How are CPU coolers made

Processor cooling

Until Intel's 486 processor, cooling a processor was never an issue. The processors got lukewarm at most. The surrounding air inside the computer case was completely sufficient for cooling. Since the processors exceeded the clock frequency of 66 MHz, at least passive cooling by a heat sink has been necessary. But at some point that was no longer enough. At significantly higher clock frequencies, so much heat is generated in the processor that it can only be operated with an active cooler. An active cooler is a heat sink on which a fan also ensures an air flow.

Today, modern processors are not only developed for computing power, but also for energy efficiency. It is usually less about saving energy than about getting heat problems under control. While a few years ago it was not possible to do without active cooling, there are more and more processors that are powerful enough for a specific application or environment and can manage with pure passive cooling or even without it.

Reasons for the heat problems

In addition, a brief explanation for non-electronics technicians in advance of why heat develops in a processor and thus a heat problem occurs: The processor is an electronic semiconductor component that is operated with voltage and current. Every electronic component or device consumes electricity or electrical power. Electric power is a product of voltage and current. If voltage or current increases, the power increases many times over. Power is often converted into heat. This is known from a light bulb that gets hot after a long period of use. Heat is a problem for electronics. Especially when there is too much of it. Electronic components, especially processors, do not tolerate very much of it. Many components are densely packed into a processor. If there is too much heat in the processor, it will destroy the processor.

Processors consist of a large number of transistors (electronic semiconductor components). If you look at a single transistor, it works like a switch. It is either on or off. This represents the digital bit with the values ​​"0" and "1". In both conditions, the power is zero watts. The transistor does not give off heat when it is switched on or off. This effect is one of the most important characteristics for the success of digital technology.

A processor now consists of many millions of transistors. This begs the legitimate question, why does a processor get so hot anyway? There are three reasons for this:

  1. Frequent redistribution of power during switching due to a high clock rate.
  2. Increase in leakage current due to structure reduction.
  3. Parts of a processor have nothing to do but use electricity.

On the one hand, the transistor acts like a capacitor when switching from 0 to 1 and from 1 to 0. This means that the transistor is "charged" or "discharged" when switching. It takes a while. During this time, neither the current nor the voltage is zero. And that is why electrical power is generated in the switchover period, which is converted directly into heat. The more frequently it is switched, the more power and the more heat is generated. Switching in the megahertz range is still almost problem-free. But even in the gigahertz range, which is normal with today's processors, power consumption increases sharply. Depending on the processing load, a different number of millions of transistors switch. Only a very small current flows through a single transistor. But all together come to peak values ​​of over 100 amperes.
The result: The heat becomes so great that considerable measures have to be taken to cool a processor. One problem is that with twice as much power (in watts) the cooling effort does not have to be twice as large, but several times as large.

The second reason that leads to the generation of heat is the increasing number of transistors and the consequent reduction in size of the semiconductor structures. The small structures mean that the insulation layers in the transistor are very thin. It can happen that a leakage current flows. The leakage current increases with every further reduction in the size of the structure. You cannot limit it by lowering the clock rate. It always flows. It can only be limited by taking special measures in the manufacture and use of other insulation materials. It is there in any case. Sometimes more sometimes less. But, the leakage current increases the performance and thus heat generation.

The third reason arises from the increasing variety of functional units that work in parallel or independently of one another in a processor. All functional units in the processor switch to the rhythm of the processor cycle. The parts of the processor that actually have nothing to do also consume energy. This can be remedied by energy-saving functions that can switch off parts of a processor.

Processor cooling basics

When it comes to cooling processors or an entire computer system, it is usually not about cooling measures, but about cooling measures. This means that cold is not supplied, but heat or heat is dissipated. During this process, an air flow is generated which, due to its principle, draws in colder air. In order for the airflow to work, the computer case must of course have openings through which colder air can be drawn in and warmer air can be drawn out. One problem with such openings is, of course, that dust and dirt can get into the housing, which in the long term can also damage the electronics.

The heat is usually dissipated with air or water in combination with a heat sink that is permanently connected to the heat source (e.g. processor). Unfortunately, the heat sink only has a limited thermal conductivity and low heat distribution. Therefore, heat is dissipated by air flow or water flow. A higher air flow allows more heat to be dissipated per volume. A higher air flow is generated by fans or ventilators rotating faster, but this also creates a higher level of noise.

Quieter cooling is generally possible with liquids. The liquid cooler is attached directly to the component. It is a cooling element through which liquid flows, which causes neither noise nor vibrations. This type of cooling is more complex and is used when air cooling is no longer sufficient.

Cooling measures

There are several things you can do to keep a processor cool. One measure, purely passive, is used in almost all semiconductors. For example with a transistor or a voltage regulator, which have to dissipate a high power loss and whose housing is not sufficient for this. A heat sink is used to add a heat-dissipating component to the housing.
With active cooling, a fan is attached in addition to the heat sink, which ensures that the heated heat sink cools down faster. Active cooling has the disadvantage that it requires additional space and electricity.

passiveHeat sink with cooling fins and painted black
activeas above, but with an additional fan installed

Sometimes it happens that a processor slows down its performance despite sufficient cooling. This could be due to the fact that the voltage converters on the motherboard, which are located directly next to the CPU socket, benefit little from the cooling measures. The voltage converters get hot under full load and the processor shuts down its performance.

The optimal cooling

  1. One or more thermal sensors determine the actual temperature.
  2. The fan control compares it with the fixed setpoint temperature.
  3. The fan control calculates a new value for the command value.
  4. The fan motor changes its speed and thus also changes its cooling capacity and temperature.
  5. Then the control loop starts all over again.

Thermodynamics: Fundamentals of Thermal Conduction

The transport of thermal energy (heat or cold) only takes place when there is a temperature difference between the elements or materials. The greater the temperature difference, the more powerful the heat energy flows.
The thermal conductivity of an element is given in watts per meter and per Kelvin. The following table shows the cooling capacity of the elements used for cooling. Basically, the colder a material feels at room temperature, the better it dissipates heat.

elementHeat dissipation in watts per meter and per Kelvincomparable
air0.003 W / mKHeat insulator
water0.6 W / mKHeat insulator
steel20 to 30 W / mKHeat conductor
silicon160 W / mKHeat conductor
Aluminum (pure)221 W / mKHeat conductor
copper393 W / mKHeat conductor
silver410 W / mKHeat conductor
gold310 W / mKHeat conductor

Processor cooler

With current processors, the CPU die sits on a carrier board. This board is called a chip carrier. All of the waste heat from the processor is generated on the tiny die surface. Because the heat is so concentrated in a tiny area, it is a difficult task for CPU developers to build future processors in such a way that they can be safely cooled.
A heat distribution plate, called a head spreader, is located above the die. The heat spreader distributes the waste heat and prevents dangerous hotspots on the die, which can lead to crashes and in extreme cases to the destruction of the processor. In addition, the heat spreader prevents damage to the sensitive silicon die when installing the cooler. The heat sink of the cooler sits on the head spreader. In between there is a thermal pad or thermal paste.

The heat dissipation is slowed down by the contact resistance between the die and the head spreader and between the head spreader and the cooler base plate. So that the thermal energy of the processor can be dissipated over the relatively small area of ​​the die or heat spreader, the material of the heat sink must be a good heat conductor. Usually the heat sink is made of aluminum or copper. The heat sink ensures that the waste heat from the processor is dissipated to the air in the PC case. In order for this to work, according to the law of thermodynamics, the processor chip must be warmer than the metal of the cooler. And that, in turn, has to be warmer than the ambient air. This means that there has to be a regular exchange of air inside the housing, otherwise the heat dissipation from the processor to the ambient air is not strong enough. Unfortunately, air is a poor conductor of heat. The transfer of thermal energy from the smooth metallic surface of the cooler to the air molecules takes place only very slowly. Therefore, large amounts of air and metal surfaces are necessary.

However, spatial limits stand in the way of cooler growth. For this reason, the thermal energy exchange is accelerated by a fan mounted on the heat sink. The heat sink, together with the fan, is referred to as a cooler. Although the air is only partially suitable for cooling a processor, the fan is used for secondary processor cooling. This only works if the air is exchanged as often as possible (through movement). A simple fan is sufficient here, which ensures that there is always enough cool air flowing between and around the heat sink, sometimes not exhausting. The best results are achieved by setting the fan speed in direct relation to the processor temperature. To do this, the control electronics must have access to a precise temperature signal. These values ​​usually come from the thermal diode in the processor. In addition, computer cases are equipped with fans that transport warm air out of the case.

Temperature distribution in the heat sink

Temperature distribution in the case of a poorly thermally conductive base plate Temperature distribution with a base plate that conducts heat well Temperature distribution with a heat pipe

Commercially available heat sinks are made of aluminum with ribs that increase the radiating surface. Together with a fan, the principle of processor cooling is achieved. In order to improve the efficiency of the cooler, a mixed copper-aluminum design was developed, which consists of a copper base plate for quick heat distribution and light aluminum fins. In order to increase the effectiveness even more, heat pipes are used to conduct part of the heat from the base plate into the colder fins.

Examples: processor cooler

Thermal pad and thermal paste

In the early days of processor cooling, the processor cooler was mounted directly on the die (chip). The unevenness, although not visible, between the touching surfaces caused poor heat distribution. For this reason, a thermal paste or thermal pad was applied between the die and the heat sink. If the installation was careless, the die was damaged and the processor was destroyed. To avoid this, the processor manufacturers at some point installed a permanently installed heat spreader (IHS) over the die. Here, too, it is necessary to apply a heat-conducting paste or a heat-conducting pad. It should be noted that thermal pads can only be used once. If the cooler is detached from the processor, a new thermal pad must be used.
There are different thermal pastes with silicone or silver components. You should keep your hands off thermal pastes with aluminum, as they can conduct electricity.
You should be economical when applying the thermal paste. The paste may only be applied thinly. It should only compensate for the invisible bumps.

Tips when choosing a cooler

  • Many fans are only approved up to a certain clock rate and TDP (Thermal Design Power) (recommended by the manufacturer).
  • A thermal paste is usually better than a thermal pad.
  • Particularly powerful fans are also particularly loud.
  • Papal fans are considered to be particularly quiet. However, the heat dissipation is less.

Tips for cooler installation

  1. The heat sink should be installed with the motherboard removed.
  2. The upper side of the processor and the lower side of the heat sink must be carefully cleaned with a soft cloth.
  3. The irregular surface of the die and the underside of the heat sink must be compensated for with thermal paste or a thermal pad. Both are applied to the die or the head spreader.
  4. The amount of thermal paste should be as small as possible, since the thermal conductivity value is only between 1 and 8 W / mK anyway. If the processor has a heat spreader, a pea-sized blob of thermal paste is sufficient. If you put the cooler directly on the die, then it should only be a fraction of it. If the paste is applied too thickly, it will cause bubbles to form. And air is known to be a poor conductor of heat.
  5. The contact pressure of the heat sink is sufficient to ensure that the thermal paste is distributed without bubbles. The additional distribution of the conductive paste is not necessary.
  6. Without a heat spreader, self-builders have to put the cooler on with great care. The corners of the processor surface are quickly damaged and a total failure of the CPU is therefore inevitable.
  7. The cooler has to be hooked in on one side and then on the other. The retaining clips must then be locked in place. When placing the heat sink on the processor surface, it is essential to ensure that the heat sink is not tilted when the retaining clips are inserted. A good hold must be guaranteed so that the heat sink does not slip during operation or transport and destroy the processor or the motherboard.

Water cooling

Water cooling is a recurring attempt to effectively cool a high-performance computer. Unfortunately, water is just as unsuitable for cooling as air. Both water and air are heat insulators. They are very poor conductors of temperature. The cooling is only created by the constant movement of water and air. With water cooling, hoses form a closed water circuit in the computer system. Special heat sinks are connected to the hoses. A pump, which is often attached outside of the computer housing because of its size, keeps the water in constant motion, thus keeping the heat sinks cool.
Water cooling has another horse's foot. By moving the air around the processor, voltage converters and memory modules also create cooler air. With water cooling, there is no air movement, which means that the processor will eventually throttle the clock rate if the voltage converters get too hot. So what use is a cooled processor when the voltage converters are smoking.

Water cooling is an elaborate measure to cool a computer. Installation is not that easy and there is always the risk of water leaking and damaging the computer. Usually distilled water is used, which cannot conduct electricity. But once the water has escaped and mixes with small amounts of dust or dirt, that also becomes conductive.
For a long time, water-cooled systems were a nice gimmick for PC hobbyists. However, there were only very rarely ready-made systems on the market. No dealer wants to guarantee such a system. That is why systems for water cooling are usually only available for retrofitting. The hobbyist is responsible for the retrofitting and the resulting damage.

Other related topics:

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage devices, drives and important hardware components.

I want that!

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage devices, drives and important hardware components.

I want that!