Gaming results
We deliberately use graphics subsystems that allow for multi-GPU operation on a particular platform. For the ASUS (Intel P965), Foxconn (Intel i975X) and MSI (ATI RD580, AM2) we've used ATI's X1900 XTX and CrossFire cards. For the ECS nForce 680i SLI we've used a pair of GeForce 7900 GTXs. The choice of graphics cards may not allow for complete comparisons between motherboards, but, we feel, it's more a real-world indication of the kinds of GPUs that would be bought by readers who opt for a particular chipset, thereby keeping an upgrade path for multi-GPU operation open in the future.
If we look at the Far Cry and Quake 4 results for the three ATI-based graphics subsystems we see that the COMMANDO has the measure of the Foxconn 1975X and, thanks to the superior gaming performance of the Intel Core 2 Extreme X6800, beats up on the AMD Athlon 64 FX-62.
Splinter Cell: Chaos Theory performance is broadly similar across all platforms, irrespective of CPU and graphics card.
Adding in a second card to form multi-GPU usage and raising the resolution to 1920x1200 with 4xAA 8xAF, we can identify two factors at play. The first is that CrossFire setups beat out the SLI (but that would change if we installed a couple of GeForce 8800 GTXs, though). The second, and what's pleasing, is that the x4 bandwidth of the second mechanical x16 PCIe slot doesn't hinder performance at all. Indeed, the COMMANDO'S multi-GPU performance is the second fastest here, just beaten out by the Foxconn board.
Performance summary
The ASUS COMMANDO matches the Foxconn 975X7AB in practically every benchmark, including, somewhat surprisingly, our high-resolution, multi-GPU Splinter Cell: Chaos Theory benchmark. The COMMANDO hasn't put a foot wrong thus far.Overclocking
It would be criminal to conclude on this review without a word or two on overclocking.We've heard fantastic reports of the board hitting 575MHz+ FSB (2300MHz QDR) during internal testing at ASUS. Our sample, though, wouldn't go much above 390MHz without losing stability, as per our overclocking methodology. We tried tweaking various voltages lines, swapping out the memory for some higher-performing modules and the usual tricks in obtaining high FSBs, but to no real avail.
Changing the B2-stepping Core 2 Extreme X6800 for a B0 model and loosening general timings improved FSB performance markedly, and we were able to hit 490MHz FSB with reasonable stability, so the final FSB overclock is dependent upon the components as much as the inherent ability of the board. That is, in part, why we're generally apprehensive of publishing overclocking results because, as the maxim states, your mileage will vary.