All clocks are at manufacturer stock speeds unless specifically noted as part of an overclocking test.
Along with the hardware, some amount of watercooling gear is chosen. This can vary by the testing setup depending on what is being tested at the time. In the case of this test we are using:
GPU Block (As applicable, see below) – Danger Den 8800GTX Waterblock
Radiator – BlackIce GTX 240mm
Fans – two 120mm AC Ryan Blackfire 4 Kameleon
Pump – Laing Vario D5 12v @ max setting
Tubing – 1/2" ID Clearflex
D-Tek FuZion CPU Block v.2.0, Intel LGA-775 on universal bracket
Danger Den TDX CPU waterblock for Intel LGA775 Processors
Understanding the test methods is vital to understanding the results of a test, as well as allowing maximum repeatability. Here at bit-tech we strive to make our tests fair, unbiased, and repeatable by anyone with the time or inclination to do so. The following methods and assumptions were used in this test:
The test setup is built in a climate controlled room free of unnecessary clutter. Once assembled, the system was started on air cooling and the operating system (Windows Vista SP1) was patched fully. Speedfan 4.34 was used to measure the temperature of the CPU cores before, during and after testing as prescribed below. Ambient temperatures were also taken from a digital thermometer on the wall approximately one metre from the test setup.
For this test, Ambient room temperature was 22 degrees Celsius.
The system was booted with the test CPU block and no GPU block, and left to idle at the desktop (no screen saver) for one hour to allow the fluid temperatures to come to equilibrium, and temperature was then recorded.
Prime95 was then started on each core using Small FFT to assure maximum CPU usage, which was also monitored via a sideboard widget to assure proper load. Temperatures were recorded at 10, 20, 30, and 60 minutes and averaged. No cores showed any significant change in temperature (greater than one degree) after 20 minutes. Once 60 minutes passed, Prime95 was stopped on all cores and the system was allowed to return to idle. Temperatures were recorded at 10, 20 and 30 minutes though no readings showed deviations past 10 minutes.
The system was then shut down and drained. The graphics card was switched out for an exact copy of the OCZ 8800GTX with stock heatsink removed and Danger Den 8800GTX waterblock installed. The GPU was run before the CPU in the loop. Testing was repeated as above.
Once a test was completed both with and without a GPU block, the CPU block was switched and the process was run again. Upon completion of this, the test was restarted from scratch. Results were averaged amongst the two runs to develop the reported findings on the next page.