As transistor sizes continue to decrease, chip manufacturers pack even more on to the processor. In 1992, a 486/DX2 66 MHz CPU consumed about 7 W of power with 1.2 million transistors. It didn't even require a cooling fan. In 2003, the Itanium II debuted at 1 GHz, using up to 130 W with 220 million transistors. By 2005, it is reported that Intel will release processors in the 6 GHz range. The final method for cooling these chips currently is unknown.
Your microprocessor-based manufacturing equipment must be kept cool if it is to work continuously and efficiently. According to one cooling system manufacturer, liquid could be the next evolutionary phase in computer cooling systems. Some microprocessor manufacturers have been devising water-based concepts for years, anticipating the point when air cooling simply would reach its limit.
Why Use Liquid? Heat-producing devices in a typical computer enclosure are air cooled. Generally, this involves mounting a heat sink and fan to each component. Heat generated from the CPU is transferred into a metal heat sink, where a fan blows air across its wider surface area. While altering a heat sink's size and makeup can improve the effectiveness, it is still limited because air absorbs and transfers heat very slowly. To help counteract this, the fan can be run at a higher speed, but most people know what that means: High performance has become equated with high noise. As systems continued to be upgraded, required heat sinks became larger and louder.
Of liquids, water (after mercury) conducts heat the fastest. Its thermal conductivity is about 30 times greater than that of air. It also holds more heat; it takes more than four times as much heat to raise the temperature of water as it does air.
These basic physical attributes give liquid cooling considerable advantages, but not only in temperature.
Cooling 101: Moving HeatAs the First Law of Thermodynamics implies, matter and energy cannot be created or destroyed, only converted between the two. Likewise, heat -- the movement of energy from a hotter object to a cooler object -- never is eliminated but only moved elsewhere. This is the role of all cooling systems.
To accomplish this, there are three primary modes of heat transfer. Some forms of heat transfer can be duplicated using multiple methods (both natural and forced), but every cooling system uses these same basic processes.
- Conduction, the transfer of heat through matter with no net displacement of the matter.
- Convection, the circulatory motion of a gas or liquid caused by the variation of its density and the action of gravity.
- Radiation, the process of transferring heat by emitting electromagnetic energy in the form of waves or particles.
No matter how flawless a surface may appear, it is highly irregular on a microscopic level. A CPU and cold plate or heat sink may only touch at 0.01 percent of their total surface areas. Because the rest of the surfaces are separated by air, many high-heat sources would fail without a thermal interface material. Not surprisingly, the best thermal interface compounds contain elements with high thermal conductivity. These are subsequently more expensive, and so most heat sinks use cheaper graphite (gray/black pads) or silicon/zinc oxide (white paste) compounds.
“Thermal conductivity” is the measure of heat flowing through a length, which is not to be confused with “thermal conductance,” or the measure of heat though a surface. Thermal conductivity is the amount of heat a particular substance can carry through it in unit time. Usually expressed in W/moK, the units represent how many watts of heat can be conducted through a one-meter thickness of said material with a 1oK temperature difference between the two ends. Table 1 shows examples of thermal conductivity.
Specific heat capacity is the amount of heat a particular substance can hold. Typically expressed in kJ/kgoK, the rate depicts how many kilojoules of energy are required to change the temperature of one kilogram of said substance by 1oK. Table 2 show examples of specific heat capacity.
Copper is the preferred cold plate material for cooling systems. It is extremely close to silver in performance but only one-sixth the cost. Like most metals, however, copper does not hold heat for long, so the heat must be absorbed by something else.
The specific heat capacities show water to be the best liquid for holding heat. Practically, it also is the best for transferring it.
This would indicate the ideal configuration is to use copper to transfer heat from the processor, then use water to absorb and move away the heat. Although many other factors are involved, this is the basic foundation of a liquid cooling system.
Liquid Cooling DesignAt first glance, liquid cooling looks simple. Liquid is pumped through a cooler, it absorbs heat, and it is cooled back down with a radiator. But, because you are working with liquid, the design principles become more complicated.
Water Conducts Electricity. Obviously, you do not want the system to leak. This seems to be a primary concern of those new to liquid cooling, but in reality, leaks are not a significant risk. Short of deliberate damage or perhaps a shipping catastrophe, a professional system does not leak.
Hose clamps are used internally as a precaution in some systems. Each device is designed to fit snugly with the next, and although not recommended, the system could run correctly without being clamped.
Pump Reliability. The pump is a liquid cooling system's “heart.” In conformance with that analogy, a similar effect can happen to the system if this component failed. Submersible centrifugal pumps are suitable choices for this type of application. Using dual pumps will add to the system's stability, but the primary reason for dual pumps is to increase liquid pressure.
Corrosion. Many types of chemical reactions can be present in a cooling system that uses liquid. The most common is galvanic corrosion, which is caused by different metals in an electrolyte (in this case, water). These varying electrode potentials can create a battery effect, damaging the anode metal.
The worst situation in which corrosion might occur is by using regular tap water in a liquid cooling system. Tap water contains numerous trace elements that can accelerate galvanic corrosion. For this reason, distilled water should be used.
Liquid Pressure and Flow Rate. It is a common misconception of liquid cooling that higher flow rates will equal better temperatures. The liquid in every cooling system constantly loses pressure as it travels; friction in the tubing and other obstructions causes it to lose energy. The more coolers that are in the system and the longer the hose, the greater the liquid pressure drop. In a traditional single-pump cooling system, liquid pressure can decline sharply upon adding multiple cooling devices to the system.
Dual-pump systems are designed to cool simultaneously many hardware devices such as processors, hard drives, video cards and motherboard chipsets. Although up to six coolers are recommended, this number can reach as high as 10, depending on the chosen coolers. The key is liquid pressure.
So why would the highest flow rate not always be favorable? Many cooling kits include a single pump with flow rates of 10 l/min or more. Figure 1 indicates that the higher the flow rate, the lower the liquid pressure will be. For a relatively low pressure system, it is analogous to a garden hose: You can turn the water all the way on (high flow), but it surges farther if you hold your thumb over it (high pressure).
The other issue relating to flow rate is the heat exchanger. Every liquid heat exchanger has an optimal flow rate; that is, a margin -- often quite small for advanced components -- in which it provides its highest efficiency. Without knowing the margin, pairing an effective pump with a heat exchanger can be a difficult task. A flow rate that is too low or too high can waste efficiency.
In cooling processors, this optimal flow rate for dual-pump systems generally is between 1.1 and 2.3 l/min. The temperature gain beyond this range is not significant despite increasing the flow rate.
Hose Diameter. Hose diameter is another consideration in a liquid conductive cooling device. Most systems using a single, large pump require large diameter tubing, typically 0.375 to 0.5" (10 to 13 mm), to maintain this essential flow rate. Increasing the length of tubing further reduces flow rate. Exceeding the critical point where increasing the tubing diameter no longer matches the temperature drop reduces cooler effectiveness.
CPU CoolerAlthough a CPU cooler is a critical component in every system, it is not solely responsible for performance. Efficiency is related to the entire collection of components in a liquid cooling system, so altering something as minor as airflow over the heat exchanger can affect how well a CPU liquid cooler operates.
Copper is the most practical metal to use in a liquid cooler cold plate. It provides a high degree of thermal conductivity and is abundant at reasonable cost. Unlike a fan and heat sink, a liquid cooler does not need to be made completely of metal. Only the areas between the heat source (such as the CPU) and liquid pass significant amounts of heat. A liquid block made completely of metal may be excessive and will be heavier and more expensive.
The internal design of a CPU cooler is everything. Poorly designed liquid coolers will require a higher flow rate to maintain system efficiency. Like air-cooled heat sinks, they also tend to be larger and heavier, which can place physical stress on the processor and motherboard.