If you reached here, should have already noticed that the frequency of operation of the processors (and also of the memories and other circuits) is not fixed but defined by the combination of the frequency of the bus (or the base frequency, in the case of I7) and a multiplier.

As an athlete who can run faster or slower, the processor clock can be adjusted up or down according to need. You can get more performance by increasing the frequency to the maximum supported by it (overclock), often going 40 or even 60% than the frequency pattern, or, if desired, reduce the frequency (underclock), making the heat less processor and consume less energy.

The obvious motivation is to extract more performance from the processor without having to pay more for it, which is always good news in times of crisis. Usually, even the worst can overclock processors to provide an additional gain of 8 or 10% and some specific processors offer 50% or more.

The Pentium E 5200, for example, uses 200 MHz FSB (with 4 transfers per cycle, resulting in the announced 800 MHz) and multiplier of 12.5x. As with almost all current processors, the multiplier is locked, so any change in frequency need to be adjusting the frequency of the FSB via the setup. Increasing the frequency to 266 MHz processor will work at 3325 MHz, and to increase it to 300 MHz you get 3.75 GHz, frequency in which the lowly Pentium E can overcome the majority of Core 2 Quad in games and other applications not optimized for the use of 4 cores.

Naturally, the stable operation depends on the cooler, the frequency of memories, and other factors, but the basic idea is this.

The explanation for such large margins for overclocking is the way the processors are manufactured and sold. To cut costs and simplify the production, manufacturers use the same production line to produce all processors of a given family. We have then a single line for all processors based on Penryn core, for example, all of which go Core 2 Duo, Core 2 Quad (obtained by combining two dual-core processors) and the Pentium E 0045 micron.

Even two processors manufactured on the same wafer are never exactly equal, since small differences of focus during lithography and small variations in the surface of the wafer are some processors that are more perfect than others, which leads to differences in frequency supported the clock and even the electric consumption of each unit.

After the wafer is cut, the processors are tested individually and the best units are intended for families with higher clock, while the other goes to the line of low cost. In a third level we have processors with defects in parts of cache L2, or some of the nuclei, which are part of the die off and are sold as E Pentiums or Celerons.

The process of separation and classification of processors is called "binning", and the use of cores with small defects is called "die-harvesting," two words you hear much to read technical material.

This process aims to extract the maximum of functional processors each wafer and at the same time, enabling the segmentation of the market with processors high, medium and low cost out of these factories and executives happy with the profits:

The opening room for overclocks generous on some models is the fact that the Clocks of processors sold are not determined by the actual capacity of processors, but by market studies, which calculate the prices according to demand and offerings of competitors . In this, the sum is simple fact that there is always more demand for cheaper processors (which are sold from $ 50 abroad) than the high-end models (which can cost up to $ 999).

Therefore, manufacturers end up being forced to sell processors perfectly healthy (able to work at frequencies much higher) as models of low cost. Therefore, it is more advantageous to buy the cheapest processor in the line (or the second cheapest, when the difference is small) and simply cover the difference with a good overclock, investing the savings on other components of the micro. The less money you have to spend, better be a good overclock.

In general, the processors are the best dual-core models that use a lower frequency of FSB, such as Pentium and, working with bus of 800 MHz (compared to the 1066 Core 2 Duo). The reason is simple: with the exception of the Extreme series of processors and Black Edition, all models sold have the multiplier locked, so that increasing the frequency of the FSB is the only way to overclock. With a low FSB, you have a much greater margin to overclock, without a super the motherboard.

A certain amount of luck is also an important factor, since buying a processor with a margin of greater or lesser overclock is a kind of lottery. The basic trick is always looking for the latest series of manufacturing, since the manufacturing process is always improved over time, causing the processors produced overclocks support increasing, until a new technique of production is introduced, repeat the cycle. Almost always, the best series to overclock are rightly among the last produced, when the model is to be discontinued.

Another key component is the motherboard that needs to offer a good range of adjustment of frequencies and voltages, in addition to being able to maintain a stable frequency and higher than the default

In general, any midrange card from Gigabyte, Asus and even the ECS will provide a sufficient number of options, but it is important to prevent the boards from Intel, which despite the good quality offer virtually no adjustments.

In the motherboard is always the best place to invest the money saved in the processor. If you can pay a little more for a plate with solid capacitors, based on a chipset for workstations (such as the Gigabyte EP45-UD3L and Asus P5Q, which are based on P45), you hardly will regret.

There are also boards for enthusiasts, such as how the Asus Rampage II Extreme (based on X58 chipset), which are much more expensive and offer differential circuits such as voltage regulators oversized, with more complete sets of options in the setup, copper and heat extractors -pipes for cooling the chipset and MOSFETs.

They undoubtedly offer margins overclock higher than the mainstream cards of the same manufacturers, but the difference in cost does not pay. They are more geared for those who have money to spend and want to simply get the highest possible frequency and not for those who are concerned about the cost-benefit.

Finally, the question of general cooler and cooling the PC. The coolers that come with processors in-a-box is sized to meet the processor frequency in default, but offer low margins to make overclock. In many cases, you can go relatively far away using the stock cooler (if you use a well ventilated cabinet), but the temperature becomes a limiting long before the processor reaches its maximum.

To achieve the best results, it is best to buy a retail processor (the models without the cooler, which is a bit cheaper) and provide 150 or 200 to buy a real cooler of larger size. The basic rule is that you need a cooler capable of dealing with a gap high-end model with a frequency similar to that which you want to achieve with the overclock.

The question of heat dissipation is a major reason for the dual-core processors allow almost always more generous overclock that quad-colors. With only two nuclei, the thermal dissipation of the processor at full load is nearly the half of a quad-core processor architecture of that operating the same frequency.

Much is said about the risks to overclock, especially in the most extreme cases. It is true that cases of processors and even damaged the motherboard are relatively common, but often they are caused by excessive increases in stress or abuse and not due to the increase in the clock itself.

In general, we can say that simply increasing the frequency of operation, without moving the tensions offers a very small risk, since the processor will simply stop by any one error processing to exceed the maximum frequency supported. In this case, just clear the CMOS (if necessary) and start again.

The next level of risk is to increase the tension slightly (somewhere between 4 and 6%), which remains within the tolerance of the circuit and allows the processor to stabilize frequencies a little higher. The greatest risk in this case becomes the heating, as by increasing both the clock and voltage, the thermal dissipation of the processor will grow in geometric shape. In normal situations, the overheating is not a big risk (the processor will simply stop or reduce the clock to reach a certain temperature, avoiding harm), but prolonged use can be dangerous.

The third level is a lot to increase the voltage of the processor or memory (many come to increase by 15% or more), which is really dangerous and can cause damage quickly. Even if the processor operates with a stable voltage 15 or 20% above the standard, you can be sure that life will be much lower.

We come then to the distinction between two categories of overclockers. The first is the normal users who just want a little more performance, without paying more. If you are one of them will want to avoid extreme overclocks and large increases in tension, since the idea is that the micro continue working in a stable for several years.

The second category is that of the enthusiasts, who simply want to achieve the best possible performance, even for a short time. Being part of that group means investing a good time to test various combinations of settings and run benchmarks, and spend good money on coolers and other accessories and, occasionally, in a processor or a motherboard new.