|
March 4, 2001
Summary
There is quite a bit of confusion about the different testing methods for heat sink fans and the validity of the different approaches. This article is aimed to shine some light on the individual parameters playing into the performance of different heatsink designs, using in part some information presented at the IDF Spring 2001 and put into context. Different methodologies are discussed and the pitfalls inherent to each method. Because of the heterogeneity of the different CPU designs, each class of CPU has its own requirement for HSF testing. As examples, the Thermaltake SuperOrb and Volcano II are used and compared with each other.
|
[an error occurred while processing this directive]
When looking at the different reviews of heatsinks currently available, one can't help but notice that there is no such thing as consistency among the different reviews. Everyone uses different testing criteria and methods, ranging from a calibrated heat source to simply running the test system in idle and load and reporting the temperature values as displayed by the mainboard's hardware monitor. To say it upfront, there is no perfect way of measuring the efficacy of any heatsink and there are no perfect guidelines either.
An eye-opener in some respect was the session about Thermal Design Guidelines for the P4 at the 2001 Spring IDF. As most of us are aware, the P4 is currently the fastest (in terms of clockspeed) and the most power demanding microprocessor on the market, further blessed with an enormous amount of heat dissipation that dwarfs any other CPU. Needless to say that also the heatsink / fan combos manufactured by a variety of companies need to adhere to very strict guidelines to comply with Intel's specs on that matter.
It is interesting, though, to look at what design factors for the optimal HSF are:
Thickness of the base plate
Material of the base plate (copper or aluminum)
Surface (height and width or HW) of the fins
Thickness of the fins (the thinner the fins are, the higher is the thermal resistance, resulting in a temperature gradient across the height of the fins)
Pitch of the fins (how many fins per distance)
Width of the gap between fins (basically pitch minus thickness)
CFM value of the fan
Pressure gradient between intake and outflow
Power consumption of the fan
rpm of the fan
Noise of the fan
All these parameters have been thoroughly analyzed and can be used to calculate the performance of an optimal HSF for a first pass design. Reality, however, shows that on average, an optimal HSF is still 50% better than the actual product. Still, there are a few rules of thumb. Copper is somewhat twice as good as aluminum. Doubling the height of the fins is equivalent to a 4-fold increase in fan power. Noise generation of a fan increases with the 5th power of the RPM. Thickness of the base plate is negligible in terms of overall performance. All of these parameters and guidelines need to take into account that, beyond a certain point, the returns are diminishing.
Even though the above explanations are quite intuitive, there are a few points that are, at least surprising. Needless to say that the average person would not think about a fan running twice as fast to be 32 times louder. Interesting also is the idea that the base thickness is of no relevance. The last point only makes sense if one knows the P4 design and what the specific Intel guidelines are. In order to move the heat away from the P4 core, Intel has added a copper slug around the silicon to increase the surface area that can be used to transfer the heat to the HSF. In addition, Intel suggests a 10 minutes interval for the integration of temperature deltas, making immediate temperature changes a factor that is not even taken into account. Furthermore, the design guidelines don't even consider anything below 7 mm base thickness.
|
