To test the modules, I first set them to stock speeds and let MemTest do it’s thing overnight. I woke up to see zero errors, which allowed me to further tweak and test. It took me an afternoon to finish testing, and here are my results:
I quickly found out that these modules are not for overclocking a great deal. The modules use BH-5 chips, a Winbond UTT variant. BH-5 never usually allows much headroom for overclocking, at least to my liking. It has very little trouble handling voltage you throw at it, and that’s why they chose to use them for this kit, I suppose.
Even though it pains me to not be able to take the modules to at least DDR500 speeds, I realize that these are not really meant for overclocking. They are developed for gamers who want superb memory speeds and super tight timings. Loosening the timings is not expected with BH-5, but the goal is to keep the timings as low as possible, and in this case, I’m extremely satisfied with the 2-2-2 timings.
For benchmarking, I used EVEREST Ultimate Edition and SANDRA 2005. To test out the effects of overclocking in games, I used Half-Life 2, Doom III and Far Cry. The games were all run in 640*480, except for Far Cry which was 800*600. Anti-Aliasing and Antisotropic Filtering are disabled.