XP vs Vista

Posted at: Tuesday 2nd October 2007 by James Morris

James Morris puts Windows XP and Vista head to head in a gruelling set of benchmarks to see which Microsoft operating system runs your games and applications the quickest.

We combined a 2.66GHz Intel Core 2 Duo E6700 with the now standard 2GB of DDR2 memory. We chose to use Crucial Ballistix PC2-6400 rather than value memory, so we had a little overclocking headroom. These core components were installed in a Gigabyte N680SLi-DQ6 motherboard based on Nvidia's nForce 680i SLI chipset - the high-end gamer's favourite. Although the Gigabyte had decent on-board audio, we disabled this and used a Creative SoundBlaster X-Fi. We wanted to test the implications of Windows Vista regarding sound, thanks to its discontinuation of hardware support for DirectSound3D. We also used an identical pair of 500GB Hitachi Deskstar T7K500 hard disks, one for each of our operating systems, installed with an identical set of testing applications apiece.

As DirectX 10 is so important for Windows Vista, for our game testing we called upon a pair of the latest DirectX 10 graphics cards - one each from both ATI and Nvidia. To keep the pricing, power and memory quotient as close as possible, we pitted a Sapphire Radeon HD 2900XT against a BFG Nvidia GeForce 8800GTS OC 640MB, rather than a GeForce 8800 GTX or Ultra, both of which are considerably more expensive than anything that ATI offers in terms of graphics products. We also wanted to find out if either vendor offers better relative Vista performance. We used Nvidia ForceWare driver 163.44 for both XP and Vista, and ATI Catalyst 7.8 build 50974 (Windows Vista) and build 50960 (Windows XP).

Our starting point was our own Media Benchmarks. Firstly, we ran all three tests that comprise our new 2007 suite of 2D performance tests to provide an overall view of system speed. Then we added the video encoding portion of our 2005 edition benchmarks - a gruelling task for even the latest processors. For a more rounded selection of everyday tasks, we also wanted to assess MP3 encoding. Windows Media Player 11 is available for both XP and Vista, and everyone running Windows is likely to have a copy, so we chose that for our CD ripping test. We also ripped two hours of video from a DVD disc to H.264 MPEG-4 AVC using HandBrake, and copied 2GB of files from one Deskstar to the other. We ran all our 2D benchmarks with a mild overclock, pushing the FSB from 266MHz to 300MHz. This provided a 3GHz Core 2 Duo instead of a 2.66GHz one, and 900MHz rather than 800MHz DDR2. All of these 2D tasks were performed with Nvidia graphics only, with the exception of Maxon Cinebench 9.5; one portion of this calls for hardware OpenGL, so we run that on both Nvidia and ATI graphics card for comparison.

Gaming performance is a fundamentally important issue when comparing Windows XP with Vista, particularly since graphics hardware is addressed in a rather different way when using the latter. We put together a comprehensive collection of current and recent gaming releases, including a few we know to have caused problems for Windows Vista, such as S.T.A.L.K.E.R.: Shadow of Chernobyl. We also ran Need for Speed: Carbon 1.4, Far Cry 1.4, Prey 1.3, F.E.A.R. 1.08, Company of Heroes 1.71, S.T.A.L.K.E.R. 1.0003, The Elder Scrolls IV: Oblivion 1.2.0416UK, and 3DMark06 1.1.0 for good measure. In each case, we set the screen to the 16:10 widescreen resolution of 1,680 x 1,050, turned anti-aliasing on or to 2x (whichever setting was available for the game in question), and anisotropic filtering to 8x (again, where this was available). All detail and graphics rendering settings were pushed to the maximum.