11 Versions Of WINE Benchmarked

[Tillin9]

Is it possible to perform the benchmarks with an ATI/AMD graphics card with the 8.02 fglrx driver? I am mostly interested to know if the benchmarks actually run with this graphics card and driver.

Thanks,
Gordy

Depends on your card. ATI/AMD has more or less unofficially dropped support for r300 cards and fglrx doesn't really seem to like my 9700 Pro at all, having all kinds of artifacts. My X850 at work, however, runs fine with fglrx. My guess is that if you have a newer card, and fglrx works for you, WINE will work too.

[Vadi]

Hm, maybe it's compiling from source that counts, but I ran 3DMark2001 SE on my nVidia GeForce 8600, 512MB, 169 drivers and my results were lower than yours roughly by 20%.

(don't care much atm, don't do any gaming on wine)

[Gordy]

Is it possible to perform the benchmarks with an ATI/AMD graphics card with the 8.02 fglrx driver? I am mostly interested to know if the benchmarks actually run with this graphics card and driver.

Thanks,
Gordy

[Thetargos]

I didn't have to do anything special to wine for it run 3DMArk. I find it kind of ironic, though, that the "simple" shaders of the Pixel Shaders and Nature tests aren't correctly displayed, but the Advanced Pixel Shaders test runs flawlessly (which test is indeed more advanced for Wine, I wonder?)

I think I'll update the AppDB of 3DMark2001

[Tillin9]

I had the same artifacts under fglrx when it still worked with my card, Radeon 9700 Pro. My guess is WINE doesn't fully support Direct3D shaders. WINE is only DirectX 7 feature complete and while most of 8, 9 and some of 10 are supported, maybe some random (and rarely used by real games) shader extensions are not.

Might want to update the 3DMark 2001 AppDb: http://appdb.winehq.org/objectManage...rsion&iId=4544 Some people posted later versions and didn't mention artifacts.

Also, did they finally fix the bug where ntoskrnl.exe and winedevice.exe need to be disabled for 3D Mark to run? My 9.55 doesn't want to start without the library overrides, but as I upgraded Debian after building WINE (and don't really feel like rebuilding 9.55 when I know 9.56 is due soon), it may just be me.

[Thetargos]

All of those tests you are referring to are known to not be affected by CPU speed, as they are mainly vid card benches. Whenever I have run them in Windows, overclocking my CPU will rarely change the scores and even if it does, the difference is negligible. Even so, the cpu utilization is always pegged in Windows.

I understand that reasoning... in Windows, under Wine, however, since the translation has to occur, I thought it would be higher (something in the realm of 20%, I was kind of expecting to see)

[Max Spain]

It was interesting to note that when running the tests in a window, the CPU utilization was amazingly low. I had assumed that since wine translated DirectX shader instructions to OpenGL it would toll the CPU more, it did not (CPU utilization in GKrellm was of about 5% per core )

All of those tests you are referring to are known to not be affected by CPU speed, as they are mainly vid card benches. Whenever I have run them in Windows, overclocking my CPU will rarely change the scores and even if it does, the difference is negligible. Even so, the cpu utilization is always pegged in Windows.

[Thetargos]

Ok, here are the shots I took. They link to 90% quality 1x1x1 JPEGs.

Like I said in the previous post, there are missing effects in the Nature and Pixel Shaders tests (as you can see bellow), but not so in the Advanced Pixel Shaders test... Is the same for others as well?

System:

CPU: AMD Athlon X2 5600+
GPU: nVidia 8500 GT 512MB
RAM: 2GB 800MHz DDR2

Software:

OS: Fedora 8 x86_64
Driver: NVIDIA 169.09 packaged from Livna
Wine: 0.9.54

It was interesting to note that when running the tests in a window, the CPU utilization was amazingly low. I had assumed that since wine translated DirectX shader instructions to OpenGL it would toll the CPU more, it did not (CPU utilization in GKrellm was of about 5% per core )

3DMark 2001 SE

Nature



Pixel Shaders


Advanced Pixel Shaders



High Quality (PNG) shots:

Nature 1
Nature 2
Pixel Shaders
Advanced Pixel Shaders 1
Advanced Pixel Shaders 2

[Thetargos]

When the previous "7 versions of Wine bechmarked" article came out, I went and fetched 3DMark 2001 SE. All tests got comparable results to the article, however I noticed two tests had some "artifacts" or rather missing elements, particularly the Pixel Shaders test and the Nature test. Both these test showed mssing element sin the shader of the water, where the water was simply a shiny mirror-like surface (missing the wrinkles in the water). Was this the case too for the tests performed by Phoronix? I'll see if I'm able to pull a screenshot or two.

[phoronix]

Phoronix: 11 Versions Of WINE Benchmarked

Last December we had published benchmarks of seven versions of WINE, which covered up through the WINE 0.9.50 release. We had used two versions of Futuremark's 3DMark suite for testing, and with that we had found the performance to be stable in some cases while in later WINE releases we had found some performance losses. With the WINE project on a consistent two-week release cycle, we are looking at the WINE 3D performance and this time going back with the past eleven releases.

http://www.phoronix.com/vr.php?view=11861