System Builder Marathon, December 2010: $500 PC

Overclocking

When the time came to tinker outside of spec, our first order of business was to determine if the dormant processing core could be successfully unlocked. We had a stable fourth core in September, but this time our system refused to even boot as an Athlon II X4. The motherboard’s design, unfortunately, made resetting the system a bit tedious, as accessing the clear CMOS jumper requires removal of the graphics card.

While unlocking was a failure, our Athlon II had a good deal of headroom for increasing core speed. The chip’s VID (Voltage ID) was 1.3 V, as opposed to the September machine's 1.4 V. Once again, the motherboard overvolted a bit beyond that. At stock voltage, stability was found beyond 3.4 GHz.

We only took the CPU up to 1.4 V in the BIOS, which CPU-Z reported as 1.432 V at idle and 1.464 V during load, finally settling for a respectable 3.875 GHz (250 * 15.5). This Mushkin RAM was found to have little headroom, so the 250 MHz reference clock worked out for keeping our RAM within spec (at DDR3-1333). Maximum available CPU-NB frequency was 2500 MHz, but we settled for 2250 MHz due to the motherboard's cooling and lack of adjustable CPU-NB voltage.

The Sparkle GeForce GTX 460 also had a decent amount of headroom for increased frequencies, and the quiet cooler was effective for overclocking without overriding the auto fan settings. Starting at a reference 675 MHz core and 1350 MHz shaders, our card topped out at 835/1670 MHz, respectively. We didn’t push the GDDR5 memory all the way to its breaking point, calling it quits after raising it from 900 MHz (3600 MT/s effective) up to 1060 MHz (4240 MT/s). These frequencies were then dialed back to 823/1640/1050 (4200) during our overclocked set of benchmarks.