Windows XP: What the Numbers Mean
To understand how Windows XP compares with previous versions of Windows, we collected more than 500 test results on 86 different system configurations. Our desktop test-beds included systems with processors ranging from 233-MHz Pentium IIs to 2-GHz P4s. We configured the lower-end systems (less than 866 MHz) with 64MB and 128MB of RAM, and we used 128MB and 256MB of memory on the higher-end computers. The Business Winstone 2001 and Content Creation Winstone 2001 tests were used to measure system performance when running common business programs and working with files. We ran all tests on multiple duplicate systems to ensure accurate and reliable results.
As we expected, Windows XPs performance was very similar to that of Windows 2000, with a few notable exceptions. For instance, Windows XP, which is the heir to Windows Me, proved much better suited than Windows 2000 to gaming and 3-D graphics.
When users install Windows XP, they have to choose between FAT32 and NTFS file formats. Both Home and Professional default to and recommend NTFS. FAT32 configurations generally perform faster, but NTFS handles file usage and access better, thanks to a smaller cluster size.
Windows XP is smart enough to turn off some visual effects to improve performance, based on processor speed. On our 233-MHz systems, 9 of the operating systems 16 visual-effects options were turned off automatically; on the PIII/550s, Windows XP disabled 1 effect; and on the 2-GHz systems, all options were enabled. The user can always go into the System Properties control panel and decide which features to enable or disable.
We extensively tested upgrade versions of Windows XP Home and Professional. Upgrading was seamless, even on the 233-MHz systems. Our benchmark test results were slightly lower on the upgraded systems than on systems with clean installations, though generally the difference was imperceptible and within a reasonable margin of error. When possible, however, its better to make a fresh installation rather than upgrading from an already choking operating system.
Windows XP had no problem recognizing all the different devices that were in our test machines, and the OS tended to perform better when it used digitally signed device drivers. The OS was even able to detect some new devices that arent yet on the market when we simply plugged them in. And Windows XP installed the correct drivers, creating a true plug-and-play experience.
We tried to install and run various games and other software and hardware on our test systems. Most recent games worked, but some older software had installation problems. If you run into this situation, check to see whether a new version of your software or a patch is available to correct the situation. Most of the hardware was easy to install, but we did run into some difficulties when we tried to use an older CD-R drive. By the time you read this, most manufacturers will be rolling out digitally signed drivers for their devices and updated versions of their software.
Compared with most previous Microsoft operating systems, Windows XPs stability is outstanding—on a par with Windows 2000. Unstable drivers are often the cause of system crashes, but Microsofts driver compatibility programs let the company thoroughly test manufacturers drivers before they are available on the market.
Click here to view the results for Business Winstone 2001 and Content Creation Winstone 2001 with Pentium II/233 MHz and Pentium III/550 MHz processors. |
Click here to view the results for Business Winstone 2001 and Content Creation Winstone 2001 with Pentium III, Athlon and P4 processors. |
Click here to view the results for Home and Pro XP Upgrades. |
Start
-Up and Shutdown”>
Start-Up and Shutdown
Online Extra
We tested start-up and shutdown times on all our test systems (for complete results, visit www.pcmag.com/xp). We found an interesting trend on the older systems. On PII/233 systems, Windows XP took as long as the old OSs to boot. Even on the PIII/550s, Windows XP took as long as Windows 98 SE and longer than Windows Me. This was mainly caused by the older BIOS. For the 866-MHz systems and up, Windows XP offers a significant improvement in system start-up, mainly because of its parallel-tasking capability. Shutdown, however, doesnt seem to be Windows XPs forte. In most cases, Windows XP took slightly longer than the other OSs to turn off the system.
Click here to view the results. |
3D WinBench and 3
-D Game Performance”>
3D WinBench and 3-D Game Performance
To test Windows XPs 3-D performance, we ran the 3D WinMark tests from 3D WinBench 2000 Version 1.1, as well as three games: Infogamess Unreal Tournament, Interplay Entertainments MDK2, and id Softwares Quake III. All tests were run at 1,024-by-768 resolution with a 32-bit color depth. The monitor was set to a refresh rate of 85 Hz.
3D WinBench tests the graphics adapter, graphics driver, Direct 3D software, data bus and AGP slot, and all crucial components for games and gamers demanding increasingly realistic imagery. 3D WinBench uses the DirectX 7 API, not DirectX 8.
For Windows XP, we used the drivers that came with the OS. For earlier Windows versions, we downloaded and used the latest graphics card drivers available from the vendors Web sites. Since we ran our tests before Windows XP was released to the public, we could not get updated drivers from the card manufacturers. There should be new drivers available when Windows XP ships, and these will most likely improve performance.
Using the same OS core and drivers, Windows XP Home and Professional performed nearly the same. On the Pentium III systems, Windows XP outperformed the other operating systems. Interestingly, the Pentium 4 performed best with Windows Me. And The 1.4-GHz Athlon scores were lower than expected. We suspect that there may have been a driver conflict between the Athlon systems AGP bus and Windows XPs default drivers.
The Pentium III/866 results were indistinguishable across OSs. The Pentium 4 scores were also close, but there were more noticeable variances. In Unreal Tournament and MDK2, the Windows XP systems fell behind the leader, but they were slightly faster on Quake III.
Click here to view the results of the 3-D Game Performance. |
Click here to view the results of the 3D WinBench. |
Application Load
Application Load Times
When we tested application load times, Windows 2000 was the fastest, often with Windows Me alongside it. We werent surprised that Windows NT 4.0 often took the longest. Notably, while Windows XP didnt load applications as quickly as Windows 2000 and Me, the new OS is designed to monitor usage patterns and optimize your system over time for faster application loading.
We found that on the higher-end systems—the 1.4-GHz Athlon and 2-GHz P4—Windows 2000 and Me, loading applications more quickly, though marginally.
Click here to view the results. |
BatteryMark
BatteryMark
eTesting Labs Business Winstone BatteryMark measures a notebooks battery life by performing many of the same tasks as our Business Winstone 2001 benchmark tests to simulate real-world conditions. All notebooks had either Intel SpeedStep or AMD PowerNow! processors and were run in battery-optimized mode (the lowest clock speed and voltage). We upgraded the preinstalled OSs to Windows XP Pro.
For the most part, there was little difference in battery life between the notebooks running Windows XP and those running their default operating systems. The new OS still uses the same ACPI (Advanced Configuration and Power Interface) power technology found in Windows 2000 and Windows Me. The Dell Inspiron 8100 did have 15 minutes less battery life with Windows XP than with Windows Me.
Click here to view the results. |