Is 2d acceleration really this bad?

jpilotus · Feb 13, 2005, 08:57 PM

Ok... I've installed a new motherboard (NF7-S) and decided to try the latest (6_8_0-8.8.25) driver out once more in the hopes that it will play nice with the new chipset. I was obliged to disable the ATI internal agpgart, and use the kernel's nvidia agpgart. So far it seems to be stable.

There's just one thing. 2d acceleration is baad.

When switching workspaces, for instance... I see the wallpaper draw before the windows fill in... and scrolling in rxvt takes about 50% CPU time and comes in waves. I found some mention that disabling dri for the fglrx driver with the "no_dri" option would help, at the expense of losing 3d acceleration. It seems not to help at all, though. For the sake of argument, I ran "x11perf -copypixpix500" with the fglrx driver and DRI enabled, then ran it again with the fglrx driver and DRI disabled and, finally, with the X.org generic radeon driver...

EDIT: Oops... just discovered that completely disabling glx and DRI by commenting out the "load "glx"" and "load "dri""' statements in xorg.conf does yield fast 2d acceleration. Please ignore the bit about "no_dri" not working. Disabling DRI does work if you do it correctly.

Another Edit: Gawd... I spoke too soon. Disabling DRI and glx does make the x11perf test run at high speed but, the slow window drawing and slow scrolling issues don't get any better.

Code:

fglrx With DRI Enabled:
==============================================

x11perf - X11 performance program, version 1.5
The X.Org Foundation server version 60801000 on :0.0
from yomomma
Sun Feb 13 20:27:20 2005

Sync time adjustment is 0.0182 msecs.

   3600 reps @   1.5171 msec (   659.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5166 msec (   659.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5174 msec (   659.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5498 msec (   645.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5327 msec (   652.0/sec): Copy 500x500 from pixmap to pixmap
  18000 trep @   1.5267 msec (   655.0/sec): Copy 500x500 from pixmap to pixmap


fglrx With DRI Disabled:
=================================================

x11perf - X11 performance program, version 1.5
The X.Org Foundation server version 60801000 on :0.0
from yomomma
Sun Feb 13 20:32:02 2005

Sync time adjustment is 0.0193 msecs.

   3600 reps @   1.5426 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5425 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5421 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5425 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.5424 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap
  18000 trep @   1.5424 msec (   648.0/sec): Copy 500x500 from pixmap to pixmap

X.org 6.8.1 generic 2d radeon driver:
=====================================

x11perf - X11 performance program, version 1.5
The X.Org Foundation server version 60801000 on :0.0
from yomomma
Sun Feb 13 20:36:31 2005

Sync time adjustment is 0.0159 msecs.

  28000 reps @   0.2014 msec (  4970.0/sec): Copy 500x500 from pixmap to pixmap
  28000 reps @   0.2014 msec (  4960.0/sec): Copy 500x500 from pixmap to pixmap
  28000 reps @   0.2014 msec (  4960.0/sec): Copy 500x500 from pixmap to pixmap
  28000 reps @   0.2014 msec (  4970.0/sec): Copy 500x500 from pixmap to pixmap
  28000 reps @   0.2014 msec (  4970.0/sec): Copy 500x500 from pixmap to pixmap
 140000 trep @   0.2014 msec (  4970.0/sec): Copy 500x500 from pixmap to pixmap

That means that, for the 2d operation being tested here, the fglrx driver is over seven times slower than the X.org generic "radeon" driver!

I know that slow 2d is a known issue with fglrx but, as the subject of my post asks, is it really this bad? Is there anything I can do to speed up 2d with the fglrx driver while still retaining DRI?

Thanks for listening,
Jonathan

Specs are:

CPU: AMD AthlonXP-M 2600+ (Barton)
RAM: 1024 MB PDP SYSTEMS DDR400@2-3-2-11-1T
MOBO: Abit NF7-S Rev. 2.0 Nvidia Nforce2 Ultra-400 Chipset
GPU: Celestica Radeon 9800-Pro 128MB
Linux 2.4.29 (Slackware 10.0)
X11R6.8.1 (X.org source; home built; optimized for Athlon-XP)
ATI Video Driver fglrx_6_8_0-8.8.25
IceWM 1.2.14

skrionius · Feb 14, 2005, 05:21 AM

Hmm. I have a similar HW, except 9600XT. I did not have to disable anything, rather enabling helped. Maybe you do not have XAA (2D acceleration) succesfuly shot up. You should look at xorg.log. (Maybe you have already done so, I am just guessing. )

Albatorsk · Feb 14, 2005, 07:33 AM

Yup, it's slow. Unfortunately, it's so slow that UAE, which I happen to use, really suffers when using fglrx.

Code:

X.org 6.8.2 built in radeon driver:
   8000 reps @   0.6502 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap
   8000 reps @   0.6511 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap
   8000 reps @   0.6507 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap
   8000 reps @   0.6504 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap
   8000 reps @   0.6511 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap
  40000 trep @   0.6507 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap

Xorg 6.8.2 with fglrx driver:
   1600 reps @   4.0022 msec (   250.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.0028 msec (   250.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.0444 msec (   247.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.1877 msec (   239.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.1855 msec (   239.0/sec): Copy 500x500 from pixmap to pixmap
   8000 trep @   4.0845 msec (   245.0/sec): Copy 500x500 from pixmap to pixmap

This was benchmarked on a dual AthlonMP 1800+ and a Radeon 9600.

Albatorsk · Feb 14, 2005, 07:39 AM

Quote:

Originally Posted by skrionius

Hmm. I have a similar HW, except 9600XT. I did not have to disable anything, rather enabling helped. Maybe you do not have XAA (2D acceleration) succesfuly shot up. You should look at xorg.log. (Maybe you have already done so, I am just guessing. )

Perhaps we could see your results when running "x11perf -copypixpix500" with fglrx and the Xorg radeon driver?

skrionius · Feb 14, 2005, 08:06 AM

As soon as I get home I will try it (in about 4 hours). Maybe it is slower than Xorg stock radeon driver, I just did not compare it, because I found it usable, and no slow scrolling and such I have detected so far. Anyways, I am curious if it is slower or not, regardless the fact I didnt find it slow.

jpilotus · Feb 14, 2005, 11:28 AM

Quote:

Originally Posted by Albatorsk

Yup, it's slow. Unfortunately, it's so slow that UAE, which I happen to use, really suffers when using fglrx.

Code:

X.org 6.8.2 built in radeon driver:
   8000 reps @   0.6502 msec (  1540.0/sec): Copy 500x500 from pixmap to pixmap

Xorg 6.8.2 with fglrx driver:
   1600 reps @   4.0022 msec (   250.0/sec): Copy 500x500 from pixmap to pixmap

This was benchmarked on a dual AthlonMP 1800+ and a Radeon 9600.

That's a little bit faster, relatively, than what I see on my system. (6x vs. 7x ratio).

I suppose I haven't come across any apps yet where it's actually debilitating. The scrolling thing in rxvt is more a visual annoyance than a real hindrance to usability. It's just that it's mighty frustrating to experience such dog-slow 2d performance on my fancy-dancy high-powered video card.

Thanks for the input, guys... waiting on those numbers, skrionius.

Delkster · Feb 14, 2005, 04:22 PM

Radeon 8500LE, XFree86 4.3 (Debian), Athlon XP 2600+.

fglrx:

Code:

Sync time adjustment is 0.0275 msecs.

   2400 reps @   2.1282 msec (   470.0/sec): Copy 500x500 from pixmap to pixmap
   2400 reps @   2.1268 msec (   470.0/sec): Copy 500x500 from pixmap to pixmap
   2400 reps @   2.1308 msec (   469.0/sec): Copy 500x500 from pixmap to pixmap
   2400 reps @   2.1284 msec (   470.0/sec): Copy 500x500 from pixmap to pixmap
   2400 reps @   2.1273 msec (   470.0/sec): Copy 500x500 from pixmap to pixmap
  12000 trep @   2.1283 msec (   470.0/sec): Copy 500x500 from pixmap to pixmap

radeon:

Code:

Sync time adjustment is 0.0387 msecs.

  12000 reps @   0.4667 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap
  12000 reps @   0.4668 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap
  12000 reps @   0.4667 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap
  12000 reps @   0.4668 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap
  12000 reps @   0.4670 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap
  60000 trep @   0.4668 msec (  2140.0/sec): Copy 500x500 from pixmap to pixmap

Just for reference.

jpilotus · Feb 14, 2005, 08:06 PM

Quote:

Originally Posted by skrionius

Hmm. I have a similar HW, except 9600XT. I did not have to disable anything, rather enabling helped. Maybe you do not have XAA (2D acceleration) succesfuly shot up. You should look at xorg.log. (Maybe you have already done so, I am just guessing. )

Sorry... I meant to reply to this post this morning, and forgot. Xaa is enabled for the fglrx driver...

Code:

jnthn@:~$ grep -i xaa /var/log/Xorg.0.log
(II) Loading sub module "xaa"
(II) LoadModule: "xaa"
(II) Loading /usr/X11R6/lib/modules/libxaa.a
(II) Module xaa: vendor="X.Org Foundation"
(II) fglrx(0): Using XFree86 Acceleration Architecture (XAA)

/me shrugs

Jonathan

lordvader · Feb 14, 2005, 09:48 PM

I would jump in and say, even with dri disabled, 2d sucks.
Just try viewing some HDTV (1080i). The opensource drivers play it back beautifully, but ATI's drivers really struggle. Heck, they struggle with 576p !

mooninite · Feb 14, 2005, 10:23 PM

Please add notes to Bug #7.

http://ati.cchtml.com/show_bug.cgi?id=7

jpilotus · Feb 14, 2005, 10:53 PM

Quote:

Originally Posted by mooninite

Please add notes to Bug #7.

http://ati.cchtml.com/show_bug.cgi?id=7

Done! Thanks for reminding me about the Bugzilla.

skrionius · Feb 15, 2005, 02:56 AM

OFF Yet I did not have the time to test it, I had to mess up with my kernel upgrading it to 2.6.11-rc4-ck1...sorry! I will make my mobile beeping for me to do the test

JonSvenJonsson · Feb 15, 2005, 07:18 AM

fglrx-64 against r300 (radeon) 64bit:

fglrx-64

Code:

Sync time adjustment is 0.0293 msecs.

   3600 reps @   1.4791 msec (   676.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.4787 msec (   676.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.4894 msec (   671.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.4788 msec (   676.0/sec): Copy 500x500 from pixmap to pixmap
   3600 reps @   1.4865 msec (   673.0/sec): Copy 500x500 from pixmap to pixmap
  18000 trep @   1.4825 msec (   675.0/sec): Copy 500x500 from pixmap to pixmap

and r300/ radeon-64 :

Code:

Sync time adjustment is 0.0402 msecs.

  20000 reps @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap
  20000 reps @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap
  20000 reps @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap
  20000 reps @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap
  20000 reps @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap
 100000 trep @   0.3079 msec (  3250.0/sec): Copy 500x500 from pixmap to pixmap

Just for the record

cheers Jon

skrionius · Feb 16, 2005, 02:46 AM

Quote:

Originally Posted by Albatorsk

Yup, it's slow. Unfortunately, it's so slow that UAE, which I happen to use, really suffers when using fglrx.

Code:

Xorg 6.8.2 with fglrx driver:
   1600 reps @   4.0022 msec (   250.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.0028 msec (   250.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.0444 msec (   247.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.1877 msec (   239.0/sec): Copy 500x500 from pixmap to pixmap
   1600 reps @   4.1855 msec (   239.0/sec): Copy 500x500 from pixmap to pixmap
   8000 trep @   4.0845 msec (   245.0/sec): Copy 500x500 from pixmap to pixmap

This was benchmarked on a dual AthlonMP 1800+ and a Radeon 9600.

Measured: I have 2500+ Barton, and R9600XT, and I have ~450/sec values running the test. You got the worst values here, maybe there is some configuration problem too?

JonSvenJonsson · Feb 16, 2005, 03:53 AM

I think it has something todo with memory-bandwith (not the cards but the main memory). I have an athlon64 with ddr400 ram and i get 676/sec with an radeon9600pro.

cheers Jon

Albatorsk · Feb 16, 2005, 06:58 AM

Quote:

Originally Posted by skrionius

Measured: I have 2500+ Barton, and R9600XT, and I have ~450/sec values running the test. You got the worst values here, maybe there is some configuration problem too?

But then again, everything in your system that can influence this test is faster than in mine. Faster videocard, faster memory, faster CPU.

Besides, I doesn't explain why the open source driver is 6x faster than fglrx on the same computer.

JonSvenJonsson · Feb 16, 2005, 07:06 AM

Well because ati uses an old codebase for the 2D part ??? AFAIK the codebase is based on X-radeon 2D driver, so it would be interesting to see the radeon-driver from lets say XFree-4.2 or so.
cheers Jon

jpilotus · Feb 16, 2005, 11:45 AM

Quote:

Originally Posted by JonSvenJonsson

Well because ati uses an old codebase for the 2D part ??? AFAIK the codebase is based on X-radeon 2D driver...

Is that really true? I don't know what the "X-Radeon" driver is... I presume you mean something developed for an ancient version of X.

/me waves at mtippett

It sure would be nice to have a response from one of the ATI guru developer types around here.

Is the 2d code really some stone age relic? Is it being actively devoloped or, has it been pushed to the wayside to make room for its sexy big sister, OpenGL? I mean... 3d acceleration is nice... it's why I bought a 9800-pro in the first place. Speedy 2d performance is a necessity as well, don't you agree?

mooninite · Feb 16, 2005, 03:14 PM

If you guys could test run it on the new drivers and see if there are improvements it would be a big help.

-- Tested it on a similar system (P4 2.6, radeon 9700) and 8.10 nets the same results as the first poster. Oh well, we can hope for the future.

jpilotus · Feb 16, 2005, 05:17 PM

Quote:

Originally Posted by mooninite

If you guys could test run it on the new drivers and see if there are improvements it would be a big help.

-- Tested it on a similar system (P4 2.6, radeon 9700) and 8.10 nets the same results as the first poster. Oh well, we can hope for the future.

Just got the new driver installed, and I now get 665.0/sec for an earth-shattering 0.09% increase.

Ah well. 3d seems to be shaping up quite well with this release. Maybe one day they'll have time to deal with 2d performance.

Feb 13, 2005, 08:57 PM	#1	Advertisement (Guests Only) Login or Register to remove this ad
jpilotus Radeon HD 6670 Join Date: Sep 2004 Posts: 40	Is 2d acceleration really this bad? Ok... I've installed a new motherboard (NF7-S) and decided to try the latest (6_8_0-8.8.25) driver out once more in the hopes that it will play nice with the new chipset. I was obliged to disable the ATI internal agpgart, and use the kernel's nvidia agpgart. So far it seems to be stable. There's just one thing. 2d acceleration is baad. When switching workspaces, for instance... I see the wallpaper draw before the windows fill in... and scrolling in rxvt takes about 50% CPU time and comes in waves. I found some mention that disabling dri for the fglrx driver with the "no_dri" option would help, at the expense of losing 3d acceleration. It seems not to help at all, though. For the sake of argument, I ran "x11perf -copypixpix500" with the fglrx driver and DRI enabled, then ran it again with the fglrx driver and DRI disabled and, finally, with the X.org generic radeon driver... EDIT: Oops... just discovered that completely disabling glx and DRI by commenting out the "load "glx"" and "load "dri""' statements in xorg.conf does yield fast 2d acceleration. Please ignore the bit about "no_dri" not working. Disabling DRI does work if you do it correctly. Another Edit: Gawd... I spoke too soon. Disabling DRI and glx does make the x11perf test run at high speed but, the slow window drawing and slow scrolling issues don't get any better. Code: fglrx With DRI Enabled: ============================================== x11perf - X11 performance program, version 1.5 The X.Org Foundation server version 60801000 on :0.0 from yomomma Sun Feb 13 20:27:20 2005 Sync time adjustment is 0.0182 msecs. 3600 reps @ 1.5171 msec ( 659.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5166 msec ( 659.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5174 msec ( 659.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5498 msec ( 645.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5327 msec ( 652.0/sec): Copy 500x500 from pixmap to pixmap 18000 trep @ 1.5267 msec ( 655.0/sec): Copy 500x500 from pixmap to pixmap fglrx With DRI Disabled: ================================================= x11perf - X11 performance program, version 1.5 The X.Org Foundation server version 60801000 on :0.0 from yomomma Sun Feb 13 20:32:02 2005 Sync time adjustment is 0.0193 msecs. 3600 reps @ 1.5426 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5425 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5421 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5425 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap 3600 reps @ 1.5424 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap 18000 trep @ 1.5424 msec ( 648.0/sec): Copy 500x500 from pixmap to pixmap X.org 6.8.1 generic 2d radeon driver: ===================================== x11perf - X11 performance program, version 1.5 The X.Org Foundation server version 60801000 on :0.0 from yomomma Sun Feb 13 20:36:31 2005 Sync time adjustment is 0.0159 msecs. 28000 reps @ 0.2014 msec ( 4970.0/sec): Copy 500x500 from pixmap to pixmap 28000 reps @ 0.2014 msec ( 4960.0/sec): Copy 500x500 from pixmap to pixmap 28000 reps @ 0.2014 msec ( 4960.0/sec): Copy 500x500 from pixmap to pixmap 28000 reps @ 0.2014 msec ( 4970.0/sec): Copy 500x500 from pixmap to pixmap 28000 reps @ 0.2014 msec ( 4970.0/sec): Copy 500x500 from pixmap to pixmap 140000 trep @ 0.2014 msec ( 4970.0/sec): Copy 500x500 from pixmap to pixmap That means that, for the 2d operation being tested here, the fglrx driver is over seven times slower than the X.org generic "radeon" driver! I know that slow 2d is a known issue with fglrx but, as the subject of my post asks, is it really this bad? Is there anything I can do to speed up 2d with the fglrx driver while still retaining DRI? Thanks for listening, Jonathan Specs are: CPU: AMD AthlonXP-M 2600+ (Barton) RAM: 1024 MB PDP SYSTEMS DDR400@2-3-2-11-1T MOBO: Abit NF7-S Rev. 2.0 Nvidia Nforce2 Ultra-400 Chipset GPU: Celestica Radeon 9800-Pro 128MB Linux 2.4.29 (Slackware 10.0) X11R6.8.1 (X.org source; home built; optimized for Athlon-XP) ATI Video Driver fglrx_6_8_0-8.8.25 IceWM 1.2.14 Last edited by jpilotus : Feb 13, 2005 at 09:14 PM.

Feb 14, 2005, 05:21 AM	#2
skrionius Sad Radeon Renegade Join Date: Mar 2004 Location: Hungary Posts: 221	Hmm. I have a similar HW, except 9600XT. I did not have to disable anything, rather enabling helped. Maybe you do not have XAA (2D acceleration) succesfuly shot up. You should look at xorg.log. (Maybe you have already done so, I am just guessing. ) __________________ SK - http://zaza.uw.hu/mrp/index.html http://jcrpg.blogspot.com Abit NF7 (nforce2), 2500+ Barton, 1GB, 250G Hitachi HDD ex-owner of an Asus R9600XT - fglrx 8.21.7 \| radeon r300 Mesa CVS Nvidia 8800GTS 320MB gentoo linux (since 2003) kernel 2.6.22

Feb 14, 2005, 08:06 AM	#5
skrionius Sad Radeon Renegade Join Date: Mar 2004 Location: Hungary Posts: 221	As soon as I get home I will try it (in about 4 hours). Maybe it is slower than Xorg stock radeon driver, I just did not compare it, because I found it usable, and no slow scrolling and such I have detected so far. Anyways, I am curious if it is slower or not, regardless the fact I didnt find it slow. __________________ SK - http://zaza.uw.hu/mrp/index.html http://jcrpg.blogspot.com Abit NF7 (nforce2), 2500+ Barton, 1GB, 250G Hitachi HDD ex-owner of an Asus R9600XT - fglrx 8.21.7 \| radeon r300 Mesa CVS Nvidia 8800GTS 320MB gentoo linux (since 2003) kernel 2.6.22

Feb 14, 2005, 10:23 PM	#10
mooninite Radeon HD 6970 Join Date: Jan 2005 Posts: 673	Please add notes to Bug #7. http://ati.cchtml.com/show_bug.cgi?id=7 __________________ CAPS LOCK: ITS LIKE THE CRUISE CONTROL FOR AWESOME

Feb 15, 2005, 02:56 AM	#12
skrionius Sad Radeon Renegade Join Date: Mar 2004 Location: Hungary Posts: 221	OFF Yet I did not have the time to test it, I had to mess up with my kernel upgrading it to 2.6.11-rc4-ck1...sorry! I will make my mobile beeping for me to do the test __________________ SK - http://zaza.uw.hu/mrp/index.html http://jcrpg.blogspot.com Abit NF7 (nforce2), 2500+ Barton, 1GB, 250G Hitachi HDD ex-owner of an Asus R9600XT - fglrx 8.21.7 \| radeon r300 Mesa CVS Nvidia 8800GTS 320MB gentoo linux (since 2003) kernel 2.6.22

Feb 14, 2005, 09:48 PM	#9
lordvader Radeon HD 6670 Join Date: Aug 2003 Posts: 47	I would jump in and say, even with dri disabled, 2d sucks. Just try viewing some HDTV (1080i). The opensource drivers play it back beautifully, but ATI's drivers really struggle. Heck, they struggle with 576p !

Feb 16, 2005, 03:53 AM	#15
JonSvenJonsson Radeon HD 6970 Join Date: Mar 2004 Posts: 581	I think it has something todo with memory-bandwith (not the cards but the main memory). I have an athlon64 with ddr400 ram and i get 676/sec with an radeon9600pro. cheers Jon __________________ Closed source Linux Kernel modules are illegal

Feb 16, 2005, 07:06 AM	#17
JonSvenJonsson Radeon HD 6970 Join Date: Mar 2004 Posts: 581	Well because ati uses an old codebase for the 2D part ??? AFAIK the codebase is based on X-radeon 2D driver, so it would be interesting to see the radeon-driver from lets say XFree-4.2 or so. cheers Jon __________________ Closed source Linux Kernel modules are illegal

Feb 16, 2005, 03:14 PM	#19
mooninite Radeon HD 6970 Join Date: Jan 2005 Posts: 673	If you guys could test run it on the new drivers and see if there are improvements it would be a big help. -- Tested it on a similar system (P4 2.6, radeon 9700) and 8.10 nets the same results as the first poster. Oh well, we can hope for the future. __________________ CAPS LOCK: ITS LIKE THE CRUISE CONTROL FOR AWESOME Last edited by mooninite : Feb 16, 2005 at 04:55 PM.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Tri monitor setup... Poor 2d acceleration, no video acceleration... ATI TV = garbage.	Arin	Radeon Technical Support	4	Mar 5, 2007 10:16 PM
No 3D Acceleration	big-byrd	Linux	5	Feb 28, 2004 05:08 PM
O' where has my acceleration gone!?	elmerc	Radeon Technical Support	7	Nov 13, 2002 10:55 PM
Why do we need 2D acceleration?	lackyboofey	General Radeon Discussion	3	Oct 4, 2002 01:21 PM
No DVD Acceleration	ryanb18	Radeon Technical Support	8	Sep 14, 2002 09:15 PM