I did some apples-to-apples 3d benchmarking today on a dual boot system. It looks offhand like Windows is about twice as fast as Wine on these benchmarks on my system.
cpu: e8400 ram: 4GB video card: nvidia GT 240 Ubuntu 32 10.04 video drivers: 195.36.15 Vista 64 video drivers: 8.17.11.9713
Test procedure:
On Ubuntu, grab latest wine from git, apply patch from http://bugs.winehq.org/show_bug.cgi?id=9210 so 3dmark06 will run. On Windows, install cygwin and a few of its packages (cabextract, wget, svn, etc.) Then on both systems do svn checkout http://winezeug.googlecode.com/svn/trunk/ winezeug sh winezeug sh yagmarktest
That will run the tests in a loop and save a report for each run in the current directory.
Results at http://kegel.com/wine/yagmarkdata/e8400/
Summary: Ubuntu: 6 runs 3dmark2000: min 4145, median 4154, max 4170 3dmark2001: min 18884, median 19261, max 19520 3dmark2006: min 17315, median 18154, max 18498
Vista: 2 runs: 3dmark2000: 8206, 8219 3dmark2006: 30278, 32384
Notes: 1) 3dmark2001 hangs for me on vista 64 bits, so no results for that test on that OS, I had to edit yagmarktest to remove it from the loop. Does it work for anyone else? 2) nvidia drivers occasionally crash X when running or quitting 3dmark2000 or 3dmark06; it's happened to me three times. 3) test script is fragile, don't touch the mouse or keyboard while it's running 4) shadows appear to be missing on wine in firefly forest in 3dmark06 and giant mecha truck test in 3dmark01? 5) format of my log files is ugly and will change.
One note of caution is that video card manufacturers have been known to blatantly cheat on these tests by rigging their drivers with special hacks when they detect 3dmark is running. As far as I know the code may still be there.
Since the Linux drivers probably don't have these hacks, if they are indeed still enabled you wouldn't get an Apples:Apples comparison even on the same hardware.
Thanks, Scott Ritchie
On 05/01/2010 11:08 AM, Dan Kegel wrote:
I did some apples-to-apples 3d benchmarking today on a dual boot system. It looks offhand like Windows is about twice as fast as Wine on these benchmarks on my system.
cpu: e8400 ram: 4GB video card: nvidia GT 240 Ubuntu 32 10.04 video drivers: 195.36.15 Vista 64 video drivers: 8.17.11.9713
Test procedure:
On Ubuntu, grab latest wine from git, apply patch from http://bugs.winehq.org/show_bug.cgi?id=9210 so 3dmark06 will run. On Windows, install cygwin and a few of its packages (cabextract, wget, svn, etc.) Then on both systems do svn checkout http://winezeug.googlecode.com/svn/trunk/ winezeug sh winezeug sh yagmarktest
That will run the tests in a loop and save a report for each run in the current directory.
Results at http://kegel.com/wine/yagmarkdata/e8400/
Summary: Ubuntu: 6 runs 3dmark2000: min 4145, median 4154, max 4170 3dmark2001: min 18884, median 19261, max 19520 3dmark2006: min 17315, median 18154, max 18498
Vista: 2 runs: 3dmark2000: 8206, 8219 3dmark2006: 30278, 32384
Notes:
- 3dmark2001 hangs for me on vista 64 bits, so no results for that
test on that OS, I had to edit yagmarktest to remove it from the loop. Does it work for anyone else? 2) nvidia drivers occasionally crash X when running or quitting 3dmark2000 or 3dmark06; it's happened to me three times. 3) test script is fragile, don't touch the mouse or keyboard while it's running 4) shadows appear to be missing on wine in firefly forest in 3dmark06 and giant mecha truck test in 3dmark01? 5) format of my log files is ugly and will change.
On Sat, May 1, 2010 at 5:26 PM, Scott Ritchie scott@open-vote.org wrote:
One note of caution is that video card manufacturers have been known to blatantly cheat on these tests by rigging their drivers with special hacks when they detect 3dmark is running. As far as I know the code may still be there.
Since the Linux drivers probably don't have these hacks, if they are indeed still enabled you wouldn't get an Apples:Apples comparison even on the same hardware.
Scott, please bottom post on wine mailing lists.
As for the driver hacks, you might try renaming the executable (e.g., 3dmark.exe ->3dmarktest.exe). Some of those hacks only check the executable name, and renaming the executable restores the 'real' behavior. Quake 3 was well known for this..
On Sun, May 2, 2010 at 4:08 AM, Dan Kegel dank@kegel.com wrote:
I did some apples-to-apples 3d benchmarking today on a dual boot system. It looks offhand like Windows is about twice as fast as Wine on these benchmarks on my system.
cpu: e8400 ram: 4GB video card: nvidia GT 240 Ubuntu 32 10.04 video drivers: 195.36.15 Vista 64 video drivers: 8.17.11.9713
Do you happen to have a 4k/5k series ATI card to attempt this test set with as well?
I'd find it interesting to see their relative performance as well as the notable short comings (things not rendered) ATI still have under Wine. Some one asked me the other day how well they'd fair with a new ATI card under Wine and Linux and I had no idea about an answer.
On Sun, May 2, 2010 at 1:03 AM, Edward Savage epssyis@gmail.com wrote:
video card: nvidia GT 240
Do you happen to have a 4k/5k series ATI card to attempt this test set with as well?
I have three nvidia cards: 8500gt, gt220, gt240. (I also have an old ati rage pro which probably isn't worth testing.)
I'd find it interesting to see their relative performance as well as the notable short comings (things not rendered) ATI still have under Wine. Some one asked me the other day how well they'd fair with a new ATI card under Wine and Linux and I had no idea about an answer.
Maybe someone with current ATI hardware can run yagmark following the steps I gave. It's pretty easy. - Dan
When doing such benchmarks also try to write down the frame rates of the individual tests. The 3dmark score is some sort of weighted average (the result of the cpu test in it is also quite important).
From what I remember the scores in 2000/2001/2003 were actually good
but 2005/2006 used to be quite slow for some reason.
Roderick
On Sat, May 1, 2010 at 8:08 PM, Dan Kegel dank@kegel.com wrote:
I did some apples-to-apples 3d benchmarking today on a dual boot system. It looks offhand like Windows is about twice as fast as Wine on these benchmarks on my system.
cpu: e8400 ram: 4GB video card: nvidia GT 240 Ubuntu 32 10.04 video drivers: 195.36.15 Vista 64 video drivers: 8.17.11.9713
Test procedure:
On Ubuntu, grab latest wine from git, apply patch from http://bugs.winehq.org/show_bug.cgi?id=9210 so 3dmark06 will run. On Windows, install cygwin and a few of its packages (cabextract, wget, svn, etc.) Then on both systems do svn checkout http://winezeug.googlecode.com/svn/trunk/ winezeug sh winezeug sh yagmarktest
That will run the tests in a loop and save a report for each run in the current directory.
Results at http://kegel.com/wine/yagmarkdata/e8400/
Summary: Ubuntu: 6 runs 3dmark2000: min 4145, median 4154, max 4170 3dmark2001: min 18884, median 19261, max 19520 3dmark2006: min 17315, median 18154, max 18498
Vista: 2 runs: 3dmark2000: 8206, 8219 3dmark2006: 30278, 32384
Notes:
- 3dmark2001 hangs for me on vista 64 bits, so no results for that
test on that OS, I had to edit yagmarktest to remove it from the loop. Does it work for anyone else? 2) nvidia drivers occasionally crash X when running or quitting 3dmark2000 or 3dmark06; it's happened to me three times. 3) test script is fragile, don't touch the mouse or keyboard while it's running 4) shadows appear to be missing on wine in firefly forest in 3dmark06 and giant mecha truck test in 3dmark01? 5) format of my log files is ugly and will change.
On Sun, May 2, 2010 at 2:07 AM, Roderick Colenbrander thunderbird2k@gmail.com wrote:
When doing such benchmarks also try to write down the frame rates of the individual tests. The 3dmark score is some sort of weighted average (the result of the cpu test in it is also quite important). From what I remember the scores in 2000/2001/2003 were actually good but 2005/2006 used to be quite slow for some reason.
OK, I'll try to parse out the subscores. - Dan
Am 01.05.2010 20:08, schrieb Dan Kegel:
Results at http://kegel.com/wine/yagmarkdata/e8400/
Summary: Ubuntu: 6 runs 3dmark2000: min 4145, median 4154, max 4170 3dmark2001: min 18884, median 19261, max 19520 3dmark2006: min 17315, median 18154, max 18498
Vista: 2 runs: 3dmark2000: 8206, 8219 3dmark2006: 30278, 32384
It looks like you've mixed up the results somewhere. The score should be higher on the lower 3Dmark version. (3dmark 2000 on vista has 30278 points - http://kegel.com/wine/yagmarkdata/e8400/gt240/vista/yagmark-e8400-Vista-2010...)
Cheers Rico
2010/5/2 Rico Schüller kgbricola@web.de:
It looks like you've mixed up the results somewhere. The score should be higher on the lower 3Dmark version. (3dmark 2000 on vista has 30278 points - http://kegel.com/wine/yagmarkdata/e8400/gt240/vista/yagmark-e8400-Vista-2010...)
That'll teach me to edit data tables manually. Next one will be automatically generated. - Dan
-
Austin English -
Dan Kegel -
Edward Savage -
Rico Schüller -
Roderick Colenbrander -
Scott Ritchie