ENCYCLOPEDIA OF RADIO ELECTRONICS AND ELECTRICAL ENGINEERING Improved cooling of microprocessors. Encyclopedia of radio electronics and electrical engineering Encyclopedia of radio electronics and electrical engineering / Computers Recently, the practice of "overclocking" microprocessors, that is, their operation at a higher clock frequency than prescribed by the manufacturer, has become widespread. This is based on a large reserve of technical capabilities of processors and often (if the motherboard chips allow it) is fully justified. Moreover, a fast processor costs much more than a slow counterpart. However, one of the main obstacles to increasing the clock frequency is the inevitable overheating of the processor, which requires better heat removal from it. First of all, let's figure out why the temperature of the microprocessor increases with an increase in the clock frequency and what troubles this leads to. The power consumed by the processor from the power source and dissipated in the form of heat into the surrounding space consists of two components: static and dynamic. The static part of the power is consumed by logic elements that are in a stable position. In the general case, it depends on the state of the element (logical 0 or 1), but since there are millions of them in the processor, it remains constant on average. Dynamic power is spent on transferring a logic element from one state to another. At this time, the transistors forming the element open and close, the capacitances of junctions and connecting circuits are recharged, and other processes occur that cause a short-term increase in power consumption. It can be assumed that a certain portion of electrical energy is consumed for each switching. The more frequency the element switches, the more such portions it consumes per unit time and the more power is dissipated. It must be said that the ratio between dynamic and static power for logical elements of different types is not the same. For example, the fastest ESL (emitter-coupled logic) elements today have practically no dynamic component and the power they consume is almost independent of frequency. Elements of the CMOS structure, on the contrary, consume almost no energy in the static mode. All power consumption is dynamic and directly proportional to the switching frequency. Other types of logic occupy an intermediate position. Any LSI, including a microprocessor, contains many elements, sometimes of different types, and the amount of thermal energy released always depends to some extent on the operating (clock) frequency, increasing with its increase. As is known, the overheating of a heat-generating system, i.e., the temperature difference between its surface and the environment, is proportional to the dissipated power. Developers and manufacturers of microprocessors take this into account as one of the factors that determine the maximum allowed clock frequency. As the clock frequency increases, the temperature of the microprocessor will inevitably increase. Even if we neglect the trivial "burning" - the complete failure of the microcircuit, overheating leads to very unpleasant consequences. As the temperature rises, the noise immunity characteristics of logic elements deteriorate. This is due to the fact that the resistance of open transistors increases, and closed - decreases. As a result, the levels of logical 1 and 0 approach and the interference, whose amplitude at normal temperature was insufficient to switch the element, becomes dangerous. It has been proven that there is a certain critical temperature, above which the probability of failure increases sharply (for example, from a value of the order of 10-15 h-1 to 10-7 h-1), although the element continues to work. For a Pentium II processor containing 7,5 million transistors, this means that failures will occur almost every hour. Failure sometimes goes unnoticed, spoiling, for example, just one digit of the result of calculations. In more dangerous cases, it causes the control computer to issue the wrong command to the managed object. When a glitch corrupts a jump command in an executable program, the computer will typically "hang", executing a nonsensical sequence of commands. Hangups are also associated with thermal breakdown of the most loaded elements of the processor. Such a breakdown is usually reversible, and after cooling in the off state, the computer's performance is restored. From my experience (I have an AMD 5x86/133 overclocked to 160 MHz) I can say that if the fan was accidentally turned off, the processor would "hang" after running for eight hours, but after turning on the fan everything returned to normal. Measurements (by applying a conventional thermometer) showed that the processor began to hang at a surface temperature above 41°, and at 40° it worked normally. Thus, overheating of the microprocessor leads to an increase in the intensity of failures in its operation and even to failures. All this must be well understood and taken into account when an attempt is made to overclock the processor to higher clock speeds. The main conclusion is that. that it is necessary to take care of removing the increased amount of heat and cooling the processor to a temperature below the critical one. For cooling, heat sinks are used - metal plates with a sufficiently large surface. Unfortunately, the heat sink efficiency does not increase in proportion to its area. It is increased by blowing the heat sink surface with a fan. It must be said that most of the processors used in modern computers are designed to work with a blown heat sink (it is called a "cooler" from the word cool - cold), without which it is forbidden to operate them. So we can only talk about increasing the efficiency of this device. Fortunately (or unfortunately), there is a reserve. Due to the unevenness of the surface, the standard heat sink does not adhere tightly to the microprocessor case, there is a layer of air between them that prevents heat transfer. Thermal resistance (the so-called coefficient of proportionality between the temperature difference at the boundaries of the layer and the transmitted thermal power, measured in degrees per watt) of the layer can be reduced by making it thinner and filling it with a substance that conducts heat well. The first is achieved by grinding the contacting surfaces, the second - by lubricating them with a special paste. To achieve the goal, you have to work a little. On a flat surface (it is better to take a sheet of glass), put sandpaper and. moisten it well with machine oil and straighten it, sand the surface of the heat sink. adjacent to the processor. This should be done without pressure in a circular motion, constantly adding oil and turning the part like this. so that the entire surface of the thermal contact is ground evenly. You need to start with coarse sandpaper, gradually moving to a finer one (up to "zero"). When the surface becomes evenly matte-mirror, grinding can be stopped and the heat-conducting paste can be dealt with. KPT-8 paste is sometimes found on sale, but this is rare and far from everywhere. In its absence, you can get by with improvised means. Of all liquids, mercury has the highest thermal conductivity, but due to the toxicity of vapors, electrical conductivity and high chemical activity, it should not be used. It is followed by water (thermal conductivity 0,648 W/m rad.), but it is electrically conductive and evaporates quickly. Of the non-drying liquids, the thermal conductivity is maximum for glycerin (0,283 W/m rad.). In addition, it increases with increasing temperature (for other liquids, it decreases). Take some glycerin and add about twice the volume of aluminum powder to it. Grind and stir this mixture well to form a uniform, viscous silvery paste. It should stick and smear, but keep its shape and not spread. This paste does not conduct electricity. but you should still avoid getting it on the boards of computer nodes and microcircuit pins. Using a brush, apply a small amount of fins to the contact surfaces of the processor and heatsink. Some try to spread more, naively believing that since the paste is heat-conducting. it should be applied thicker. On the contrary, the smaller the better. It is necessary that the layer be as thin as possible and evenly cover both surfaces, displacing air and filling in all irregularities. Carefully place the heatsink onto the processor and move (lap) it slightly to expel any remaining air and excess paste in the gap. Do not forget to fix the heat sink, and on it the fan and connect it. Now everything is ready. To check, "drive" a processor test in the Troubleshooter system for a couple of hours, and if no failures are found, you can work calmly. Author: I. Korznikov, Yekaterinburg See other articles Section Computers. Read and write useful comments on this article. Latest news of science and technology, new electronics: Artificial leather for touch emulation
15.04.2024 Petgugu Global cat litter
15.04.2024 The attractiveness of caring men
14.04.2024
Other interesting news: ▪ Trade-in to Samsung wearables ▪ Supercapacitors made from cigarette butts ▪ Intel Achievements in Silicon Photonics ▪ Compact 36V DC-DC power modules LMZM33602/3 News feed of science and technology, new electronics
Interesting materials of the Free Technical Library: ▪ section of the site Civil radio communications. Article selection ▪ article Tsiolkovsky Konstantin. Biography of a scientist ▪ article What drug was officially issued to Wehrmacht soldiers? Detailed answer ▪ article A simple bayonet with two hoses. Travel Tips ▪ article Stabilizer for four voltages. Encyclopedia of radio electronics and electrical engineering
Leave your comment on this article: Comments on the article: Shamshi My temperature reaches 60 degrees, and the monitor starts to turn on and off. All languages of this page Home page | Library | Articles | Website map | Site Reviews www.diagram.com.ua |