorg.conf is working beautifully now, except for one thing: my screen sometimes switches off for half a second, and then comes back right now, like it never happened. This has never happened before to me. What do you think must be causing it? Is it dangerous for my screen, or should I just ignore it?

2. Where do I find out about the default options? man xorg.conf has the most elaborate explanation of what goes in there, but it's not sufficient. And I don't want to keep bugging you guys about these things

Does anybody know where I can get hold of a detailed explanation of all options that go in each section? As eldragon pointed out, it's definitely better to have as minimal an xorg.conf possible so that less can go wrong, and frankly, that is the Arch way - KISS, right?

Dr4go · 2009-05-04 14:46:21

@ eldragon: Damn, you are right... It seems that intel removed the option ExaOptimizeMigration somewhere in the latest drivers (it was documented before)... Maybe also the default values changed, so I completely agree, that...

@ wnwek: ...you should wipe out ExaOptimizeMigration, ExaNoComposite, DRI and AccelMethod from the device's section. You can also remove the extensions' section, because Composite also should be enabled by default.
Regarding the screens' section I'd recommend, that:
- you add the option "DefaultDepth 24",
- remove all subsections, except the Depth 16 and 24 and
- specify the resolutions you want to have, example:
Modes "1280x800" "1024x768" "800x600" "640x480"

But... those things are minimal changes and should not be that important to have a working config.

@ keratos: The reason is simple... because Intel removed xaa, exa and dri completely starting at driver v2.7.99.x

Greetings,
Dr4go

EDIT: Typos.

2nd EDIT: @ wnwek: Also check out man exa.

3rd EDIT: Again @ wnwek: The permissions for DRI are also not needed anymore.

Last edited by Dr4go (2009-05-04 15:52:40)

keratos · 2009-05-04 16:56:34

Dr4Go: I know those specific accel methods have been removed. My point precisely. Intel's GMA+UXA does doesnt cut it in terms of performance, using a variety of soak tests to gather the evidence.

The legacy drivers outperform the newer - or at least that combination with kernel 2.6.29 does. Maybe kernel space code will improve in 2.6.30 but for now, the legacy options are the best all round option for related intel chipsets.

Why dont people just install sensible benchmark tools and post the output of comparisions here if doubt prevails.

Simples :0

Last edited by keratos (2009-05-04 16:58:32)

Arch Linux

#51 2009-04-30 16:47:36

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#52 2009-04-30 16:49:56

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#53 2009-04-30 17:12:28

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#54 2009-04-30 17:58:30

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#55 2009-04-30 18:08:15

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#56 2009-05-01 00:19:36

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#57 2009-05-01 00:34:06

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#58 2009-05-01 06:32:10

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#59 2009-05-01 08:49:07

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#60 2009-05-03 22:21:51

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#61 2009-05-04 03:07:29

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#62 2009-05-04 05:21:40

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#63 2009-05-04 06:40:03

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#64 2009-05-04 07:06:03

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#65 2009-05-04 07:10:03

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#66 2009-05-04 07:11:43

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#67 2009-05-04 07:42:45

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#68 2009-05-04 08:02:20

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#69 2009-05-04 08:52:11

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#70 2009-05-04 08:55:34

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#71 2009-05-04 09:00:36

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#72 2009-05-04 14:01:53

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#73 2009-05-04 14:30:34

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#74 2009-05-04 14:46:21

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

#75 2009-05-04 16:56:34

Re: [UXA] Anyone getting good perfomance on Intel 945gm?

Board footer