Test protocolSo as to collect our data in a realistic environment, all meaurements were taken in a closed casing. We opted for the Sonata III from Antec, very popular and considered a reference in this area. We also used an Intel based platform with a Rampage Extreme from Asus based on the X48 chipset. We went for a Q6600 CPU (revision G0).
Noise pollutionTo measure noise levels, we used a sonometer placed more or less 60 cm from the casing, like this:
We measured noise levels in idle and noise levels after 30 minutes in load. Ambient noise levels were measured at 34.8 dB, a fairly low figure.
Temperature was measured in two ways: using various thermal sensors and infrared thermography. Infrared thermography gives you an overall visual look at temperature thanks to a 320x240 frame of infrared emissions taken with the Ti25 from Fluke.
To collect all the data, we first left the system in idle for 45 minutes. Then we wrote down the scores from the various sensors with dedicated software and we opened the casing and rapidly took a thermal image.
The system was closed directly afterwards and we moved on to measurements in load. We carried out the 3Dmark 2006s Pixel Shader and Prime to put the CPU in constant load as we compared graphic cards with very different performance levels. CPU usage of course increases when going from 50 to 500 fps. After 30 minutes we briefly stopped the 3Dmark test to collect the data which wasn’t logged and then started the test again immediately. 15 minutes later we opened the case so as to take a thermal image. Then we wrote down maximum temperatures recorded before opening the case.
The following measurements were taken:
The fixed motherboard sensors: MB (motherboard, placement not detailed), SB (southbridge), NB (northbridge) and PWR (power supply)
A motherboard custom sensor: we placed it on the northbridge heatsink
The integrated CPU sensor: we took the temperature of the hottest core
The integrated GPU sensor
The integrated hard drive sensor
An external sensor placed behind the casing, 1 cm away, where the flow of air comes out of the graphics card
Between each card tested, we let the temperature of the motherboard drop back down to room temperature, which was noted each time and controlled using the air conditioning if necessary. We allowed a variation of plus or minus 1°C as we are not able to control the temperature of the room any better than this.
Notes on infrared thermography
It is important to note that images obtained may be affected by components with lower heat radiation than average as they may seem cooler than they actually are. Thermal emissivity (heat radiation) is the percentage of energy emitted from an object or body that has the same temperature. Our measurements are based on a 95% emissivity, which corresponds to the general rule.
This is not however the case for some components such as shiny metal surfaces. We approached this problem on the motherboard by putting neutral adhesive tape on parts such as USB connnections, network connections and so on. We didn’t however do this for the GeForce logo on the graphics cards, which appears to be colder than the rest of the card, not because of any paranormal qualities but simply because the heat radiation of the logo is lower.