Gaming Benchmarks: Windows 7 vs. Ubuntu Linux

[locovaca]

I went on and ran the benchmarks myself to check my doubts about Ubuntu having unfairly compositing enabled in Michael's benchmarks.

It is not unfair if Ubuntu ships with Compositing enabled, which they do. Just because they have not implemented a disable on demand feature as Windows has speaks more to Microsoft's tuning of their OS for a wider audience than Ubuntu.

Again, if this is a huge issue, file a bug report with Ubuntu. Until Ubuntu changes their policy, Michael won't (and shouldn't) touch the setting in his benchmarks.

[curfew]

I went on and ran the benchmarks myself to check my doubts about Ubuntu having unfairly compositing enabled in Michael's benchmarks.

As my operating systems I've got Windows 7 Professional and Arch Linux (using Unstable repo), both 64-bit. Arch has Linux 2.6.35 kernel and latest beta drivers from Nvidia, Windows 7 drivers were whatever the Windows Update fed me. I forgot to install the latest Windows drivers from nvidia.com, sorry. For the hardware part, my system consists of an Intel Core 2 Duo E7200, a GeForce 7900 GS and 4 GBs of RAM.

I chose to do Lightsmark 2008 and Nexuiz. The Nexuiz benches were ran with 32-bit binaries due to the fact Alientrap doesn't supply us with a 64-bit build for Windows. As the benchmark I executed timedemo using a built-in demo "demo1" (run command 'timedemo demos/demo1' in the Nexuiz console).

I'm a bit suspicious of the validity of my Lightsmark results because the benchmark did run in windowed mode on Linux but in fullscreen on Windows. Therefore I manually changed my desktop resolution to be that of the test's and that resulted as a slightly higher FPS. Dunno.

The results were:

*** Lightsmark 2008
*** Drivers: 0x AA, 0x AF forced
Windows: 96.2 fps
Linux: 98.7
Linux*: 102.6

*** Lightsmark 2008
*** Drivers: 4x AA, 4x AF forced
Windows: 72.4 fps
Linux: 72.3
Linux*: 77.2

* Resolution/fullscreen correction

*** Nexuiz: default settings
*** Drivers: 0x AA, 0x AF forced
Seconds -- fps min | avg | max
Windows: 31 -- 39 | 64 | 121
Linux: 30 -- 39 | 68 | 122

*** Nexuiz: maximum detail ('ultra' preset + everything else maxed)
*** Drivers: auto settings
Seconds -- fps min | avg | max
Windows: 120 -- 9 | 17 | 32
Linux: 114 -- 10 | 18 | 36

*** Nexuiz: 'normal' preset
*** Drivers: auto settings
Windows: 48 -- 26 | 40 | 67
Linux: 30 -- 38 | 67 | 121 ??? This must've been a Nexuiz preset mismatch, insane difference.

Short: Linux won every benchmark. I ran all the benchmarks 2-3 times and the differences were minimal and nothing was left for interpretations.

[blackshard]

Now, it is my understanding that Windows no longer has to convert all OpenGL instructions to Direct3D, thus they can get similar performance, but to beat Linux by this much... again, I am guessing this is b/c these tests aren't making extensive use of double precision.
Does anyone know something concrete?

Best/Liam

Windows never had to convert OpenGL to D3D.
Every OpenGL activity is care of the video card driver. ATi and Nvidia have their fully optimized OpenGL drivers that directly call the hardware, instead less important video chip vendors are used to do OpenGL wrapping to D3D (like SiS with their integrated video cards).

With Vista and 7 things changed to kill OpenGL, but I think video chip vendors have found the way long time ago to avoid stupid Microsoft impositions.

[V!NCENT]

Looks like Nvidious cares more about their Windows users than their Linux users. Now what are the Nvidious fanboys going to say about their beloved blob?

They know that nVidia doesn't care because they know nVidia doesn't release spec. They don't care about that because otherwise they would have bought Radeons. So they will not respond to your trolling. Especialy not because they care about what works right now and they know that Catalyst in that regard is no match for the nVidia blob.

[jmcharron]

Looks like Nvidious cares more about their Windows users than their Linux users. Now what are the Nvidious fanboys going to say about their beloved blob?

Lol!!! Nvidia's driver performance is basically uniform. I can vouch for that...

http://www.phoronix.com/forums/showthread.php?t=24798

[RealNC]

Looks like Nvidious cares more about their Windows users than their Linux users. Now what are the Nvidious fanboys going to say about their beloved blob?

Go away, troll. The results mean that the Linux blob is as good as the Windows one. Unless you're blind.

Shoo. Go away.

[stan]

Looks like Nvidious cares more about their Windows users than their Linux users. Now what are the Nvidious fanboys going to say about their beloved blob?

[RealNC]

This is probably why Windows will become a major gaming platform.

Are you living under a stone? Windows is and has been a major gaming platform for ages. Even MS-DOS was a major gaming platform.

[kraftman]

So Michael ran all the tests with Compiz enabled while Windows didn't have compositing on? Insane.

This is the whole point. There should be some note Ubuntu has compiz enabled. While it has compiz enabled I don't understand the Phoronix conclusion:

With Lightsmark and Nexuiz when more taxing on the graphics card the NVIDIA graphics driver under Windows moved forward, but with the most demanding Unigine tests the performance was about the same. Of course, these results are just representative with regard to NVIDIA's proprietary driver on both platforms,

Nvidia drivers moved forward under Windows, because... Ubuntu has compiz enabled?

[locovaca]

"The stock Ubuntu Lucid packages besides NVIDIA's proprietary driver remained with their default versions and configuration."

So Michael ran all the tests with Compiz enabled while Windows didn't have compositing on? Insane. Or does Ubuntu unplug Compiz upon running a 3D application? I don't think Compiz has that kind of a feature built-in.

If the defaults for each OS aren't the same, sounds like you could contribute to making Ubuntu sane then, eh?