Testbed and Methods
The notebook’s hard drive was formatted in NTFS before the tests. Then we installed Microsoft Windows XP Professional SP2 with DirectX 9.0c, system drivers (from the included disc), and Windows Media Encoder 9.0 with Windows Media Player 9.0. We also installed Windows Media Player 10.0 for such tests as PCMark 2005 and SYSMark 2004 SE.
The following settings were used for the tests:
- Power-saving services – Off
- Audio subsystem – Off
- Network services – Off
- Maximum screen brightness
- Maximum display resolution (1400x1050)
- Windows Taskbar is Unlocked
- Windows Taskbar hides automatically
- Classic Desktop theme
- No background image on the Desktop
- No screensaver
- Low security level
- Pop-ups blocking disabled
There were two exceptions: we returned to the Windows XP desktop theme for PCMark 2005 since the program required that. And for SYSMark 2004 SE to work normally, we had to roll each parameter back to its default (as they are set right after you install Windows).
Two power modes were used. First, we selected the Always On power mode for maximum performance and the shortest battery life. Then we switched to the Max Battery mode for the maximum battery run-down time.
- Performance benchmarks: synthetic (SiSoftware Sandra 2005, SiSoftware Sandra 2007, PCMark 2004 1.3.0, PCMark 2005 1.2.0), office and multimedia (SYSMark 2004 SE, Business Winstone 2004, Multimedia Content Creation Winstone 2004), and games (3DMark 2003 3.6.0, 3DMark 2005 1.2.0, 3DMark 2006 1.0.2, Quake 3, Quake 4, Unreal Tournament 2003)
- Battery life tests (Battery Eater Pro 2.60)
There are three test modes in Battery Eater:
- Classic (the system is under a high and evenly distributed load)
- Reader’s test (the pages are browsed through each 15 seconds)
- Idle mode
We used the first two modes as they are in Battery Eater, but in the Idle mode (when the test utility doesn’t put any load of its own on the notebook) we played a DVD movie.
If you are staying in touch with the Mobile section of our site, you may have noticed we begin to use SiSoftware Sandra 2007, the software suite for remote testing and diagnosing of personal computers, servers, PDAs, smart-phones, small home and office networks as well as large enterprise networks. The SiSoftware 2007 suite features an updated, enhanced-functionality interface, runs on three platforms (Win32 x86, Win64 x64, WinCE ARM), contains 13 etalon tests independently developed for each of the platforms, 34 informational modules, and supports a large range of hardware components thanks to the developer’s collaboration with Intel, AMD, ATI, SiS and VIA. The program is supported in six languages and has a free Lite version for personal and educational purposes.
We’ll go on using SiSoftware Sandra 2005 for a while, but we’ll abandon it eventually as soon as we collect enough statistical data with the new version of the benchmark.
We’ve also added SYSMark 2004 SE, which is intended to reveal a system’s overall performance under different types of load. It simulates a user who is solving practical tasks in a few popular applications and takes multi-tasking into account. As a result, the benchmark issues a few ratings that are indicative of the system performance under different loads. SYSMark 2004 SE is mainly positioned as a tool for testing desktop systems and includes applications that are not often run on mobile computers. That’s why the results for each test load will be shown and commented upon separately.
We now also use PCMark 2005 which carries on the tradition of complex benchmarks of the series and uses fragments of real-life applications as tests. This makes it somewhat more relevant for end-users as opposed to fully synthetic benchmarks. After running a series of 11 tests on the different components of the system, the program calculates an overall performance score in units called PCMarks. PCMark 2005 can check out a computer at processing HD video and encoding audio, and offers enhanced tests of the CPU and hard disk under multi-threaded load. The overall score is calculated by the formula: PCMark Score = 87 x (the geometric mean of the basic tests), where the geometric mean is calculated as (Result 1 x Result 2 x…)/the number of results.
We’ve also changed the version of Futuremark’s 3DMark suite that we use for testing notebooks. From now on we will benchmark notebooks with discrete graphics cores in 3DMark 2005 and 3DMark 2006. Compared with the previous version, 3DMark 2005 uses Shader Model 2.0x/3.0 instead of Shader Model 1.x, provides full compatibility with Shader Model 2.0, includes more complex tests (over a million polygons per each frame), and employs normal maps. 3DMark 2006 brings support for HDR, Uniform Shadow Maps, and multi-core CPUs. It is overall oriented at Shader Model 3.0, but two out of its four graphics tests work within the Shader Model 2.0 framework. We’ll still be using 3DMark 2001SE Pro for benchmarking integrated graphics cores, though.
And finally, we’ll use the latest version of Quake in addition to the older one. There was no standard demo record in Quake 4 , so we had to create it by ourselves and will use it in every following review of notebooks on our site so that different notebooks could be compared under identical conditions.