Need help solving Xorg slowness.
Ray Van Dolson
rayvd at bludgeon.org
Wed Oct 31 09:58:46 PDT 2007
Hi all, I've been living with slowness / sluggishness on my system for
a long time now. Always have attributed it to lousy Radeon drivers
though, and thought I might see some better performance from the brand
new fglrx drivers that just came out. Alas, I did not and so I thought
it was at least worth a try tracking this down before I either chuck
the card for an NVidia or revert to Windows. :)
The card itself is an ATI Radeon 9800XT (ASUS rebranded) -- I believe
this is using an r350 chip? In any case, while a bit dated it still
should be plenty fast for 2D, but within Xorg it feels at times like
I'm using the vesa driver (I'm not btw :). I've tried using both the
open source 'radeon' driver as well as the fglrx driver in various
forms.
I'm running on Fedora 7 currently with Xorg v1.3.0 + whatever patches
Fedora applies. Kernel is 2.6.22.9. lspci output for my video card:
01:00.0 VGA compatible controller: ATI Technologies Inc RV350 NJ [Radeon 9800 XT] (prog-if 00 [VGA])
Subsystem: ASUSTeK Computer Inc. Unknown device c000
Flags: bus master, stepping, 66MHz, medium devsel, latency 255, IRQ 17
Memory at e8000000 (32-bit, prefetchable) [size=128M]
I/O ports at d800 [size=256]
Memory at d7000000 (32-bit, non-prefetchable) [size=64K]
Expansion ROM at e7fe0000 [disabled] [size=128K]
Capabilities: [58] AGP version 2.0
Capabilities: [50] Power Management version 2
This is on an AMD T-Bird 2100+ (1.3GHz?) with a GB of RAM. ASUS A7V333
motherboard.
Symptoms are that resizing windows and anything "redrawing" really is
laggy. A lot like when you are running the vesa driver althought not
quite that bad perhaps. Firefox in particular when switching between
tabs takes noticeably longer than it does on my NVidia-based system at
work. It just "feels" sluggish.
My current config and X logs are here:
http://www.bludgeon.org/~rayvd/xorg/xorg.conf
http://www.bludgeon.org/~rayvd/xorg/Xorg.0.log
Currently am using fglrx as you can see, but can switch to radeon for
testing if necessary. Performance doesn't change much between the two.
I've also run oprofile on the system, and noted the following (this
during a lot of window resizing and Firefox tab switching and
alt-tabbing):
# opreport
CPU: Athlon, speed 1300.17 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000
CPU_CLK_UNHALT...|
samples| %|
------------------
28251 33.6017 libfb.so
20255 24.0913 no-vmlinux
6029 7.1709 Xorg
CPU_CLK_UNHALT...|
samples| %|
------------------
5999 99.5024 Xorg
30 0.4976 anon (tgid:3414 range:0x110000-0x111000)
# opreport -l
CPU: Athlon, speed 1300.17 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000
samples % image name app name symbol name
20255 24.0913 no-vmlinux no-vmlinux (no symbols)
15714 18.6902 libfb.so libfb.so fbCopyAreammx
4751 5.6508 libfb.so libfb.so fbSolidFillmmx
4732 5.6282 libfb.so libfb.so fbCompositeSolidMask_nx8x8888mmx
Full results here:
http://www.bludgeon.org/~rayvd/xorg/opreport.txt
http://www.bludgeon.org/~rayvd/xorg/opreport_l.txt
>From a quick Google, it appears that some of these calls (fbCopyAreammx
in particular) are non-accelerated fallback functions. Could this be
the source of my problem?
I am using XAA vs EXA (XAA is default) as EXA didn't appear to help
much or didn't work right last time I tried. Can try again.
What else can I do to track this issue down? Happy to open a bug here
and provide whatever help / testing is necessary.
Any thoughts? I _know_ this card has the horsepower to be much faster
(it is very very fast still in Windows).
Thanks in advance!
Ray
More information about the xorg
mailing list