Benchmark results and analysis

Test Constants

All tests were run with BIOS settings at defaults, other than the AGP aperture size being set to 256mb, and memory timings set to "aggressive".

The testbed system components were:

All tests were run under Windows XP, standard installation without any service packs. The chipset drivers were the manufacturer supplied software, as they came with their respective boards.

The benchmark software used was:

Missing results

Results are missing for the Abit board as it wouldn't run DDR400 with 2 Dimm modules on"aggressive" timings. 1.75ghz / 166FSB results for the Biostar are missing due to lack of a cooler that would fit the board and still cope with the heat output.

Memory Bandwidth

ScienceMark 2.0 was used for these tests. To determine what benefits dual DDR brings over a standard configuration, I ran benchmarks with 2 modules, in slots DIMM1 and DIMM2, which share the same memory bank, and do not support dual DDR when they are both used. All figures are in MB/s

Dual DDR is failing to give any more memory bandwidth at all, on any board at any speed. This would seem to be because of the AMD architecture, because although there is 150% of the normal bandwidth between the RAM and the northbridge, communication between the CPU and the northbridge is already running at the same throughput as normal DDR ram, and is not speeded up by using dual DDR, and so becomes a performance bottleneck. This is the same reason that asynchronous memory speeds have no benefit on AMD systems. The pentium 4 architecture doesn't have this bottle neck, as the pipeline from the CPU to the north bridge is "quad pumped" which gives twice as much data throughput MHz for MHz than DDR allows.

When reading the ScienceMark results, bear in mind there is a margin of error up to 100MB/s either way. Even taking this into account, the dual DDR is still showing no benefits, and there are few differences between the motherboards, with the albatron and the chaintech showing results consistently a little higher than the rest. This is a point in favour of the albatron, as you wouldn't expect a budget board to perform better than its more expensive peers, evenif only by a small margin.