LiamC said:
What people tend to forget is that while copper can absorb more heat than aluminium by volume, heat is carried away form the heat source more efficiently by aluminium than copper. The purpose of a heatsink is twofold, absorb the heat and dissipate it. Good thermal conductivity (aluminium) is essential. Copper has the marketing and mindshare - that doesn't make it intrinsically better.
The thermal conductivity of copper is higher, so it also dissipates heat better than aluminum,
all other things beings equal. That is wherein the difference lies. Since copper is much heavier than aluminum, copper heatsinks are generally smaller than aluminum heatsinks for the same processor in order to keep the weight within certain limits. Although the superior thermal conductivity of the copper can partially compensate for this, the overall smaller size necessary for the copper heatsink gives rise to the myth that aluminum dissipates heat better than copper. It doesn't. I made liquid heat sinks for thermoelectric modules of the same size and design from both materials, and the copper ones were almost 50% better. Copper absorbs heat better than aluminum because its density is higher, hence more mass to heat. Heat absorption only matters in heat sink design in case the fan fails. A higher heat absorption can give you a few minutes more until the CPU dies from overheating, but other than that, there is no benefit.
The end result of all this is that aluminum heatsinks are indeed better than copper ones unless the copper ones can be made the same size(and number of fins). In that case, the copper heat sink will be better by 25% to 50%, but the heavier weight may exceed the limits that the heat sink attachment mechanism can handle. So we arrive at the same conclusion for different reasons.
Having read many heat sink reviews, I find most of the methodology in this area flawed. The only way to properly test a heat sink is to apply a fixed heat load to the base, and measure the temperature rise above ambient of the heat sink base directly under the heat source. Divide the temperature rise by the input heat load, and you get the heat sink's thermal resistance(the lower the better), which is the only true measure of it's effectiveness. For example, I recently tested a P4 heat sink which I was thinking of using to cool thermoelectric modules. At 50 watts heat load the temperature rise was 11.75° C
above ambient temperature. Therefore, the thermal resistance was 0.235°C/watt. The heat sink was 83 mm x 70 mm with 27 fins of 30 mm height. By comparison, my copper liquid heat sink was 40 mm x 50 mm x 12.7 mm high, and had a thermal resistance of 0.02°C/W at 20 gph water flow.
One thing that should improve the efficiency of any existing heat sink by about 15% is to use a tip-magnetic fan. This will improve airflow to the center of the heat sink, where the heat load is, due to the smaller fan hub. Another way is to increase the number and/or height of the fins. However, past a certain point, the high density fins limit the airflow and efficiency begins to drop, so there is an inherent limit to how efficient a heat sink can be in a given size. Smoothness of the base plate is very important, as is contact pressure. The smaller the gap filled with relatively low conductivity thermal grease, the cooler the microprocessor. I typically make sure any heat sink I use for TE modules is smooth to within 0.0005 " in order to minimize the thermal losses at the interface.
