Skip to contentSkip to conversation

3d performance very low

DISCUSSION Hanno Böck 11 participants 19 messages Jul 28, 2010 – Aug 1, 2010

lore
Share & tools
lore
#0 lore raw

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

From the logs everything seems to be fine. Any ideas what I could look for what's wrong here?

kernel 2.6.34-rc6, xorg-server-1.8.2, xf86-video-intel-2.12.0,T61 laptop, card is
00:02.0 VGA compatible controller: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c)

dmesg says:
agpgart-intel 0000:00:00.0: Intel 965GM Chipset agpgart-intel 0000:00:00.0: detected 7676K stolen memory agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xe0000000 [drm] Initialized drm 1.1.0 20060810
i915 0000:00:02.0: power state changed by ACPI to D0
i915 0000:00:02.0: power state changed by ACPI to D0
i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16 i915 0000:00:02.0: setting latency timer to 64
ACPI: Battery Slot [BAT0] (battery present)
i915 0000:00:02.0: irq 45 for MSI/MSI-X
[drm] set up 7M of stolen space
[drm] initialized overlay support
Console: switching to colour frame buffer device 175x65
fb0: inteldrmfb frame buffer device
drm: registered panic notifier
acpi device:01: registered as cooling_device2
input: Video Bus as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/LNXVIDEO:00/input/input3
ACPI: Video Device [VID] (multi-head: yes rom: no post: no) [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

xorg log says:

[    24.079] (==) Depth 24 pixmap format is 32 bpp
[    24.079] (II) intel(0): [DRI2] Setup complete
[    24.079] (II) intel(0): [DRI2]   DRI driver: i965
[    24.079] (**) intel(0): Tiling enabled
[    24.079] (**) intel(0): SwapBuffers wait enabled
[    24.079] (==) intel(0): VideoRam: 262144 KB
[    24.079] (II) intel(0): Allocated new frame buffer 1408x1050 stride 5632, 
tiled
[    24.119] (II) UXA(0): Driver registered support for the following 
operations:
[    24.119] (II)         solid
[    24.119] (II)         copy
[    24.119] (II)         composite (RENDER acceleration)
[    24.119] (II)         put_image
[    24.119] (II)         get_image
[    24.119] (==) intel(0): Backing store disabled
[    24.119] (==) intel(0): Silken mouse enabled
[    24.119] (II) intel(0): Initializing HW Cursor
[    24.147] (II) intel(0): RandR 1.2 enabled, ignore the following RandR 
disabled message.
[    24.152] (==) intel(0): DPMS enabled
[    24.152] (==) intel(0): Intel XvMC decoder enabled
[    24.152] (II) intel(0): Set up textured video
[    24.152] (II) intel(0): Set up overlay video
[    24.152] (II) intel(0): [XvMC] i965_xvmc driver initialized.
[    24.152] (II) intel(0): direct rendering: DRI2 Enabled
[    24.152] (--) RandR disabled
[    24.152] (II) Initializing built-in extension Generic Event Extension
[    24.153] (II) Initializing built-in extension SHAPE
[    24.153] (II) Initializing built-in extension MIT-SHM
[    24.153] (II) Initializing built-in extension XInputExtension
[    24.153] (II) Initializing built-in extension XTEST
[    24.153] (II) Initializing built-in extension BIG-REQUESTS
[    24.153] (II) Initializing built-in extension SYNC
[    24.153] (II) Initializing built-in extension XKEYBOARD
[    24.153] (II) Initializing built-in extension XC-MISC
--
[    24.153] (II) Initializing built-in extension RENDER
[    24.153] (II) Initializing built-in extension RANDR
[    24.153] (II) Initializing built-in extension COMPOSITE
[    24.153] (II) Initializing built-in extension DAMAGE
[    24.430] (II) AIGLX: enabled GLX_MESA_copy_sub_buffer
[    24.430] (II) AIGLX: enabled GLX_INTEL_swap_event
[    24.430] (II) AIGLX: enabled GLX_SGI_swap_control and 
GLX_MESA_swap_control
[    24.430] (II) AIGLX: enabled GLX_SGI_make_current_read
[    24.430] (II) AIGLX: GLX_EXT_texture_from_pixmap backed by buffer objects
[    24.430] (II) AIGLX: Loaded and initialized /usr/lib64/dri/i965_dri.so
[    24.430] (II) GLX: Initialized DRI2 GL provider for screen 0
Signature · 4 lines
Hanno Böck		Blog:		http://www.hboeck.de/
GPG: 3DBD3B20		Jabber/Mail:	[email protected]

http://schokokeks.org - professional webhosting

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#1 lore raw
Quoted 10 lines · Hanno Böck

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

From the logs everything seems to be fine. Any ideas what I could look for what's wrong here?

kernel 2.6.34-rc6, xorg-server-1.8.2, xf86-video-intel-2.12.0,T61 laptop, card

The most important piece of software here is Mesa. What version of Mesa?

Quoted 91 lines

is
00:02.0 VGA compatible controller: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c)

dmesg says:
agpgart-intel 0000:00:00.0: Intel 965GM Chipset agpgart-intel 0000:00:00.0: detected 7676K stolen memory agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xe0000000 [drm] Initialized drm 1.1.0 20060810
i915 0000:00:02.0: power state changed by ACPI to D0
i915 0000:00:02.0: power state changed by ACPI to D0
i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16 i915 0000:00:02.0: setting latency timer to 64
ACPI: Battery Slot [BAT0] (battery present)
i915 0000:00:02.0: irq 45 for MSI/MSI-X
[drm] set up 7M of stolen space
[drm] initialized overlay support
Console: switching to colour frame buffer device 175x65
fb0: inteldrmfb frame buffer device
drm: registered panic notifier
acpi device:01: registered as cooling_device2
input: Video Bus as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/LNXVIDEO:00/input/input3
ACPI: Video Device [VID] (multi-head: yes  rom: no  post: no) [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

xorg log says:
[    24.079] (==) Depth 24 pixmap format is 32 bpp
[    24.079] (II) intel(0): [DRI2] Setup complete
[    24.079] (II) intel(0): [DRI2]   DRI driver: i965
[    24.079] (**) intel(0): Tiling enabled
[    24.079] (**) intel(0): SwapBuffers wait enabled
[    24.079] (==) intel(0): VideoRam: 262144 KB
[    24.079] (II) intel(0): Allocated new frame buffer 1408x1050 stride 5632, tiled
[    24.119] (II) UXA(0): Driver registered support for the following operations:
[    24.119] (II)         solid
[    24.119] (II)         copy
[    24.119] (II)         composite (RENDER acceleration)
[    24.119] (II)         put_image
[    24.119] (II)         get_image
[    24.119] (==) intel(0): Backing store disabled
[    24.119] (==) intel(0): Silken mouse enabled
[    24.119] (II) intel(0): Initializing HW Cursor
[    24.147] (II) intel(0): RandR 1.2 enabled, ignore the following RandR disabled message.
[    24.152] (==) intel(0): DPMS enabled
[    24.152] (==) intel(0): Intel XvMC decoder enabled
[    24.152] (II) intel(0): Set up textured video
[    24.152] (II) intel(0): Set up overlay video
[    24.152] (II) intel(0): [XvMC] i965_xvmc driver initialized. [    24.152] (II) intel(0): direct rendering: DRI2 Enabled
[    24.152] (--) RandR disabled
[    24.152] (II) Initializing built-in extension Generic Event Extension [    24.153] (II) Initializing built-in extension SHAPE
[    24.153] (II) Initializing built-in extension MIT-SHM
[    24.153] (II) Initializing built-in extension XInputExtension [    24.153] (II) Initializing built-in extension XTEST
[    24.153] (II) Initializing built-in extension BIG-REQUESTS
[    24.153] (II) Initializing built-in extension SYNC
[    24.153] (II) Initializing built-in extension XKEYBOARD
[    24.153] (II) Initializing built-in extension XC-MISC
--
[    24.153] (II) Initializing built-in extension RENDER
[    24.153] (II) Initializing built-in extension RANDR
[    24.153] (II) Initializing built-in extension COMPOSITE
[    24.153] (II) Initializing built-in extension DAMAGE
[    24.430] (II) AIGLX: enabled GLX_MESA_copy_sub_buffer
[    24.430] (II) AIGLX: enabled GLX_INTEL_swap_event
[    24.430] (II) AIGLX: enabled GLX_SGI_swap_control and GLX_MESA_swap_control
[    24.430] (II) AIGLX: enabled GLX_SGI_make_current_read
[    24.430] (II) AIGLX: GLX_EXT_texture_from_pixmap backed by buffer objects [    24.430] (II) AIGLX: Loaded and initialized /usr/lib64/dri/i965_dri.so [    24.430] (II) GLX: Initialized DRI2 GL provider for screen 0

--
Hanno Böck              Blog:           http://www.hboeck.de/
GPG: 3DBD3B20           Jabber/Mail:    [email protected]

http://schokokeks.org - professional webhosting

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#2 lore raw

Am Mittwoch 28 Juli 2010 schrieb Matt Turner:

Quoted 7 lines · Hanno Böck
Quoted 6 lines

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

Quoted 7 lines
Quoted 5 lines

From the logs everything seems to be fine. Any ideas what I could look for what's wrong here?

kernel 2.6.34-rc6, xorg-server-1.8.2, xf86-video-intel-2.12.0,T61 laptop, card

The most important piece of software here is Mesa. What version of Mesa?

Sorry, I forgot, latest 7.8.2.

Signature · 4 lines
Hanno Böck		Blog:		http://www.hboeck.de/
GPG: 3DBD3B20		Jabber/Mail:	[email protected]

http://schokokeks.org - professional webhosting

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw
Quoted 11 lines · Hanno Böck

Am Mittwoch 28 Juli 2010 schrieb Matt Turner:

Quoted 10 lines · Hanno Böck
Quoted 9 lines

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

From the logs everything seems to be fine. Any ideas what I could look for what's wrong here?

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

kernel 2.6.34-rc6, xorg-server-1.8.2, xf86-video-intel-2.12.0,T61 laptop, card

The most important piece of software here is Mesa. What version of Mesa?

Sorry, I forgot, latest 7.8.2.

When I tried Mesa 7.8 from Debian experimental I got stuttering 3D output, at least in neverball and glxgears. I tried libdrm2 2.4.18, 2.4.21, kernel 2.6.34, 2.6.35-rc5, and the Intel driver 2.11 and 2.12, without a change to the stuttering. The Xserver is 1.7.7. the FPS output in glxgears was even higher than with Mesa 7.7, but it looked ugly because of the stuttering. Do you have a chance to downgrade to Mesa 7.7?

Regards,
Tino

#4 lore raw

Am Donnerstag 29 Juli 2010 schrieb Tino Keitel:

Quoted 7 lines

When I tried Mesa 7.8 from Debian experimental I got stuttering 3D output, at least in neverball and glxgears. I tried libdrm2 2.4.18, 2.4.21, kernel 2.6.34, 2.6.35-rc5, and the Intel driver 2.11 and 2.12, without a change to the stuttering. The Xserver is 1.7.7. the FPS output in glxgears was even higher than with Mesa 7.7, but it looked ugly because of the stuttering. Do you have a chance to downgrade to Mesa 7.7?

Tried 7.7.1, no change in performance.

Signature · 4 lines
Hanno Böck		Blog:		http://www.hboeck.de/
GPG: 3DBD3B20		Jabber/Mail:	[email protected]

http://schokokeks.org - professional webhosting

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#5 lore raw
Quoted 9 lines · Hanno Böck

Am Donnerstag 29 Juli 2010 schrieb Tino Keitel:

Quoted 7 lines

When I tried Mesa 7.8 from Debian experimental I got stuttering 3D output, at least in neverball and glxgears. I tried libdrm2 2.4.18, 2.4.21, kernel 2.6.34, 2.6.35-rc5, and the Intel driver 2.11 and 2.12, without a change to the stuttering. The Xserver is 1.7.7. the FPS output in glxgears was even higher than with Mesa 7.7, but it looked ugly because of the stuttering. Do you have a chance to downgrade to Mesa 7.7?

Quoted 7 lines

Tried 7.7.1, no change in performance.

--
Hanno Böck              Blog:           http://www.hboeck.de/
GPG: 3DBD3B20           Jabber/Mail:    [email protected]

http://schokokeks.org - professional webhosting

You're saying it's slow, but in relation to what? Were previous versions faster, or is the comparison being made between hardware?

Matt

#6 lore raw

Am Donnerstag, den 29.07.2010, 12:24 -0400 schrieb Matt Turner:

Quoted 10 lines · Hanno Böck
Quoted 9 lines

Am Donnerstag 29 Juli 2010 schrieb Tino Keitel:

Quoted 7 lines

When I tried Mesa 7.8 from Debian experimental I got stuttering 3D output, at least in neverball and glxgears. I tried libdrm2 2.4.18, 2.4.21, kernel 2.6.34, 2.6.35-rc5, and the Intel driver 2.11 and 2.12, without a change to the stuttering. The Xserver is 1.7.7. the FPS output in glxgears was even higher than with Mesa 7.7, but it looked ugly because of the stuttering. Do you have a chance to downgrade to Mesa 7.7?

Quoted 4 lines
Quoted 1 line

Tried 7.7.1, no change in performance.

You're saying it's slow, but in relation to what? Were previous versions faster, or is the comparison being made between hardware?

In his original post he wrote that his old r200 card was faster in the q3demo benchmark [1].

Thanks,

Paul

[1] http://lists.freedesktop.org/archives/intel-gfx/2010-July/007538.html

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#7 lore raw
In reply to Matt Turner

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Am 29.07.2010 18:24, schrieb Matt Turner:

Quoted 2 lines

You're saying it's slow, but in relation to what? Were previous versions faster, or is the comparison being made between hardware?

Well I see the same problem here, using

OpenGL renderer string: Mesa DRI Intel(R) 965GM GEM 20100328 2010Q1 OpenGL version string: 2.1 Mesa 7.8.2

I get stuttering when scrolling in warzone2100. I've been using the same hardware for years, and the problem appeared some time this month, AFAIR around the time I upgraded Mesa from 7.7 to 7.8, but it could have some other cause.

Philipp
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkxRvPUACgkQbtUV+xsoLpq+QACg7HqYAaBoVVjF/+dDcmfHeBD5 puYAoKab5N4Q2uFpjDKl5CRyQGiZstFq
=CMrQ
-----END PGP SIGNATURE-----

#8 lore raw
In reply to Hanno Böck
Quoted 8 lines · Hanno Böck

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

From the logs everything seems to be fine. Any ideas what I could look for what's wrong here?

Sounds like you're hitting the vsync limit. I wouldn't know on my old 855 since no game/3D app ever gets close to 60fps anyway.

Try comparing with glxgears: if it yields 60fps, then it's the vsync. If it's some other frame rate, then never mind.

Cheers,

Rémi
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#9 lore raw

Hi,
i have the same problems with 2.11 and 2.12 intel driver and mesa 7.8. In 3D application i have annoying stuttering and in glxgears (and others 3d application) i don't obtain more than 60fps (i think is active vsync). The stuttering problem is mitigated with mesa 7.9 (but is not solved), the vsync problem is solved with downgrade to 2.10 drivers and i would to say how can i disable vsync in new driver? i watched the configure compile options of mesa and drv-intel packages, but i didn't find anything. I use Fedora 13 Amd64.

Greetings.

#10 lore raw
Quoted 4 lines · Camassa Valerio

the vsync problem is solved with downgrade to 2.10 drivers and i would to say how can i disable vsync in new driver? i watched the configure compile options of mesa and drv-intel packages, but i didn't find anything.

It's not a compile-time option. Set the environment variable vblank_mode to 0 to disable vsync, like so:

 $ vblank_mode=0 glxgears

Or set it with the driconf tool.

Signature · 4 lines
Cheers,
Sven Arvidsson
http://www.whiz.se
PGP Key ID 760BDD22

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#11 lore raw

I tried to change vblank_mode with driconf and with set param
vblank_mode, but i obtain the same fps with 2.11 and 2.12 drivers:

Quoted 1 line

311 frames in 5.0 seconds = 62.175 FPS

With 2.12 glxgears rates ~700FPS.

#13 lore raw
Quoted 11 lines · Valerio Camassa

 I tried to change vblank_mode with driconf and with set param vblank_mode, but i obtain the same fps with 2.11 and 2.12 drivers:

Quoted 2 lines

311 frames in 5.0 seconds = 62.175 FPS

With 2.12 glxgears rates ~700FPS.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

You need to add a seperate dri2 device to .drirc for vblank_mode to work there, driconf doesn't work right

    <device screen="0" driver="dri2">
        <application name="Default">
            <option name="vblank_mode" value="0" />
        </application>
    </device>

my full ~/.drirc for example

<driconf>

    <device screen="0" driver="dri2">
        <application name="Default">
            <option name="vblank_mode" value="0" />
        </application>
    </device>
    <device screen="0" driver="i915">
        <application name="Default">
            <option name="force_s3tc_enable" value="false" />
            <option name="no_rast" value="false" />
            <option name="always_flush_cache" value="false" />
            <option name="stub_occlusion_query" value="false" />
            <option name="always_flush_batch" value="false" />
            <option name="bo_reuse" value="1" />
            <option name="texture_tiling" value="true" />
            <option name="early_z" value="false" />
            <option name="allow_large_textures" value="2" />
            <option name="fragment_shader" value="false" />
        </application>
    </device>

</driconf>
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#14 lore raw

I tried to change .drirc to add the xml for dri2 but glxgears still
rating ~60FPS. Also with "$ vblank_mode=0 glxgears" it rates ~60FPS.

Il 01/08/2010 17:26, Robert Hooker ha scritto:

Quoted 45 lines · Valerio Camassa
Quoted 9 lines

I tried to change vblank_mode with driconf and with set param vblank_mode,
but i obtain the same fps with 2.11 and 2.12 drivers:

Quoted 1 line

311 frames in 5.0 seconds = 62.175 FPS

With 2.12 glxgears rates ~700FPS.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

You need to add a seperate dri2 device to .drirc for vblank_mode to work there, driconf doesn't work right

     <device screen="0" driver="dri2">
         <application name="Default">
             <option name="vblank_mode" value="0" />
         </application>
     </device>

my full ~/.drirc for example

<driconf>

     <device screen="0" driver="dri2">
         <application name="Default">
             <option name="vblank_mode" value="0" />
         </application>
     </device>
     <device screen="0" driver="i915">
         <application name="Default">
             <option name="force_s3tc_enable" value="false" />
             <option name="no_rast" value="false" />
             <option name="always_flush_cache" value="false" />
             <option name="stub_occlusion_query" value="false" />
             <option name="always_flush_batch" value="false" />
             <option name="bo_reuse" value="1" />
             <option name="texture_tiling" value="true" />
             <option name="early_z" value="false" />
             <option name="allow_large_textures" value="2" />
             <option name="fragment_shader" value="false" />
         </application>
     </device>

</driconf>

#15 lore raw
In reply to Valerio Camassa
Quoted 6 lines · Valerio Camassa

 I tried to change vblank_mode with driconf and with set param vblank_mode, but i obtain the same fps with 2.11 and 2.12 drivers:

Quoted 2 lines

311 frames in 5.0 seconds = 62.175 FPS

With 2.12 glxgears rates ~700FPS.

Look, as Owain pointed out there is no vsync problem - it's a feature. Why would you want glxgears to produce 700 images per second when your monitor is only updated 60 times a second?

Geir Ove

#16 lore raw
In reply to Valerio Camassa

Am Samstag, den 31.07.2010, 10:36 +0200 schrieb Valerio Camassa:

Quoted 5 lines

I tried to change vblank_mode with driconf and with set param vblank_mode, but i obtain the same fps with 2.11 and 2.12 drivers:

Quoted 1 line

311 frames in 5.0 seconds = 62.175 FPS

With 2.12 glxgears rates ~700FPS.

62.175 FPS ≠ 700 FPS

Why do you write it is the same?

Thanks,

Paul

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#18 lore raw
In reply to Hanno Böck
Quoted 5 lines · Hanno B??ck

Hi,

it seems I'm currently getting insanely slow 3d performance on my system. q3demo benchmark (run with ` to start console, timedemo 1, demo demo001) gives me 57 fps. I think even my old r200 got more than 100 there.

I'll happily bet that you have a screen refresh rate of 60hz.

Your rendering is synchronized to vertical blank of your screen. Move along, nothing to see here.

IIRC the patch that meant mesa take heed of the vblank_mode environment variable was not in 7.8.2, but may be found in git as commits:
234286c0f8b7d30ed49223c648d4c73c1a517ab3 and 45e2b51c853471b79004a954ce3092a253b20b77

-0-

Signature · 4 lines
Brain, v. [as in "to brain"]:
	To rebuke bluntly, but not pointedly; to dispel a source of
	error in an opponent.
		-- Ambrose Bierce, "The Devil's Dictionary"