Tweaking action
To try something new, tests were run using several different sets of BIOS settings. The results should give us an idea of just how much a little tweaking can make a difference. The KT7E was run on both a 100MHz (12x100) and 133MHz (9x133) bus, with both an anemic, untweaked BIOS and a BIOS tweaked for performance.

Performance tweaks included setting CPU decode, memory interleave and timing, and so on. About the only thing I didn't tweak was the CAS latency, since this lowly Micron stick is only rated at CAS 3.

Why go through all the trouble of showing you the difference tweaks like this can make? Because not all motherboards have a flexible system like the SoftMenu. You might be harder core if you tweak with jumpers and dip switches, but Abit deserves some credit for dumbing things down and making them easy for all of us.

Memory performance
SiSoft Sandra's memory tests are a great way to measure theoretical memory bandwidth. Because of their ability to isolate the memory subsystem, we should get a good indication of just how valuable those memory tweaks can be.

In this highly focused, synthetic playground, tweaking matters, a lot. Using the exact same hardware and simply tweaking a few easy BIOS options yields huge gains, especially when moving to a 133MHz (266MHz effective) front side bus.

Like Sandra's ALU scores, FPU performance sees a healthy boost from tweaking the BIOS settings. The FPU scores seem more limited by the bus speed, but with the Athlon's easily picked multiplier lock, a front-side bus speed boost costs only a pencil and five minutes of your time.

VIA chipsets haven't had the most stellar memory performance in the past. However, with the newer chips like the KT133E and plenty of tweaking options provided by SoftMenu, memory performance is quite impressive. Sandra's STREAM tests, of course, concentrate on memory performance alone—far more important is how the board performs in real-world applications.