Power And Noise Benchmarks
The power consumption benchmarks were very close, as we actually lowered CPU voltage to get our higher overclock. As well, we weren’t really able to overclock the graphics cards to a noteworthy level without running into heat issues.
The only curious result was the peak concurrent CPU and GPU peak load test, which showed our overclocked system used less power than the stock system did. We repeated these tests a number of times with the same results and paid close attention to the CPU and GPU clock speeds to make sure that everything was running at full speed. Still, we kept getting identical results.
This is a benchmark we thought was important to include, to demonstrate the trade-off we had to make for system stability. At stock graphics card fan speeds, the system would sometimes crash during 1920x1200 benchmarks, especially if AA was applied. To counter this, we increased the graphics card fan speed to 100% using the Nvidia System Tool, which gave us the stability we needed, but at a high cost to the amount of system noise. The end result was that our quiet Micro-ATX system became quite loud.
In retrospect, if we were to live with this system on a daily basis, we might choose to make a compromise: increasing the fan speed by a small amount, and even decreasing the graphics cards clock speeds by a small amount as well, perhaps to reference GeForce GTX 260 levels. As the benchmarks show, a slight increase or decrease in graphics card clock speeds has a negligible effect on graphics performance, but it can have a marked effect on stability. Slowing it down a bit while speeding up the fans a small amount would likely supply us with solid stability and have a negligible impact on performance, supplying a quieter and more tolerable long-term solution.