Setup and Performance
Please Give Us Some Feedback
| Carbon Black |
Black is Beautiful
|(Review by MS, October 7, 2003)|
|Cooling and Environmental Factors
All runs were performed in triplicate and in reverse order within approximately 2 hours to minimize the impact of climatic or other environmental changes.
The system was booted into Windows XP Professional Corporate Edition and left idle for a few minutes until the processor temperature (measured on the on-die diode) stabilized. After that, 3DMark2001SE was launched as application, which resulted in an immediate temperature spike with soon thereafter return to idle values. As soon as the idle temperature was reached, the 3DMark2001SE "Benchmark" was launched at the default settings of 1280x1024, no AntiAliasing.
Temperatures were logged using the log function of MBM5 with a 5 sec polling interval and imported into an Excel spreadsheet. Since MBM5 only reports full centigrades, all values are rounded. However, in repeated runs, the differences in temperatures at any data point were no more than +1 centigrades. Averaging the individual runs under the same conditions would not contribute more to any clarity, therefore, we show one representative log graph for each run. All runs are triggered to the same starting point.
We used three different compounds:
The different compounds and how they compare during a single run of 3DMark2001SE
Overall, the results are extremely similar, however, Carbon Black still resulted in the highest temperatures, followed by Arctic Silver (original). AS is highly viscous and diluting it in a 1:2 fashion with Carbon Black improved the performance of both compounds. Arctic Silver "Ceramique" gave the overall best results.
On the time-Axis, the individual events within the Benchmark are clearly recognizable as temperature transients with loading phases apparent as drop in temperature that are consistent between all runs of all compounds tested here.
As additional control we ran the same series of test runs on the next day after swapping the mainboard for an ASUS A7N8X (Dual Channel nForce2), which gave us different readings on an absolute scale, however, on a relative scale, the results were comparable. (Baseline started at 39 centigrades and under load the temperatures went to 49 centigrades, however, relative humidity had dropped to 36% and ambient temperature was 19 centigrades, in addition, the calibration of the output on both boards will be different).
<= back next page =>