Testbed and Methods
The tests were performed on a closed and fully assembled case and at a constant ambient temperature maintained by an air conditioner. I also took care to lay out the cables and wires in such a way that they didn’t hinder free circulation of air. The fan speeds were set as to achieve the lowest noise level using an external speed controller. Unfortunately, the fans in the system cases didn’t have velocity sensors so I couldn’t see their resulting speeds.
The following configuration was assembled in the tested PC cases:
- ASUS P5WDG2-WS mainboard
- Intel Pentium Extreme Edition 955 (3.46GHz; Presler core, Hyper-Threading off)
- Thermaltake Big Typhoon cooler (1200rpm)
- 512MB DDR2 SDRAM
- MSI RX1650XT 256MB graphics card
- Seagate Barracuda 7200.8 200GB hard disk drive
- Seasonic SS-401HT power supply
This is far ahead of any regular user configuration as well as of a lot of overclocked ones in terms of CPU heat dissipation. This should suffice for testing CPU coolers as well as system cases.
There were four test modes:
- CPU Burn: two copies of the CPU Burn utility running
- VGA Burn: Unreal Tournament 2004 is running with the maximum graphics quality settings
- HDD Burn: two file sets (a 7GB folder with many small files and a 40GB folder with movies) are being copied simultaneously from one partition to another
The temperatures of the CPU and mainboard were read with ASUS PC Probe which was supplied with the mainboard. The temperatures of the GPU, VRM transistors, and the chipset were read with Scythe KamaMeter. The HDD temperature was reported by HDD Thermometer. The temperatures were read only after they had fully stabilized. The ambient temperature remained constant at 25°C throughout the tests.
The noisiness of the preinstalled system fans is discussed below. The noises of the PSU, CPU and graphics card didn’t interfere as I disabled their coolers when listening to the system fans.
You can learn more about the Scythe KamaMeter here. I’ll only tell you where I put the speed sensors. First, two of them were put on the South Bridge and on the PCI-X bus controller:
But the temperature of the PCI-X bus controller is hardly important for an ordinary user, so one sensor moved to the graphics card:
I couldn’t put the thermocouple right on the graphics core, so I measured the temperature of the spot where the heat pipes had contact with the base. This is sufficient for test purposes, but I plan to modify the graphics card’s heatsink to place the thermal sensor in an even better position.
The temperature of the North Bridge…
…and of the transistors of the CPU power circuit was measured in the same way:
I had problems with the thermal sensors that wouldn’t hold on the scorching-hot elements. So, I had to fasten the sensor with thermal glue and fix it in place with a tiny piece of polypropylene.