Metals are quite good heat conductors. But some are substantially better conductors than others. There can be conflicts between material cost, manufacturing concerns (e.g optimum alloys for extrusion vs casting vs bonding) and optimum heat conduction. Getting heat to distant parts of a heat sink is not usually a problem unless fins are very large and very thin. The obvious answer to transferring more heat to air is to increase the surface area of the heat sink by adding more fins. However, additional fins tend to block airflow thus trapping heated air between the fins and reducing the transfer of heat. Some heatsinks claim to be designed to induce "chimney effects" where air warmed in the valleys between the fins rises and is replaced by cooler air from outside the sink. Heatsink design-- especially airflow analysis -- seems to be very complex. Large differences can exist between observed performance of heatsinks that appear similar in size, shape, and material.
A heatsink that transfers heat efficiently may not be easy to manufacture even if it works. Some designers bypass the whole issue of proper design of a passive heatsink by mounting a fan on the heatsink. One problem with fans is that their failure is likely to be accompanied by severe overheating of the device they are supposed to be protecting. Fans are also noisy. Elaborate measures may be added to detect a fan that has stopped turning.
In older PCs, there were a number of devices where heat was a problem, especially the CPU, the power supply, the video circuitry and disk drives. The trend over time has been for devices other than the CPU and video card to generate less heat and for CPUs to generate more. Thus heatsinking has become almost entirely a CPU/video issue. In the case of CPUs introduced after the year 2000 -- which approach or even exceed 100 watt dissipation -- heatsinks threaten to become so massive that there is concern about damage to the CPU sockets from the weight of the heatsinks.
Shrouds are increasingly being used on heat sinks using fans both in order to control airflow and to reduce fan noise. Shrouds are generally designed on the assumption that the fan will blow air up into the shroud rather than sucking air past it into the fan.
Heat pumps using the Pelitier Effect to "pump" heat electronically through a stack of specially prepared plates have found some use. A heatsink is required at the warm end. Unlike passive or fan equipped heatsinks, a heat pump can become cold enough that water may condense out of humid air at the cooled end. Moisture and electronics do not mix. Similarly, water or other cooling fluids can be used in a sealed cooling system, but the working fluid usually must be pumped, and leaks are a constant worry. Few PC owners are anxious to keep a plumber on retainer. It is reportedly possible to operate a PC immersed in mineral oil, but problems such as oil working into peripherals via the connecting cables discourage this in practice.
Pragmatically, heat removal in PCs is by means of convection. In theory, heat can be lost by radiation, but at PC temperatures that is not very practical. Conduction to the case is sometimes used in laptops/notebooks some of which become uncomfortably warm and a few of which are reported to be capable of injuring a user who operates one in their lap. Case conduction is rarely used in desktop PCs or servers. See the articles below for an extensive analysis of heat flow. The same material is covered in a somewhat different context in articles available at most libraries and online discussing housing insulation except that the goal of the latter more often is to keep heat in rather than to get it out.
Return To Index Copyright 1994-2002 by Donald Kenney.