Graphics Overclocking: Getting The Most From Your GPU

Benchmarks: ATI And D.O.T.

As we’ve already pointed out, overclocking using MSI’s D.O.T. only works with the outdated 8.542 driver, so that’s what we’re using here. In this test, we were interested in seeing what kind of benefits driver-level overclocking offer. To this end, we used the default setting as well as the Sergeant and Commander levels of the D.O.T. software. To see what kind of an impact CPU overclocking would have, we also ran all tests with our Core i7-920 once at 2.67 GHz and once at 3.8 GHz.

One thing to bear in mind when looking at the minimum and maximum frame rates is that they represent the absolute peaks and troughs and may last only as long as a single frame. Additional power provided by overclocking the CPU doesn’t seem to have much of an impact at all. Indeed, in Far Cry 2 you can easily compensate a slower CPU by overclocking the graphics card instead.

The “Commander” setting represents the highest O/C-level offered by the D.O.T. system and offers performance increases between 1.1 (average) and 2.4 (minimum) fps. With the CPU overclocked to 3.8 GHz, the maximum saw the greatest gains of up to 6 fps. Ideally, overclocking would allow the graphics card to break the 25 fps barrier at a setting like 1920x1200 with 8xAA and 16xAF in order to avoid choppiness and lag. Sadly, the minor bump from 19.0 to 20.1 fps won’t make any kind of difference.

Left 4 Dead doesn’t benefit much from an overclocked CPU either, while an overclocked graphics card can boost performance by up to 5.1 fps. The “Sergeant” setting proved completely stable. When we switched to “Commander” mode and enabled AA as well, the additional strain on the card caused the benchmark to crash from time to time. At that setting, the GDDR5 memory is running 100 MHz over ATI’s reference clock speed, which translates into a jump from 3,600 MHz to 4,000 MT/s.

MSI’s fan speed profile reacts to overclocking and audibly spins up the fan when higher, more demanding D.O.T. settings are selected. Nevertheless, the performance gains are rather small. To corroborate this, we ran another quick benchmark at 1280 x 1024 and witnessed similarly small performance gains there as well. The low overall results may stem from the high quality settings we used for this benchmark. Reducing the image quality to attain higher frame rates doesn’t make much sense, though, since these results already represent playable frame rates.