The processor in the computer occupies a dominant role, since its tasks include managing the system. When buying and operating, it is important to consider all parameters, including the heat dissipation of the processor. This article will be about this.

The principle of the processor

There is a certain sequence of actions according to which the processor performs its functions. The block that controls the processor loads data from RAM, which will later be used. After that, the instructions are placed in

When data is loaded into a memory buffer, it is written into instructions to be moved to a register later. When the logical device reads the received registers, it thereby executes the instructions written in them.

After executing the instructions, the results need to be stored somewhere, and the processor stores them in its cache for later use. If the information is needed again, then the processor will receive it twice as fast, since it is already stored in the buffer memory.

If the information stored in the buffer is no longer required, then the processor moves it to main memory, since it can store more information, and the cache frees up space for new data.

Such a parameter as processor heat dissipation is one of the main ones, since this calculation helps to find out which cooler is best to install.

The main characteristic in choosing a processor is the core. That is, the more of them, the more threads, and this, in turn, increases the performance of the central processor.

Just as important as the heat dissipation of the processor. The fact is that this parameter becomes smaller every year and it is measured in nanometers. There are millions of transistors on the manufacturing process, which are responsible for the transmission of direct and alternating voltage, converting it into a binary number system.

The frequency of the processor indicates its speed of information transfer and data processing. That is, the higher the frequency, the more productive the processor. On some of them you can see an additional letter, which makes it clear that the device is designed for overclocking. In everyday use, it is of no use, but in games, processor performance plays an important role, because you have to read a huge amount of data. It is worth overclocking the processor only if it is really necessary and the cooling system allows it. Similar procedures are carried out through the BIOS menu.

TDP, or heat dissipation, shows how much heat is generated when the processor is running. Based on these data, the user selects a cooler that will be able to cool it and the power supply. The maximum heat dissipation of the processor is specified in the device parameters.

The processor cache can be multi-level, and each level has a unique amount of memory and data transfer rate. The higher the level of the memory buffer, the slower it works, but it has more memory. So, the first level cache communicates with the processor faster because it is closer, but the memory bank is very small. But the third level can have 8 megabytes of memory, but be farther from the processor chip, so it has to access it through the first two caches.

Of course, when purchasing a processor, it is important to consider its generation and series. These parameters are in the first place, because they determine which socket must be selected for the processor, as well as the motherboard.

These were all important parameters. And the maximum heat dissipation of the processor allows you to find out about its power consumption.

Data calculation

Many users after the overclocking procedure are interested in how to correctly calculate the heat dissipation of the processor. The example uses a processor that has a TDP of 65W.

Almost all processors after overclocking increase their frequency by about thirty percent. Thus, you need to take the original number 65 and multiply by 30. After that, divide the resulting number by 100. It turned out to be 19.5. You can round up to twenty. Now it turns out that the heat dissipation of the overclocked processor is 85 watts.

You can also increase the frequency by a greater number of percent, only in the calculations you need to substitute the figure by which the initial heat dissipation of the processor is multiplied.

From the previous paragraph, we figured out how to find out the heat dissipation of the processor. We continue to discuss this topic further.

CPU temperature monitoring programs cannot show how much CPU heat dissipation actually is. For example, "Aida" or BIOS can only show the temperature that is at the moment, but during active operation, the processor heats up a lot, and this is difficult to track. Therefore, the most reliable option to find out the maximum heat dissipation of the processor is to make calculations.

When purchasing a processor from this manufacturer, it is important to consider the generation and series. For Intel processors, heat dissipation depends on what temperature the device can withstand. With these data, you can select the appropriate cooling system.

Processors from this manufacturer, which are designed for mobile computers, have a heat dissipation of 35 watts. It is not difficult to cool such chips if you use the simplest cooler with copper pipes and a heatsink. The frequency of such processors is from two to three Hz.

The heat dissipation of processors for desktop computers of this line reaches 50 watts. Here you will need more serious cooling, but during operation the temperature does not rise to high levels. Thus, the cooling system can be represented as an aluminum radiator without tubes, with a fan.

The reduced process technology consumes less power, but other innovations include an increase in the number of cores to four and an increase in operating frequencies. These settings increase the power consumption of the CPU.

The heat dissipation of the i5 processor does not even reach 75 W, but manufacturers recommend installing a 95 W cooler. It's all about overclocking potential, that is, the standard frequency is about 3.5 Hz, and you can overclock the processor up to 4.2 Hz. With increasing frequencies, the heat dissipation of the processor also increases.

The heat dissipation of the i7 processor is practically the same as the previous generation, that is, 65-75 watts. For such CPUs, you will need better cooling in the form of a tower cooler. Since this generation has a different connector for connecting, the cooler must be selected according to size.

An interesting comparison can be made with the two models 7700 and 7700K. The letter at the end means that the processor is designed for overclocking, but the performance gain is only ten percent. Most importantly, with such a minimal increase, the heat dissipation of the processor increased by 30 watts.

Subsequent models have heat dissipation reaching the bar of 200 W, despite the fact that their frequency is 5 Hz. That's what it is - the heat dissipation of Intel processors.

The heat dissipation of AMD's entry-level FX processors reaches 95 W at an operating frequency of 3.5 Hz. The uniqueness of the processors of this manufacturer is that the price-quality ratio differs from competitive models. So, the processor of the latest model can cost 25-30 percent less than the competitor's counterparts.

In this case, AMD uses its unique socket AM3 +. Returning to heat dissipation, we can distinguish models with a TDP of 125 W, and their frequency barely reaches 4 Hz. CPUs that have surpassed the 220W bar are capable of operating at a starting frequency of 4.7Hz. However, it is not known how the cooler will cool the CPU if you try to overclock it. Of course, you can refer to the calculations presented above, but the overclocking percentage of processors from this manufacturer is unpredictable.

Choosing and installing a cooler for Intel i5 and i7

So, the choice of a cooler depends on some parameters.

1. CPU heat dissipation.
2. Energy consumption.
3. Acceleration ability.
4. Connector (socket).

If the heat dissipation of the initial state of the processor allows the use of a boxed cooler, then you should not worry about cooling. When the TDP crosses the 95 W mark, then you will need to purchase a tower cooler with copper pipes, a heatsink and a fan.

The modern market has to offer various variations such coolers, but their installation on the processor remains unchanged. To begin with, it is worth removing the cover of the system unit and disconnecting all components from the motherboard in order to bring it to a horizontal position. Then a reinforcing bar is attached to the back panel so that the cooler does not deform the motherboard.

Apply a drop of thermal paste to the processor and spread it evenly over the surface. Now the heatsink is placed opposite the screw holes that come with the cooler. After that, the fan is attached, and the motherboard, together with the cooler, can be placed back into the case of the system unit.

Cooling for AMD FX

The parameters for purchasing a cooler are the same as in the previous example, but the installation is slightly different.

Having placed the motherboard in a horizontal position, it is necessary to apply thermal paste to the processor and smear it, after which a radiator with tubes and a fan is mounted. The fact is that coolers for processors from this manufacturer are installed in a completely different way. There is a lever in the lower left corner that folds out so that the cooling can be installed. Further, the lever is lowered, making a characteristic click, which means that the cooler is fixed. It is important to know that distortions and unevenness are unacceptable, as this can lead to the fact that the generated heat will not be properly removed by the tubes - and the processor will simply burn out.

The principle of operation of the cooler is simple - the processor heats up during operation, the thermal paste ensures full contact between the processor and the cooler, the tubes are engaged in heat removal to the radiator, and the fan cools it, expelling hot air. Coolers with a radiator without tubes or without a fan have a low heat sink, and with all the attributes, the processor is not in danger of overheating.

All coolers are purchased in accordance with the socket, as they differ not only in the location of the "legs", but also in size. If the diameter of the fans is incorrectly selected, then the processor will not be able to cool as it should, and in another case, the cooler simply will not fit.

Coolers with a cable for three and four pins are slightly different from each other. Three-pin cooling works at full capacity even when the processor is not under active load. Another thing - four contacts. An additional pin allows the cooler to control the fan speed in accordance with the load on the processor. Thus, during passive operation, the processor practically does not heat up - the fans spin weakly, but in games or programs where calculations are actively carried out, the fans operate at full power.

If you decide to assemble a new computer yourself, then you have to solve a number of problems associated with the selection of components. One of these problems is the choice of a cooler for the processor. In this article we will consider the main points in this not tricky case.

Standard CPU cooler

When building a new PC, it's always tempting to opt for the stock cooler that comes with the processor. After all, when using it, you can save a little.

If you are assembling a computer to work with undemanding programs, then a standard cooler will be enough. Moreover, if good air movement is organized in the case, then a standard cooler will cope even with heavy programs and computer games.

The only thing for which a standard cooler is definitely not suitable is overclocking the processor. If you plan to overclock the processor, then you should definitely take a more efficient cooling system.

Also, don't go for a stock cooler (or what it's called a "boxed cooler") if you want to build a quiet computer. Standard coolers are usually very small in diameter and because of this they are noticeably noisy. At the same time, the revolutions of such a cooler should be high, because the radiator under them is also small.

CPU cooler socket and dimensions

When choosing a CPU cooler, the first thing to consider is the size of the cooler.

If you choose a cooler that supports your processor socket, then you simply cannot install it, the cooler will not fit into the right place. If you make a mistake with the dimensions of the cooler, then you may have problems when closing the case. If the cooler turns out to be larger than the space that is between the processor and the case cover, then you simply cannot install the side cover.

In the case of particularly large coolers and compact motherboards, situations are possible when the cooler will overlap slots for RAM or even PCI Express connectors. This should also be taken into account when choosing a CPU cooler.

Not a standard cooler in a computer case

Therefore, in order not to have to take the cooler back to the store later, it is better to carefully study the characteristics of the cooler and make sure that your processor socket is in the list of supported sockets, and the dimensions of the cooler will not create problems when assembling the system.

Cooler Efficiency

When choosing a cooler for a processor, it is very important to correctly evaluate its efficiency and heat dissipation of the processor. To do this, it is best to focus on the TDP of the processor and cooler. TDP stands for "Thermal design power", which in turn can be translated as "Requirements for the cooling system." TDP is specified in Watts and indicates the amount of heat that the processor cooling system must remove. You can read more about this.

If the characteristics of the cooler indicate the TDP for which it is designed, then simply compare the TDP of the cooler with the TDP of the processor. If the TDP of the cooler is higher, then you can safely buy such a cooler. It will cope with the cooling of your processor without any problems.

But, the characteristics of coolers do not always contain information about TDP. In this case, when choosing a cooler for a processor, you have to evaluate its efficiency based on indirect factors. These factors are:

• Radiator weight. The heavier the heatsink, the more heat it can take away from the processor and dissipate into its surroundings. Therefore, the greater the weight of the radiator, the more efficient the cooling system.
• Number of heat pipes. Heatpipes transfer heat from the processor to the heatsink fins. Therefore, the more heat pipes and the larger their diameter, the more efficient the radiator.
• Number of coolers and their size. The more coolers on the radiator and the larger the diameter of these coolers, the better the radiator is blown and the more efficient the cooling system.
• Heat pipe contact. Heat pipes can contact the processor either directly or through an additional plate. The best option is direct contact. This allows the heatpipes to better transfer heat from the processor to the heatsink fins.

Processor heat dissipation is one of the main parameters that you should pay attention to when assembling a computer. The CPU is a key component on which the operation of the entire system depends. If it overheats, forced cooling mode will begin, which results in skipping cycles, that is, problems with computer performance. When the processor cannot cool down even in this way, it starts to automatically turn off so as not to fail completely. It is probably not worth talking about the dangers of a sudden shutdown of the computer, moreover, when it happens in the emergency mode of the central processor.

In order for the CPU to maintain an acceptable temperature, it needs additional cooling. That is why it is important to choose the right cooler for the processor. There are many nuances that you need to pay attention to when choosing a CPU fan, and it is also important not to forget about the main parameters when choosing it.

Why change the cooler that comes with the kit

On sale you can find central processors in OEM and BOX configurations. In terms of performance, there are no differences between these versions of the CPU of the same model, and they differ only in the bundle. The OEM version is only the CPU itself, while the BOX package includes a cooler.

Many users who do not have much experience in building a computer may get the impression that the ideal solution is to buy a BOX processor package, but this is not always the case. The coolers that come with the processor are most often of mediocre quality, and they are not able to provide cooling for the “stone” under its high load. That is, if the CPU is purchased for an office computer, where it will not face tasks more complicated than working with a browser and a text editor, then there will be no problems with cooling the processor with a cooler from the BOX bundle. But if the "stone" is planned to be used in games and other resource-intensive applications, then you need to take care of buying a more powerful cooler.

Modern processors are made up of more than 500 million transistors, each of which heats up during operation. Due to the small area of ​​the CPU, such a serious heat cannot dissipate on its own, and an additional cooler is required to remove it. The more complex tasks are assigned to the processor, the more efficient cooling is needed.

How to choose the right CPU cooler

The most important thing when choosing a cooler is to choose it according to the characteristics of the processor. Obviously, the more powerful the processor, the more heat it generates under high load. Accordingly, he needs more cooling. The heat dissipation parameter of a processor is commonly referred to as TDP, and it is measured in watts. Paying attention to the heat dissipation of the processor, we must not forget that the models also differ from each other by the type of socket. Now let's look at the selection for each of the parameters in a little more detail.

Processor socket

A socket is the size of a processor and is denoted by: AM3+, 1150, 2011-3, and other combinations of letters and numbers. Manufacturers try to standardize CPUs to certain sizes, but due to changes in manufacturing technology over time, there are already about a dozen of them available. The socket is the size of the connector on the motherboard where the “stone” itself is inserted.

Thus, when choosing a cooler for a processor, you should initially find out the processor model and check on the Internet on the manufacturer's website in what size it is made. Often coolers are suitable for several sockets due to the versatility of mountings.

CPU heat dissipation

Having decided on the size, you need to look at the heat dissipation of the processor. You can find out information about the TDP parameter of a particular processor on the official website of the manufacturer.

With the selection of a cooler according to the heat dissipation parameter for a certain processor model, everything is somewhat more complicated. The fact is that in online stores and on various sites it is quite rare to find accurate information about which processors with which TDP a particular cooler is suitable for. However, well-known manufacturers of processor fans, such as Noctua, do not hesitate to provide such information.

If you could not find information about a specific cooler model, you can use the data from the table below. Please note that the information in it is very approximate, and it is better to choose the fan option for the processor "with a margin".

How to choose a quality cooler

Having selected cooler models by processor parameters, there will still be dozens or even hundreds of fan options that can be purchased. In such a situation, you should look at the reviews about the quality of a particular cooler left by their owners. But it’s better to independently reject the least quality options from the available options by evaluating the fans according to the following parameters.

cooler base

The area that the cooler touches the processor plays an important role in cooling. Since the dimensions of the “stone” are fixed, it is practically impossible to increase this contact area. At the same time, some cooler manufacturers, in search of innovation, partially remove heat pipes on the basis of the fan. Because of this, the contact area and the efficiency of the cooler are reduced.

It is also important that the base be made without various patterns. It should be copper polished to a mirror finish. Before buying, it is necessary to inspect the cooler so that there are no cuts, bumps and other defects on its base.

Please note: In most cases, the base of the cooler is made of copper. This material is cost-effective and efficient in terms of heat transfer. There are aluminum options, but they are much less effective. In this case, copper can be coated with nickel, due to which it acquires a silver color.

heat pipes

On almost every modern cooler, you can see several heat pipes, while previously they were not used. The fact is that with an increase in the power of processors and an increase in their heat release, the existing cooling standards were no longer enough, and manufacturers decided to use a proven option - the installation of heat pipes.

The copper tube is filled with liquid and sealed on both sides. When heated, the liquid heats up and passes into a gaseous state. The gas moves to the other side of the tube and thereby removes heat. Then the steam cools, turns back into water and returns to the base of the tube. In computer coolers, the process is similar, except that there is also a porous material inside, which is necessary for the liquid to return back even when the tubes are in a horizontal position.

When choosing a cooler for the processor, you need to pay attention to how many tubes are installed. In their properties, they do not differ much, depending on the size, so the main criterion is the quantity. The minimum allowable amount for cooling a modern powerful processor is 3-4 tubes, but the more the better.

Radiator

From the base of the cooler, heat passes to the radiator, which consists of dozens of plates put on heat pipes.

The radiator can be of any shape, but it is important to remember some rules that distinguish a good option from bad:

• The larger the radiator area, the better;
• The plates should be thin, but in large numbers;
• It is better when the radiator is made of copper.

Since the radiator on the cooler is partially visible in an open computer, some companies try to make it the best way from a design point of view. It can be of various colors, shapes, plates are made at unusual angles. If the above rules are followed, then the design decisions have little effect on the quality of the cooler.

Fan

Some time ago, in pursuit of quiet computer operation, manufacturers did everything to abandon the active cooling element, that is, the cooler itself. However, a heatsink without an element to dissipate heat cannot cope with powerful processors, and fans in the processor cooler have not yet been abandoned.

When choosing a fan, it is necessary to pay attention to its size, respectively, and to the size of the blades. Coolers with large blades are not only more efficient in terms of heat dissipation from the radiator, but also quieter. There is an erroneous opinion that the cooler rotation speed is much more important, but this is not so. Speed ​​is an important parameter, but the diameter of the fan itself is more important. If you install a cooler with a small diameter of the blades, but a high rotation speed, such a PC will make a lot of noise even when performing “office tasks”.

Also, when choosing a cooler, you need to pay attention to the type of bearing on which the fan is mounted. On sale you can find options made on a rolling bearing (Ball Bearing) or sliding (Slide Bearing). Rolling bearings, which are less noisy and have a longer service life, perform better.

How to install a cooler on the processor

Having chosen the right cooler, there should be no problems with its installation. With the help of special mounts, the fan is installed on top of the processor inserted into the slots of the socket. Most often it comes with a cooler detailed instructions on its installation, which allows you to understand in general terms the principle of operation of the fastening mechanism.

Before installing the cooler is important. It is necessary to eliminate unevenness between the base of the cooler and the processor, for maximum heat transfer. Without thermal paste, the processor will not be able to work for a long time, therefore it is forbidden to neglect this step.

Quite often, technical periodicals mention such characteristics of processors as TDP, crystal temperature, maximum power dissipation, etc. However, the general public is not sufficiently informed about what each term means and how to interpret it, reviews sometimes appear not quite correct interpretations of those or other results and, accordingly, erroneous conclusions. The article discusses the issues of heat dissipation using the example of Intel processors, as well as some features of the next generation CPUs.

As you know, every entity has two extremes. With regard to microprocessors, these are performance and power consumption, and the first parameter is better known to us, since it is given the most attention in the press, and the average PC user is much less aware of the second. This knowledge is divided into two parts - empirical and theoretical, while the latter most often come down to familiarity with the mysterious abbreviation TDP (Thermal Design Point or Thermal Design Power) and the corresponding unit of measure - watt. The term TDP does not have a well-established Russian equivalent, it can be translated as "thermal design power" of the processor. The concept of TDP is most often used to characterize the thermal (thermal) performance of a microprocessor (its "hotness": the lower the better), and other things being equal, a processor with a low TDP is preferred. In addition, this indicator serves another purpose - to intimidate the consumer. Like, this processor dissipates "a lot of watts", so its use in home or office conditions is impossible.

As will be seen later, everything is determined not by the magnitude of this power, but by how efficiently we can dissipate it. The PC user receives an empirical assessment "by ear" - the computer makes noise (which is most often associated with the processor cooling system), or visually - through the BIOS or using software supplied by the motherboard manufacturer. Unfortunately, reviewers usually do not pay due attention to these characteristics, namely: not just the mention of temperature values ​​in certain places of the board, but their correct interpretation. For example, if a PC user observes a processor temperature of 100 ° C in the utility readings, you should not despair - in fact, it is much lower. At such a high temperature, the processor simply could not function, because in case of overheating, which is this value, the CPU will simply stop. And this means that such a temperature cannot be reached even theoretically.

Actually, the main purpose of the proposed material is to explain what is hidden under the mentioned characteristics and how they should be correctly understood and used. All further considerations refer exclusively to Intel microprocessors.

Introduction to process physics

First of all, let us recall some principles of power supply of microprocessors and the basics of thermodynamics in order to give an idea of ​​the range of tasks solved by the manufacturer.

The Intel microprocessor is powered by a VRD (Voltage Regulator Down) source, commonly known as a voltage converter. It converts the voltage of 12 V into the required voltage for the processor - about 1.5 V or less (Vcc - Voltage CPU Core, processor core voltage). In this case, the supply voltage on the 12 V bus with a current of 16 A (192 W), as indicated on the power supply, is converted into a supply voltage of 1.5 V, but with a current of 100 A (these figures are given solely to simplify mathematical calculations). In such a situation, of course, there is a loss of part of the power (in our case, for example, 42 W), since the converter has an efficiency of less than 100%. The final current of 100 A is supplied to the processor through several hundred pins - in the technical documentation, you can be surprised to find that most of the pins of the LGA775 socket are used to power the processor and ground.

The value of this part of the power is quite high. A 3 GHz CPU dissipates less than a 3.4 GHz CPU, but they both fall under a TDP of 95W! We will talk about the TDP parameter itself a little lower, the main thing for now is to understand that the maximum power dissipated by the processor is not the same as the TDP parameter.

The power leaving the processor is converted into heat, which must move elsewhere to equalize the heat balance. If the possibility of removing this heat from the processor was not provided, then the temperature of the CPU would rise rapidly and it would fail. Therefore, the heat generated by the processor (its crystal) must be taken away from the microcircuit and spent on an absolutely useless thing - heating the air in the room. For this, the Fan Heatsink Solution, or active cooling system, was invented. The modern design is shown in the figure (the fan is not shown there). The heat generated by the processor crystal (dark green in the figure) is removed from it in the following order: first it passes through the heat-conducting material of the microcircuit, then it enters the metal cover of the distributor (the main purpose of which is not mechanical protection of the crystal, as many believe, but uniform distribution of heat dissipated by the microprocessor crystal). After that, it moves to the so-called heat-conducting material, which is applied to the sole of the radiator and has different crystalline phases depending on the temperature (therefore, never try to remove the heat sink from the processor without first turning on the PC for 10-15 minutes, otherwise you can simply pull the processor out of the socket , especially when using Socket 478). Further, the heat enters the radiator and, with the help of a fan, goes outside the structure.

Let us recall once again that the main task of this design is to remove heat from the microprocessor and dissipate it in the surrounding space. Certain difficulties await us on this path, and the main one is related to ensuring the thermal efficiency of the device. It is a “layer cake”, each layer of which can both help and harm. Any material has its own characteristic of thermal resistance or, in Intel terminology, thermal efficiency (parameter Ψ in the documentation for the processor). This means that it will heat up, and as a result, heat can return to the processor die. Thermal resistance is measured in °C/W (less is better) and shows that when a thermal power of 1 W passes through a material, the temperature of the material will rise by this amount. For example, when one watt of thermal power passes through the radiator material with the parameter Ψ = 0.3 °C / W, its temperature will increase by 0.3 °C, at 100 W of thermal power, the heating will already be 30 °C. Adding to this value an ambient temperature of 40 ° C, without much effort we get as much as 70 ° C! And this means that sooner or later the processor will also heat up, which is exactly what we want to avoid, or at least minimize.

The author tried to evaluate the quality of thermal pastes common in the domestic market - it does not stand up to criticism. In all cases, their use resulted in the processor heatsink fan speed being 200-300 RPM faster than Intel's thermal interface material. The reason for this is the high value of thermal resistance. Of course, Intel does not release such material for its "boxed" products on its own, but when choosing a supplier, a thorough analysis is carried out in terms of price / performance. Materials with the most good performance are expensive, the same pattern is characteristic of radiators. You can make it all copper and with a huge scattering surface area, but it will come out heavy, bulky and expensive. You can use an additional fan, the air flow from which will "blow off" the heat from the surface of the radiator - cheap, but noisy. There are other exotic ways - for example, water cooling, cryogenic installations. They are more efficient, but are unlikely to get into mass production because of the high price and low reliability.

Therefore, Intel uses a number of technical solutions that ultimately give the best balance. Finding the best cooling solution is always a compromise between cost, efficiency and reliability. The total thermal heat dissipation index is the sum of the thermal resistances of each of the elements of our “pie” that are encountered along the path of thermal power. And each element can significantly affect the final integral characteristic of the thermal efficiency of heat removal.

Learn more about TDP

TDP is a value that is used to calculate the thermal efficiency of a cooling system. The widely held belief that TDP determines the maximum power dissipation of an Intel processor is fundamentally wrong.

How is TDP used? The input data for calculating the thermal efficiency of the cooling system (and eventually developing its design) are the TDP value and the maximum operating temperature of the crystal T case max . It is measured at point T case (see fig.) - geometric center on the surface of the cover of the heat distributor (note: T case is not the temperature of the crystal, as it is mistakenly believed). As an example, consider the TDP value of 95 W, which is currently used to calculate cooling systems for approximately 90% of Intel desktop processors. Tcasemax for them is approximately 70 °C (the exact value can be found in the SSpec database at support.intel.com using the SL code present on the chip label and processor carton). The formula for calculating thermal efficiency (thermal resistance) will look like this:

T case max = T ambient + TDP × Ψ,

where T ambient is the temperature of the "ambient",

Ψ = (T case max - T ambient) / TDP = (70 - 38) / 95 = 0.34 C / W.

As a result, we must design a cooling system with such thermal efficiency. And here begins the struggle between "good" (thermal efficiency) and "evil" (economical).

Imagine that we have developed such a system, now it needs to be tested. To do this, you will have to damage the surface of the cover of the heat distributor. A groove is made in it, in which one thermocouple is laid. Another is placed on the surface of the fan motor (in Fig. T ambient). With the first thermocouple we measure the temperature of the crystal, and with the second - the environment. We begin to gradually load the processor and see how our cooling system works. Upon reaching the threshold of 95 W, the temperature at the measuring point should not exceed 70 °C. The indicated power can be dissipated by only a few models out of 90% that fit “under the umbrella” of 95 W, the rest will never reach this value. For example, in the line of Intel Pentium 6×1 processors, all models dissipate up to 86 W, i.e., hypothetically, it can be assumed that this barrier will be overcome only starting from a core frequency of 3.8-4 GHz.

So, if during our measurements the temperature at this point exceeds T case max = 70 °C, something is wrong here. For example, we applied cheap thermal grease to the sole of the radiator. The question arises, how much can an Intel processor dissipate at a TDP of 95 watts. In principle, the top-end model of the family is capable of dissipating a little more, but this is only achievable by running a special Intel utility (it is not available to the general public), the task of which is to make all transistors on the processor work. With the help of commercial software, this result is almost impossible to achieve.

Now let's move on to the question of whether it is possible to use the sensor readings from the BIOS or specialized software to evaluate the efficiency of the cooling system. To do this, you need to understand what temperature the user sees in the BIOS settings or motherboard software. The fact is that there are two thermal sensors on the crystal itself. One thing, the TCC control sensor, we will temporarily forget. The second (in Fig. T diode) is a thermal diode, in which the anode and cathode are brought out to two contact pads of the processor in the LGA4 package (for the LGA775 socket). There are several models for using this sensor. For example, the board has a so-called current comparator and an ADC circuit that converts the difference between the currents of a reference and a specific sensor into a number and informs the user of this value through the BIOS or specialized software from the board manufacturer, after converting this value into temperature according to an existing template, which may be wrong. That is, when reading the number 12, which should correspond to a temperature of 40 ° C, we translate it into 47 ° C or, even worse, we read the number 16 from the sensor instead of 12, which corresponds to 70 ° C.

Thus, we see the so-called temperature of the crystal ... which has already been measured once, but in a different place and in a different way. This is where the greatest number of problems are hidden, here are a few of them. Firstly, the sensor shows the temperature in a particular place on the crystal, and if it is 100 °C at this point, this does not mean that the entire crystal has the same temperature. Its value, displayed on the monitor screen, largely determines the application software used. Namely: at 90% CPU load while playing DOOM, it will be 70 °C, and at the same 90% load in Photoshop - 55 °C. Those. the temperature at this point depends on which nearby CPU blocks are being used the most.

Secondly, the conversion circuit on the board may not be calibrated (most often the calibration correction is done through the BIOS) or simply fail, and the specialized software of the motherboard may be erroneously programmed for an incorrect value template. For these reasons, Intel strongly discourages the use of this sensor's values ​​(in BIOS or board software) to perform thermal validation work on assembled PCs. An example is , which examined the performance and thermal characteristics of the Intel Pentium Extreme Edition 955 processor on the Intel D975XBX motherboard. After taking a lot of temperature measurements with this (not recommended) sensor and getting higher values, the reviewer concluded that the maximum power dissipation of this CPU is 200 W, and not 130, as Intel claims.

Employees of one of the popular English-language Web resources faced a similar situation. When they saw that the sensor was showing abnormal temperatures of 100°C or more, they contacted Intel, and after unsuccessfully trying to fix the problem through a BIOS update (most often this eliminates abnormal readings), they had to replace the board. In addition, the experience of overclocking this processor (with an unlocked multiplier) suggests that with a standard cooling system, the Pentium Extreme Edition 955 can be overclocked to 4.2 GHz without core frequency modulation (more on that later). And it is worth recalling once again that 130 W is a design characteristic of the cooling system, not the processor. In other words, this was a confirmation of the manufacturer's recommendation not to use these values ​​for evaluating the efficiency of cooling systems.

The question arises: why such a sensor, where can it be used? Its main purpose today is to control the fan speed of the cooling system for the LGA775. The same circuit reads this sensor and, using the fourth wire of the cooling fan (connected to the motherboard), uses PWM modulation to control the fan speed. This scheme differs significantly from that used in the Socket 478 cooling system, where the fan was controlled by a temperature sensor located above the engine, under the fan cover marked with Intel. With such a scheme, it was necessary to take into account the inertia of the cooling system, and therefore the fan ran at a speed much higher than necessary, which means that the noise was higher. The temperature of the processor could rise sharply (point T diode), but we would only feel it after a long time - the temperature sensor, which is designed to immediately respond to all changes, is located at point T ambient . So I had to turn the fan at a speed of 2000, and not 1500 rpm.

On the LGA775, the T diode temperature control system instantly responds to temperature increases and increases the speed. As in the previous case, the board manufacturer may make a mistake in programming the control system and overclock the fan when it is not necessary. This problem with uncalibrated sensors or erroneous programming will be fixed in the next generation of Broadwater chipsets (i965), where the temperature reading and fan speed control circuitry is part of the system logic. In addition, the sensor(s) on the Conroe processor will become digital (the digital sensor scheme already works on the Intel Core Duo and is called DTS).

As an intermediate result, we note the following. The TDP of a processor is used as a starting point when calculating the thermal efficiency of the cooling system for that CPU. The use of a temperature sensor (T diode) for the fan speed control circuit is one of the most advanced mechanisms for reducing PC noise today, at least in terms of the processor cooling system. However, the readings from this sensor should not be used as an accurate estimate of the thermal efficiency of the processor cooling system and the thermal performance of the system.

The behavior of the CPU when overheating

We will separately consider how the Intel processor behaves when the cooling system cannot cope with heat removal. This is controlled by the second sensor on the CPU, which is completely autonomous and there is no access to it (in the figure it is T prochot). All threshold values ​​for it are "sewn up" at the factory at the manufacturing stage. There are two of them - T prochot and T thermtrip. When the sensor reaches the first value, the modulation of the processor core frequency starts. There are two schemes - TM2 and TM1. Most often, the board manufacturer decides which one to use, but Intel recommends using TM2 whenever possible. In this case, the processor multiplier changes to 12 (2.4 GHz for new samples) or 14 (2.8 GHz for old ones), and then the core supply voltage is reduced. When the temperature normalizes, the CPU returns to the nominal operating point in the reverse order. When the supply voltage is changed, the processor is available and working, while when the multiplier is changed, it becomes unavailable for 5 or 10 µs (depending on the model).

According to the TM1 scheme, the core frequency is modulated - out of 3 ms, the core is idle for 1.5 ms and works for 1.5 ms. She also has a software option to control the duty cycle. This scheme is used by utilities that reduce the noise of the cooling system. It is clear that you have to pay for this with performance, there are no miracles. The purpose of both schemes is simple: if the processor overheats, it must be slowed down, allowing it to cool down, which is better than immediately stopping work - you can at least save the files. As soon as the processor has cooled down and the sensor "felt" it, the TCC (Thermal Control Circuitry) circuit is turned off. Of course, a small hysteresis is added to avoid constant mode switching.

For TM2 and TM1, their inclusion manifests itself in the form of a slowdown in the system. If this does not correct the situation, the sensor immediately turns on the THERMTRIP circuit, all internal blocks of the processor are stopped and a signal is generated instructing the voltage converter (VRD) to stop supplying power to the CPU. The approximate value of the temperature at which this situation occurs is 90 °C. More recently, it has become possible to turn on TM1 / TM2 circuits when the VRD overheats: the processor slows down and starts to consume less, and the VRD can “take a break”. On the Pentium D, instead of the PROCHOT# signal line, FORCEPR# is used to activate the processor slowdown when the voltage converter overheats.

The presence of a separate sensor for the overheat control circuit creates a new group of problems. We can see the temperature T diode = 100 °C on the processor, and on the T prochot sensor it will reach only 70 °C, i.e., according to the readings of the first sensor, the processor should have stopped a long time ago, but it is still functioning. And again, everything is determined by the software profile, which can affect the readings of these sensors in different ways. The most annoying thing about this protection scheme is that it is disabled by default, and it is the motherboard BIOS's job to enable it. (Forgetfulness of the BIOS designer or his mistake can cost the owner of the PC dearly). The latest Conroe processors use the same sensors for both the fan speed control circuitry and the CPU's thermal management. This should eliminate the problem of inconsistent readings from the sensors. This scheme is implemented in Intel Core Duo (Yonah) - already mentioned DTS. The summary is simple: the developers of the processor are doing everything so that even if it overheats, it remains possible to continue working. Even in the event of catastrophic overheating, you don't have to worry - the CPU itself and a properly designed motherboard with the correct BIOS will not allow themselves to be burned.

Further is better

In conclusion, we will touch on one of the most important questions: what is Intel doing to reduce the power dissipation factor? There are two main ways. The first is to disable those processor blocks that are not currently in use at the microarchitecture level. This scheme is most actively used in mobile microprocessors. The second way is to make changes at the level of semiconductor materials. One of the main goals in the implementation of the 65 nm process technology was to reduce leakage currents, and this was achieved - their values ​​decreased hundreds of times. As a result, for example, we got dual-core microprocessors of the 900th models of the C-1 stepping, which “fit” in a 95 W thermal package at frequencies up to 3.4 GHz inclusive.

Naturally, the story would be incomplete without an attempt to look into the near future. Expected in Q3 this year is a desktop processor codenamed Conroe, which at launch will be the quintessence of Intel's power-efficient performance innovations. Expected 40% performance improvement (over Intel Pentium D 950) in SPECint_rate test and even higher gaming rating, while dissipating only 65W of thermal power, using more advanced fan speed control and thermal control circuitry.

The material presented in a number of places was deliberately simplified, but, we hope, has not lost its relevance. Detailed information on the thermal characteristics of Intel processors can be found at support.intel.com in the following documents: Thermal and Mechanical Design Guide (TMDG), Thermal Design Guidelines, Processor Datasheet, VRD Design Guide.

So, get acquainted - TDP. As you can see from the title, TDP stands for "Thermal Design Power". This value indicates the maximum amount of heat that the chip's cooling system must dissipate.
Manufacturers take it equal to the maximum power that the chip consumes. Power consumption is easier to measure, and eventually all of it (with the exception of negligible electromagnetic radiation) will be dissipated as heat.

History of Desktop Processors by TDP

The table below shows TDP values ​​​​for iconic (in my opinion) models of Intel desktop processors.

Model	Frequency	TDP
Pentium	75MHz	8.0W
Pentium MMX	200MHz	15.7W
Pentium II 300 (0.35µ)	300MHz	18.6W
Pentium III 600 (0.25µ)	600 MHz	43W
Pentium III 1000 (0.18µ)	1GHz	35.5W
Pentium III 1333 (0.13µ)	1.33GHz	34W
Pentium 4 1.5 (0.18µ)	1.5GHz	58W
Pentium 4 2.8 (0.13µ)	2.8GHz	68W
Pentium 4HT 672 (90nm)	3.8GHz	115W
Pentium D960 (65nm)	3.6GHz @ 2 cores	130W
Core 2 Duo E6850 (65nm)	3GHz @ 2 cores	65W
Core 2 Quad Q6600 (65nm)	2.4GHz @ 4 cores	95W
Core 2 Quad Q9550S ​​(45nm)	2.83GHz @ 4 cores	65W
Core i5-680 (32nm)	3.6GHz @ 2 cores	73W
Core i7-3930K (32nm)	3.6 GHz @ 6 cores	130W
Core i7-3770K (22nm)	3.5GHz-3.9GHz@4 Cores	77W

According to the logic of things, with a decrease in topological norms, heat release should decrease. However, the number of transistors on a chip grew much faster than the heat dissipation of an individual CMOS pair decreased. This was the reason for the pattern, which can be clearly seen in the table. The infamous gigahertz race led to the fact that the Pentium 4 set a kind of anti-record, exceeding the TDP of 100 W in its 3.8 GHz modification. Obviously, it was impossible to put up with such a situation: the computer looked more and more like a time bomb. And the conclusions were made - heat dissipation began to wane.
And don't let the high TDP of top processors like the Core i7-3930K fool you. These are very special representatives of the processor family, and those who are willing to pay a tidy sum for them will probably worry about the appropriate cooling. In general, the TDP of Intel processors for Lately decreased significantly and continues to decline.

A small digression into the history of cooling systems

Already in the era of the first Pentium processor, computers began to use active cooling, which was a ridiculously sized radiator and the same “propeller”.

pictured Intel Pentium 200 MMX with the fan removed

It was possible to get by with passive cooling, using a slightly more developed radiator, but in those days they did not care too much about noiselessness. Of course, the matter was not limited to one processor cooler, hard drives and power supplies made a significant contribution to the overall system noise.
Cooling systems have evolved smoothly in parallel with the growth of heat dissipation of processors, and finally ...

In the era of Pentium 4 processors, monstrous coolers and alternative cooling methods were developed: liquid, cryogenic, nitrogen. For those interested in history, I will give a link to the article “Millennium Coolers” by a friend LIKE OFF from 2001.

Nowadays, low noise for PC has great importance, many enthusiasts try to build a computer as silent as possible, ideally with completely passive cooling.
This is quite a daunting task. In such cases, a processor with a TDP of no more than 40W is most often used. You can choose a model with a high TDP and lower its frequency and core voltage. (Power is proportional to the frequency and the square of the supply voltage).
The result might be something like this:

With a TDP of more than 50W, it is already difficult to do without active cooling. Even if the processor cooling is passive, you need good air circulation inside the case.

The behavior of the processor when overheating

Those whose acquaintance with computers began a long time ago probably remembered the legendary video from the Tom's Hardware team. (I cannot provide a link to it for ideological reasons). These guys figured out what would happen to the processor if it was during The situation is actually quite possible: the cooler may fall off during transportation, or the fan may break in the cooling system.And finally, the most common problem is when the thermal interface between the processor and the cooling system loses its heat-conducting properties over time.
What happens when the processor temperature exceeds the limit? Obviously, nothing good, but the processor still has some self-defense. Starting with Pentium 4, when the temperature reaches about 90°C, the so-called throttling will turn on: the processor will start skipping cycles, slowing down its work and reducing heat dissipation. Of course, left without cooling, the processor will not be able to provide even the slightest acceptable performance.

Mobile Computing.

For laptops, the main aspect of TDP is power consumption, as it has a direct impact on battery life. The TDP of Atom processors, most often used in netbooks, is in the range of 2-10W, and most laptop processors are 15-40W.
According to my estimates based on network research, a 15" laptop with discrete graphics and a processor with a TDP of 35W consumes about 80W in total. You can estimate the contribution of the processor to the total power consumption of the laptop as 30-40%. Of course, this is true only at maximum load on the processor Most of the time, the processor rests, power-saving technologies come into play, and its share in total power consumption decreases.
Note that despite the low TDP of mobile processors, it is sometimes problematic to implement effective cooling inside a cramped case, so overheating in laptops is even more common than in desktops.

Conclusion

In general terms, I talked about TDP. This topic can be developed in two directions: to consider the reasons for the power consumption of CMOS circuits, which include processors, and to talk about power-saving technologies used in modern Intel processors. I propose to vote for one of my comments on this article: "Energy Saving Technologies" and "Power Consumption in CMOS Circuits".
Those with the most votes will decide the next topic. Also in the comments are welcome real stories about the fight against the temperature inside the computer, victories and defeats in it.