TEST PROCESS
Benchmarks | Features | Rankings

To test the motherboards, a constant set of hardware devices and peripherals were used. The test bench comprised of

1. A 30.7 GB IBM Deskstar 75GXP DLTA-307030 ATA/100 hard disk running at 7200 rpm (courtesy Wales Technology)
2. A 128 MB PC-133 Bravish RAM (courtesy Wales Technology)
3. An ASUS V7700 GeForce2 GTS 32 MB DDR SDRAM (courtesy Rashi Peripherals)
4. A Creative 48x CD-ROM drive
5. An Acer 211C 21-inch monitor

For Intel chipset-based motherboards, we used the Intel Pentium III 800EB processor (courtesy Ockam Infotech), while for AMD-based platforms the AMD Thunderbird 800 MHz (courtesy Zeta Technologies) was used. For the ASUS CUBX-E board, we used the Intel Pentium III 700E.

We formatted the hard disk drive before installation. The OS of choice was Windows 98SE. We noted down install times and glitches while installing Windows, as also the IRQ assignments. By keeping the set of critical components constant, we ensured that the scores were influenced only by the motherboard and its I/O architecture.

Benchmarks
To test the motherboards, we ran a series of tests, which looked at all aspects of the motherboard such as the graphics sub-system, memory architecture, etc. Other aspects such as the features and the price were also taken into consideration.

Quake III (v1.17):
This game requires no introduction and has one of the most well-oiled gaming engines. Despite being very video card dependent, it's still the best overall system analyser. In normal mode, one can totally eliminate the video card and it's the CPU, I/O subsystem and the memory bandwidth that drives the game. This gives you a clear picture on how the motherboard is able to translate the bandwidth between the different components. A good fps score indicates that the graphics card is not being bottle-necked due to the motherboard.

Evolva Bump Rolling Demo Build 944:
This Direct3D game is extremely texture-heavy and when run in 1024x768 at 32-bit colour, it totally trashes the memory and the hard-disk sub-system, swapping huge textures from the hard disk to the memory and then to the graphics sub-system. This tests the memory and I/O interface of the motherboard.

3DMARK 2000 v1.1:
This system is heavily oriented towards testing the graphics sub-system and, as such, may seem out of place in a motherboard test. However, by keeping the graphics card constant, you can gauge how the other components affect the score. There was a noticeable difference in scores with the different motherboards and this proves that the total bandwidth offered varies from motherboard to motherboard and can result in real-world performance differences.

The CPU3DMarks test gives you a non-fill-rate limited performance number. Since it runs in a 640x480 resolution, it minimises the effects of video card fill-rate to a great extent and you get a truer picture of how your CPU is performing in 3D games. Here, you can also notice how the CPU codes have been optimised by different motherboards.

SiSoft Sandra 2001 (CPU benchmark):
This new version of the popular benchmark series adds support for the Pentium 4. The CPU benchmark is totally synthetic and does not always represent real-world performance, but can definitely give a rough idea of how the motherboard is performing against other platforms.

SiSoft Sandra 2001 (Memory benchmark):
This benchmark is a sure-fire way to check the memory sub-system of your motherboard. The total memory bandwidth will vary from board to board depending on how the architecture has been optimised. The 'By SPD' detection routine for the memory was kept enabled for all motherboards. By SPD refers to the motherboards method of detecting the speed and CAS settings for the memory.

SiSoft Sandra 2001 (Drive index):
The drive index benchmark was specifically run to test the new onboard ATA 66/100-compliant controllers that are now standard with most motherboards. The drive index consists of a score compiled from random, sequential and buffered read and write speeds and is very heavy on the I/O sub-system.

Content Creation Winstone 2000:
This test places a heavy load on the hard disk drive and memory sub-system while running content-heavy applications. Widely respected, it still is a synthetic and not real-world test. But even then, differences of single decimal points should be considered, as the score scale is very small.