Since the explosive increase of electronic devices a very annoying problem occurred: how to cool down the silicon. I want to make from this article a practical one, that's why I don't want to enter into complicate physics. The electrons mobility is bigger in lower temperatures and this allows greater clock speeds and voltages applied to the silicone core. You know how important it is to keep the CPU cool these days; in the early times of computing it was unnecessary because the CPU didn't get much hotter than a low power transistor. With the clock increase the CPU needed more and more power to function so passive heatsinks became required. Later a fan was added to cool even more the heatsink surface. There are several external things which make computing unreliable, two of the most important being temperature and dust. The dust level is very hard to control, there are no simple devices able to keep the dust away. A sponge limits somehow the dust but it's not regarded as a serious solution in most media.

The good news is that we are able to control the temperature level, at least in a certain degree. Today I'll try to present you several traditional coolers which should be able to cool down your CPU no matter if it's a cheap Duron or a more expensive Pentium III. Let's go to work.
We defined several days ago the PC Hardware general testing methodology. Although we haven't defined a special section for coolers the testing procedure will be the same with a few amendments. As you may know we split the market in three segments: low price, medium price and high end. We have many coolers ready for testing today in all the three segments. From the beginning I was tempted to test the coolers in every market segment and to publish the results separately. The other option would have been to publish the results for every manufacturer that entered the roundup. In the end I decided to go for both. First I tested the coolers in every market category and in the end I make a comparison based on price and performance.
We have described almost all in the testing methodology but we haven't defined precise testing procedures so I will do it now. We have tested coolers that can work with both Intel and AMD CPUs but I preferred to run all the tests with AMD CPUs because these are more sensitive to temperature issues. I personally prefer real CPUs, no resistor simulators because these fail to simulate properly the small CPU core of the today CPUs.
In our testing procedure the case and the temperature in the case is not important because in Romania it's very hot these days (around 38-40C). The temperature in our building is around 23C which is positive because if would have been above 30C our tests hadn't had any relevance. Physics laws say that if two bodies exchange temperature the balance temperature is greater than the coldest and smaller than the warmer. With a medium temperature of about 23C we had the chance to measure the coolers efficiency.

Methodology

Below you can find the three market as defined by PC Hardware at this date:

• Low market segment - coolers targeted to Athlon below 1.1Ghz. These coolers should be able to work with any Duron / Celeron and Pentium III below 1.13Ghz. Not suitable for overclocking, but you can push to the limits a 800Mhz Duron.
• Medium market segment - coolers targeted to Athlon above 1.1Ghz and below 1.4Ghz. Any Duron / Celeron and Pentium III should work with these coolers. May be suitable for overclocking.
• High end market segment - coolers targeted to Athlon beyond 1.4Ghz. These coolers are the best overclockers on the market and can work with any CPU that fits its design.

Every cooler was installed with the interface material provided. If the cooler didn't have an interface material build in we used thermal grease. The only cooler in this situation was the Thermalright SK6. Many people like to read the temperature with infrared rays or thermistor digital devices. We decided to use the software method although the temperature reading may not be the most correct one from two reasons. In order to precise test the temperature you have to drill a hole into the heatsink base and to introduce a probe as close to the CPU core as possible. It was not possible to do this because heatsinks are different and the method wouldn't have had relevance. The second reason is that using the software method the reading is impartial: all coolers are tested on the same platform and comparisons can be made. We stressed the system as much as were able to do with SiSoft Sandra 2001 Professional and Quake III and the maximum temperature on the CPU sensor was recorded.