Re: Radeon 9200 is slower than on-board intel 815 graphic card



Hallo Henrik

Henrik Carlqvist <Henrik.Carlqvist@xxxxxxxxxxxx> writes:

Pawel <null7@xxxxx> wrote:
I just recompiled mplayer to i686 architecture, chnaged to AGP 4x,

Where you able to verify that it really got 4x? Did you find something
like this in Xorg.0.log:

(**) RADEON(0): Using AGP 4x mode

glxperf.sh:
OpenGL renderer string: Mesa DRI R200 20041207 AGP 4x x86/MMX/SSE TCL

It can be obtained from /var/log/messages:
Aug 12 15:51:00 localhost kernel: agpgart: Found an AGP 2.0 compliant device at 0000:00:00.0.
Aug 12 15:51:00 localhost kernel: agpgart: Putting AGP V2 device at 0000:00:00.0 into 4x mode
Aug 12 15:51:00 localhost kernel: agpgart: Putting AGP V2 device at 0000:01:00.0 into 4x mode


I noticed is AGP 8x configuration information inconsistency

Note: AGP mode info is different in glxperf.sh script and in /var/log/messages.
Current config:
Option "AGPMode" "8"

glxperf.sh:
OpenGL renderer string: Mesa DRI R200 20041207 AGP 8x x86/MMX/SSE TCL

/var/log/messages:
ug 12 20:03:57 localhost kernel: agpgart: Found an AGP 2.0 compliant device at 0000:00:00.0.
Aug 12 20:03:57 localhost kernel: agpgart: Putting AGP V2 device at 0000:00:00.0 into 1x mode
Aug 12 20:03:57 localhost kernel: agpgart: Putting AGP V2 device at 0000:01:00.0 into 1x mode


For AGP 4x info is consistent. Conclusion: 4x is the highest available for my hardware.



?

At http://www.free3d.org/ there are results from a simple benchmark:

cat /proc/pci | grep VGA || lspci | grep VGA | \
colrm 1 4 ; cat /proc/cpuinfo | \ egrep "model name|MHz" ; xdpyinfo | \
egrep "version:|dimensions|depth of" ; glxinfo | \
egrep -A2 "direct rendering|OpenGL vendor" ; glxgears & sleep 25 ; \
killall glxgears

Here are my results:
bash-3.1$ ~/Scripts/glxperf.sh
0.0 VGA compatible controller: ATI Technologies Inc RV280 [Radeon 9200 SE] (rev 01)
model name : Pentium III (Coppermine)
cpu MHz : 994.404
X.Org version: 7.0.0
dimensions: 1024x768 pixels (302x232 millimeters)
depth of root window: 24 planes
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.2
--
OpenGL vendor string: Tungsten Graphics, Inc.
OpenGL renderer string: Mesa DRI R200 20041207 AGP 4x x86/MMX/SSE TCL
OpenGL version string: 1.3 Mesa 6.4.2

4993 frames in 5.0 seconds = 998.429 FPS

bash-3.1$ ~/Scripts/glxperf.sh
2.0 VGA compatible controller: Intel Corporation 82815 CGC [Chipset Graphics Controller] (rev 04)
model name : Pentium III (Coppermine)
cpu MHz : 994.209
X.Org version: 7.0.0
dimensions: 1024x768 pixels (302x232 millimeters)
depth of root window: 24 planes
direct rendering: No
server glx vendor string: SGI
server glx version string: 1.2
--
OpenGL vendor string: Mesa project: www.mesa3d.org
OpenGL renderer string: Mesa GLX Indirect
OpenGL version string: 1.2 (1.5 Mesa 6.4.2)
570 frames in 5.1 seconds = 111.648 FPS


There are two systems with 9250 cards which both get about 2600 fps. You
can compare your result with these numbers, but the glxgears benchmark is
not a very good benchmark and it doesn't say anything at all about xvideo
performance.

but unfortunatelly animation is still teared.
I decreased data stream to 1Mbps (from 8Mbps).
Quality is poor but I still can see that animation is teared(?!strange).
I will try to play with directfb video filter.

MPlayer usually warns if it can't get high enough RTC interrupt frequency.
Have you done:

echo 1024 > /proc/sys/dev/rtc/max-user-freq

Yes I have


?

I saw another interesting option in man mplayer related to directfb video driver:

buffermode=single|double|triple
Double and triple buffering give best results if you
want to avoid tearing issues. Triple buffering is more
efficient than double buffering as it does not block
MPlayer while waiting for the vertical retrace. Single
buffering should be avoided (default: single).

I will try to use it


regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc8(at)uthyres.com Examples of addresses which go to spammers:
root@xxxxxxxxxxxxx root@localhost

regards,
Pawel
.



Relevant Pages