Determining overall performance levels for motherboards is a bit more difficult compared to processors or 3D video cards. Hardware and software consistency must be maintained, benchmarks should have some impact on overall system throughput, and the operating system and driver revisions must be standard.
As this is a single-product evaluation, we're trying to keep our main hardware selection consistent, while switching the motherboard component. We are including performance scores from the ASUS A8V Deluxe and MSI K8T Neo2-FIR (K8T800 Pro) for our main benchmark comparison, as well as the MSI K8T Neo-FIS2R (VIA K8T800) and Gigabyte K8NNXP (nForce3 150) to give an idea on the performance advantage of the newer AMD64 platforms.
The Socket 939 boards will use the Athlon 64 3800+, while the Socket 754 platforms get the Athlon 64 3400+. The ratings numbers may seem far apart, but these two CPUs are very close in overall architecture/L2 cache/core speed ratios, and the differences are mostly in the platform and its memory architecture. The hardware components include 2x512-MB of OCZ PC4200 DDR, an ATI Radeon 9800XT video card, and dual Western Digital WD360 drives in RAID 0 Mode. Windows XP Pro SP1 is the operating system, with ATI Catalyst 4.7 video drivers.
The System BIOS is also an important area, and for all benchmarks, the settings have been standardized to a degree. All motherboards have been tested using the highest memory timings allowed, along with other performance settings equalized at the high-end. This tends to reward manufacturers that allow greater tweaking in the BIOS, while maintaining equal footing in terms of the components used.
We do draw the line at bumping up the FSB at standard processor settings, or any other performance enhancing trickery. There are certain System BIOS options that do this very thing, as well as some selections that artificially increase system speeds. These have either been disabled or if needed, an equivalent FSB was selected manually (and confirmed via diagnostic software) to ensure an even playing field.
The benchmark list features a wide range of system, memory, and gaming tests. These include the new revisions of Business Winstone 2004, CC Winstone 2004, and PCMark 2004, along with popular 3D game tests like AquaMark 3, Unreal Tournament 2004 & 2003, and Wolfenstein: Enemy Territory. Our default game benchmark setting is 1024x768x32, which is demanding enough for comparisons and more realistic as an actual gameplay resolution, but still allows the motherboard-CPU-memory subsystem to take control in relatively low-resolutions testing.
* Please note that unless otherwise stated, all performance graph results equate to the standard "higher is better" routine.
AMD Socket 939 Reference Systems:
Processor: Athlon 64 3800+
Memory: 2 x 512-MB OCZ PC4200 EL DIMM
Motherboards: ASUS A8V Deluxe & MSI K8T Neo2-FIR
Motherboard chip set: VIA K8T800 Pro
ATI reference drivers: Catalyst 4.7
Video Card: ATI Radeon 9800XT 256-MB
Hard-Drive: Dual Western Digital WD360 in RAID 0 Mode
IDE Interface: Serial ATA-150
CD/DVD: AOpen Combo Drive
Power Supply: ANTEC 430 Watt
DirectX 9.0b
Operating System: Windows XP Pro SP1
AMD Socket 754 Reference Systems:
Processor: Athlon 64 3400+
Memory: 2 x 512-MB OCZ PC4200 EL DIMM
Motherboards: Gigabyte K8NNXP & MSI K8T NEO-FIS2R
Motherboard chip set: VIA K8T800 Pro/nForce3 150
ATI reference drivers: Catalyst 4.7
Video Card: ATI Radeon 9800XT 256-MB
Hard-Drive: Dual Western Digital WD360 in RAID 0 Mode
IDE Interface: Serial ATA-150
CD/DVD: AOpen Combo Drive
Power Supply: ANTEC 430 Watt
DirectX 9.0b
Operating System: Windows XP Pro SP1