One of the biggest pieces of advice I have ever given to people looking for their own heatsink is to first flip the mass of metal over and look at the bottom.

A good surface finish is what we hope to see, and most of the time we'll find something a less appealing. This isn't always the be all and end of a heatsink, though it does tend to knock many coolers out of the performance circle. So why is the surface finish on the bottom of a heatsink so important?

The answer to that question lies in something called thermal contact resistance. This is described as the temperature drop across the interface between a composite system. Given a perfectly flat metal base, the degree of thermal contact resistance which exists will depend largely on surface roughness effects.

Consider what the base of the average heatsink looks like. It could be machined-flat aluminum, or perhaps the manufacturer sanded the base flat after the metal was extruded from the die. Whatever the case, most heatsinks have a base with differing degrees of surface roughness. When the heatsink is placed on top of a processor (with nothing else in between)there are contact spots interspaced with air gaps.

These gaps can be moderately small, or just about invisible, but they exist because of slight imperfections in the surface finish of the heatsink base and processor. Heat transfer from the processor is limited to conduction at the many small points where the two parts (processor and heatsink) are in physical contact, and conduction/radiation where the air gaps exist.

Contact resistance can be seen as not only the air gaps, but also the relatively small surface area of contact spots. However, the major contribution to thermal contact resistance is made by the air gaps. Now since we all want heat transfer to be as thermally efficient between the processor and the heatsink as possible we want to diminish the thermal contact resistance as much as possible.

Three possible means of reducing thermal contact resistance are; increase the joint pressure, reduce the roughness of the mating surface(s), and/or introduce an interfacial fluid of a larger thermal conductivity than air. Physics dictate that there will be some level of thermal contact resistance between the processor and heatsink materials which make direct contact, but as this area tends to be less than that of the area of small gaps so we have to focus our efforts there. Both AMD and Intel have steadily increased the contact force that a heatsink applies to the core of a processor, and manufacturers have improved the surface finish qualities of their heatsinks so we'll look towards thermal joint compounds.

As long as that thermal joint compound is of a lower thermal resistance than air (which they are) we are able to reduce that aspect of the total thermal contact resistance. However, thermal joint compound has a much lower thermal conductivity than a direct metal to metal interface so it is really important to use it sparingly. The material is only being used to fill in those very small air gaps. That's it.

The variety of thermal joint materials available today is pretty stunning, with everything from foils like Indium, lead, tin, and silver to silicon-based thermal grease. Newer silver, copper and now ceramic based thermal compounds are also widely in use but they can have drawbacks that silicon-based materials don't have. For starters, metal-based thermal compounds can conduct electricity, so they present an electrical shorting hazard to sensitive components like your processor. Ceramic and other non-silicon based thermal compounds can be mildly abrasive and if improperly removed, can lightly scratch the exposed silicon core of some processors.

Generally speaking, thermal compounds are more thermally conductive than thermal pads which may consist of non-flowing material, or flowing material. Bergquist for example makes a wide variety of these (non-flowing) thermal pads and they are generally unacceptable for current AMD and Intel processors. Thermal pads which flow (at a specific temperature) on the other hand can be very efficient, and easy to apply.

Typically, these materials are applied as a small grey, or pink patch to the base of the heatsink and can have the consistency of a wax, or bubble gum. After the heatsink has been applied and pressure and heat are both enacting on the material it will be come semi-fluid - thinning out and leaving those air gaps filled with a thermally conductive material. It is important to realize that until the material reaches its flow temperature the heatsink will not be at its most efficient. Once the material has "set" however the installation process is completed and things are good to go.

With that we have invariably reduced the amount of thermal contact resistance by filling in many of the small air gaps. Radiant thermal conductivity, or conduction by air are not very efficient ways for heat to travel, so it is always important to minimize the overall handicap each imparts on processor cooling.