XP vs Win7

[Pho3nix1991]

XP vs Win7 (Benchmarks and Charts) Completed!

I will be running some benchmarks on Windows 7 and Windows XP to give users an idea of how gaming will perform on one OS to the other. All games will be ran using DirectX9 on both and using 180.48 drivers. I will also be running each game through 3 runs on both 0xAA,0xAF and 4xAA,8xAF as most Benchmarks are ran.

Test System:
Motherboard: EVGA 750i SLI FTW
CPU: Intel E7300 2.66GHz
Memory: OCZ Platinum 2x1GB DDR2-1066
Video Card: EVGA 9800GT 512MB
Main HDD: Seagate 7200.10 250GB SATA
Monitor: Samsung 932BW (1440x900@75Hz)

The game testing list may changed dependent on Windows 7 working or not. All of them from my knowledge should work but I will make sure to list which games will not work if any. Also if any benchmarking programs or techniques fail I will swap out for other games that are used in current benchmarks.

Games:
Unreal Tournament 3
Crysis
Crysis: Warhead
Far Cry 2
Fallout 3
Half Life 2: Episode 2


XP Benchmarks:
- 0xAA 0xAF -
Crysis - Min. 21FPS, Avg. 36FPS, Max. 45FPS
Crysis: Warhead (Max) - Min. 7FPS, Avg. 19FPS, Max. 27FPS
Crysis: Warhead (High) - Min. 13FPS, Avg. 37FPS, Max.48FPS
Half-Life 2 Episode 2 - Avg. 129FPS
Unreal Tournament 3 - Avg. 62FPS
Far Cry 2 - Min. 31FPS, Avg. 33FPS, Max. 37FPS
Fallout 3 - Min. 13FPS, Avg. 62FPS, Max. 79FPS
- 4xAA 8xAF -
Crysis - Min. 18FPS, Avg. 27FPS, Max. 34FPS
Crysis: Warhead (Max) - Min. 5FPS, Avg. 11FPS, Max. 19FPS
Crysis: Warhead (High) - Min. 9FPS, Avg. 24FPS, Max.34FPS
Half-Life 2 Episode 2 - Avg. 127FPS
Unreal Tournament 3 0xAA - Avg. 62FPS
Far Cry 2 - Min. 18FPS, Avg. 20FPS, Max. 21FPS
Fallout 3 - Min. 9FPS, Avg. 42FPS, Max. 79FPS

Finished XP benchmarking as of 9:30PM on 1/12/09.
Finished W7 benchmarking as of 7:00PM on 1/14/09.

Win7 Benchmarks:
- 0xAA 0xAF -
Crysis - Min. 20FPS, Avg. 31FPS, Max. 44FPS
Crysis: Warhead (High) - Min. 10FPS, Avg. 31FPS, Max.39FPS
Half-Life 2 Episode 2 - Avg. 121FPS
Unreal Tournament 3 - Avg. 62FPS
Far Cry 2 - Min. 36FPS, Avg. 39FPS, Max. 45FPS
Fallout 3 - Min. 39FPS, Avg. 53FPS, Max. 62FPS
- 4xAA 8xAF -
Crysis - Min. 5FPS, Avg. 24FPS, Max. 33FPS
Crysis: Warhead (High) - Min. 4FPS, Avg. 21FPS, Max.30FPS
Half-Life 2 Episode 2 - Avg. 119FPS
Unreal Tournament 3 0xAA - Avg. 62FPS
Far Cry 2 - Min. 26FPS, Avg. 29FPS, Max. 31FPS
Fallout 3 - Min. 23FPS, Avg. 45FPS, Max. 63FPS

Charts

 

[johngalt]

Do realize that some of those benches, especially the gaming, may not be completely fair, as XP does not have DirectX 10 support - be sure to mention that in your results.

 

[Halaster]

Yep, make sure when doing a FPS comparison to do a fair one. Run the game in DX9 mode for an accurate comparison. Games like Crysis can be ran in either mode.

 

[Pho3nix1991]

I will be running all games in DirectX9 to show a fair comparison since DX10 Benchmarks would be completely irrelevant and I'm currently using HOCBench applications and Fraps to run these Benchmarks.

 

[johngalt]

Good call, m8. I am interested to see the results.

Also, did you do the XP benches on a vanilla XP install?

 

[I2aMpAnT]

Do realize that some of those benches, especially the gaming, may not be completely fair, as XP does not have DirectX 10 support - be sure to mention that in your results.

Shouldn't be also be fair and recognize that XP is SP3 and Windows 7 is still beta?

 

[johngalt]

Well, if you can find me a copy of XP Beta 1 I'll gladly make benches

Seriously, though, that is the point - to minimize the number of differences as low as possible - IOW to get it as close to an OS comparison as possible.

 

[I2aMpAnT]

Well, if you can find me a copy of XP Beta 1 I'll gladly make benches

Seriously, though, that is the point - to minimize the number of differences as low as possible - IOW to get it as close to an OS comparison as possible.

No, stick with SP3 XP. Then people will see how much better Windows 7 already is than XP when it's still in Beta form.

 

[johngalt]

I knew you'd see the light

 

[thegr8anand]

Good one dude. I have just installed Win 7 2 days ago and I have only tried L4D on both and am amazed that its running so well just as good as XP. I didn't want to change from Vista as I liked it but for games I changed to XP Pro Lite. But from what I have seen till now, it's looking great for 7.

Although now you have started the benches but 181.20 seem to be great as I am using them on 7 and are supposed to provide good improvements.

 

[Pho3nix1991]

For now I'm just going to stick with 180.48 since that is what I started with. I will come back and revisit the tests with the new 181 drivers later.

Also my point of comparison right now is to show gaming performance difference between Windows 7 as of this exact moment and Windows XP SP3 as of this exact moment. Yes there will most likely be performance changes for Windows 7 (hopefully for the better) but I think people will be happy with these benchmarks as opposed to no benchmarks at all.

 

[I2aMpAnT]

I knew you'd see the light

lol what light should I be seeing?

 

[Pho3nix1991]

lol what light should I be seeing?

The headlights on his car before he hits you

 

[I2aMpAnT]

The headlights on his car before he hits you

No it's my FGM-Javelin missle coming down from the sky at his car that he should be seeing

 

[Pho3nix1991]

I've finished all of the benchmarks and the charts! Hope this information is useful to everyone! =)

 

[Everlong]

Thanks for these.

Doesn't look too bad, not too far behind on XP, considering it's still in beta as well.

A good thing to point out is the minimum fps on XP is a lot lower than Win 7.

 

[splxtreme]

It looks like, the vista drivers are optimized for newer games, more so than xp's. I have been using 7 for about a month now, and each release keeps getting better. I think the fps are close enough and probably better than Vista. I did do one test on my own, and that was on 3dmark 2006. XP had a score of 13198 and 7 had a score of 12982. That is close enough for me. Also, It was an older driver, but I remember Vista had a score in the low 12,000's.

Another thing to note, is that no OS from MS has been faster than its predecessor. I think Win 7 will be the first, I hope.

 

[TheNetAvenger]

A few things to note about these sets of tests...

It is good to see that you are getting good results from Win7; however, these may not be the best reflective set of tests for everyone.

1) The 9600 and 9800 NVidia series currently have several bugs and performance issues with Vista and Win7 drivers. The 185.xx series correct some of these.

2) In addition to the current bugs with the 9xxx NVidia drivers, other cards will yeild significantly different results. For example if you have a card with lower VRAM, Vista or Win7 will pull ahead of XP because of the GPU RAM Virtualization on higher detail texture settings and AA. This means a 128mb card will get more of a boost from the Vista/Win7 WDDM than a card with 512mb of dedicated VRAM. Also shared memory solutions like onboard video also get more of a boost due to how the WDDM handles VRAM.

3) Older NVidia and ATI series are seeing more improvements in Win7 than the 8xxx and newer NVidia and ATI 2xxx and newer cards are. So a user with a Geforce 7900 should be seeing more of a jump in performance in Win7 over XP.

4) Make sure your FPS and video timing are consistent. Because you state you are running at 75Hz, this doesn’t mean the game on XP will conform to that refresh and could be shifting to a default 60Hz, thus giving more FPS due to refresh timings. As you know VSync forcing and other settings like triple buffering make differences, and they are done differently on XP than on Vista and Win7.

5) I would also suggest that you through in a few ‘older’ games that are DirectX8.1 (or even earlier) and OpenGL 1.x in your test mix. (Many gamers are MMO players and still using games written in the DX8 era)

6) Also run a few rounds of performance tests with DX10 enabled on the newer games, and see if even with the more advanced quality settings DX10 turns on, how the FPS hold up in Win7 compared to the DX9 mode in XP. Part of DX10 and DX11 when games get to either point is the performance enhancements in addition to the new lighting and larger textures allowed.

(A Side note about DX10 and DX11: unless a games is DX10 only, it is running on a hybrid DX9 engine for XP compatibility, and this means that it is usually only turning on the visual enhancements and the core engine wasn’t designed around the more efficient DX10 constructs. When games come out that are DX10 ONLY they will not only be ‘prettier’, but will run faster. Remember the Xbox 360 video system is what the Vista WDDM was designed from, and because the hardware is controlled, the Xbox 360 is actually running DX11 level gaming engine libraries. Which means that Win7 just now catches up the XBox 360. Also the DX10 and DX11 being forced on the XBox 360 is what helps it keep good performance numbers even though the hardware is 3 years old.)

Keep up the good work, and if you have another video card to through in your system to recycle the tests, give that a try for people with lower end system wanting to know how they will fair with Win7. Even an older Geforce 6600 level card is good enough for basic testing.

(Our labs are showing most games have more of an advantage across the board on various GPU/CPU combinations than what your tests are showing, so I know Win7 even looking as good as it does from your tests is still being shorted a bit in what it will offer people. –Sadly we can’t publish this information.)

 

[TheNetAvenger]

It looks like, the vista drivers are optimized for newer games, more so than xp's. I have been using 7 for about a month now, and each release keeps getting better. I think the fps are close enough and probably better than Vista. I did do one test on my own, and that was on 3dmark 2006. XP had a score of 13198 and 7 had a score of 12982. That is close enough for me. Also, It was an older driver, but I remember Vista had a score in the low 12,000's.

Another thing to note, is that no OS from MS has been faster than its predecessor. I think Win 7 will be the first, I hope.

Be careful with relying on 3DMark, as it doesn't mean games will reflect the small up and downs of the 3DMark results.

This is based on a lot of things from the benchmarkes being artificial to focusing on things that your game might not even use, or are even outdated ways to handle rendering.

A good example of how 3DMark can fail people is if you use 3DMark2001/2003 it will show a Geforce 6800 or 7800 to be faster than a 8800, even though this is really wrong. This has to do with the unfied shaders in the newer cards not being accurately represented, and also 3DMark2001 and 2003 using differnt texture formats and bump mapping, etc that are just not used in gaming, even older games.

Also use a couple of games to augment your tests, instead of relying on 3Dmark solely. You might find that a new driver makes your 3DMark go down, but pops up your FPS in your favorite games. (I have seen a 3DMark with a new driver drop a few hundred points depending on the version and yet both DX and OpenGL games were pushing 5-10% better FPS.)

Take Care...

 

[Pho3nix1991]

All the games the I benchmarked by the way were all DirectX 9.0. None of them were 8.1 and the oldest game was HL2: EP2 which runs on DX 9.0.

And I do not have an older video card seeing as my 7600GS had died a few months back in my old PC.