1. Introduction
2. What is the overclocking of graphics cards?
3. Does this make sense?
4. Is it dangerous?
5. Warranty issues
6. Junior and senior models
7. How much can I get when I overclock the video card?
8. What are synchronous and asynchronous frequencies?
9. What is the chip process and memory access time, how do they affect overclocking?
10. How to effectively improve cooling?
11. Tips for improving cooling from our overclocking expert
12. Than and how to disperse?
13. What is an "auto brake" and what is it served with?
14. Instead of concluding

Introduction

Surely almost all readers are more or less familiar with the concept of "overclocking", but not everyone can clearly imagine how to correctly and painlessly overclock their video card, and do not know some of the subtleties encountered during overclocking.

This material is intended just for beginners in overclocking, gathered to disperse their video card. Now we will try to explain quite clearly and clearly about many problems encountered in overclocking, how to solve them, and, of course, we will share some useful tips. But let's not breed extra demagogy and go directly to the matter.

What is the overclocking of graphics cards?

Under the overclocking of video cards is meant an increase in the operating frequencies of the video card. But also overclocking can be called and other ways of freelance performance increase, whether it is unlocking additional pipelines on the Radeon 9500 / 9800SE, or the inclusion of HyperZ on the Radeon LE.

Does this make sense?

Undoubtedly. Overclocking a video card is, without exaggeration, the most effective means of increasing computer performance in games and other 3D applications, except for only those cases where performance hinders the speed of the platform (read, connect the processor + memory).

Is it dangerous?

No. The chance of burning a graphics card during overclocking is much less than the processor's allowance. In general, the video card can not burn out from the overclocking, but it can overheat, although in most cases, when the graphics processor overheats, the machine simply hangs.

On the other hand, working on freelance frequencies, as well as forcing the work of any other component of the computer, significantly reduces the life of the card. And this feature could be a very serious deterrent, if not for one "but" - the life of the video card is more than eight years, and even with overclocking it is less than five years will not. And if you look at the existing technology race, in gaming computers the cards do not hold for more than two years, so if you do not plan to leave the video card for about six years, you can calmly disperse it.

Warranty issues

The main side effect is that theoretically you completely lose the warranty for the purchased video card. But do not get upset, because even if the card fails, then to prove that it was due to overclocking is very, very problematic ))

Junior and senior models

It's no secret that new models of video cards are produced by so-called "rulers". It happens in the following way - any chip comes out, then several video cards with different frequencies, and in some cases also different designs with different width of the memory bus, are released on its basis.

However, in any case, the younger model having significantly lower frequencies than the older one will be built on the same chip, and consequently, the chip installed on the younger model will in most cases be able to earn at the frequency of the older, or even higher.

But here everything is not as smooth as we would like to see it. The fact is that in the manufacture of video cards, the chips are pre-tested, and some chips that failed to pass tests at the maximum frequencies set for the older model are sent to the production of younger ones. But if we take into account the fact that modern production technology is thin enough, such a "marriage" is not so common today.

As for the memory, everything is a little worse here - the younger models are equipped with slower ones than the older chips, and it's not always possible to overclock the memory on the younger model to the higher frequencies.

In general, if you look at the percentage of the average overclocking of the younger models in comparison with the older ones, the first ones have a significant advantage due to the initial frequency margin. The older models work almost at the limit, and squeezing out additional megahertz will be more difficult.

What is the gain you can get when overclocking a video card?

Here everything depends on the testing conditions, and of course on the degree of frequency increase. Worst of all this with noname-cards produced by Chinese craftsmen and flagship models of rulers (for example, GeForce4 Ti4600 or RADEON 9700 PRO). In the first case, the cards are weakly overclocked because of poor-quality components, which Chinese craftsmen equip their products, in the second case, the boards already work almost at the limiting frequencies, as we said in the previous paragraph.

As a rule, when overclocking such cards, only 15-20% of the frequency increase can be achieved. With medium and low-end models in the lineup the situation is better, the potential for increasing the frequencies more and overclocking such cards can improve performance by 20-40%.

The best option is all sorts of overclocking samples. They can increase by 35-50%, and sometimes even more.

Now a few words about cards with a reduced structure of memory organization. There is an opinion that it is senseless to overclock the chip on such cards, but personally I completely disagree with this. The fact is that users of such cards, as a rule, play in modes such as 800x600 or 1024x768, and the low bandwidth memory capacity in such modes slightly limits the performance, but on the graphics processor the load, on the contrary, is greater.

What are synchronous and asynchronous frequencies?

The frequencies of the chip and the memory of the video card can be synchronous, that is, the same, or asynchronous, in other words, different. But what is the difference?

When the video card is working and data is exchanged between the graphics processor (chip) and the video card's memory, signals are synchronized. In case the chip and memory operate at the same frequencies, the signals pass simultaneously and do not take additional time to synchronize them, but if the frequencies are different, before the data exchange, the video card must synchronize the signals, which, of course, takes a little time.

From this, without thinking twice, you can make a simple conclusion that at synchronous frequencies the graphics card will work a little faster than on asynchronous ones. But there is one thing ...

Synchronous frequencies are advantageous only if the possible asynchronous frequencies are not too different. For example, we have the opportunity to put the maximum frequency of 450/460 and more frequency can not be set. In this case, it will be much more effective to sacrifice memory for ten megahertz players for the sake of synchronism to deliver 450/450 - in that case the video card will almost certainly be faster. However, if we have the opportunity to put frequencies, for example 475/450 or 450/480, such options will be preferable to synchronous 450/450 due to much larger resulting frequencies.

What is the chip process and memory access time, how do they affect overclocking?

Any overclocker must know what the chip process and memory access time are. Knowledge of these two definitions blinks significantly in the approximate determination of the maximum frequencies of the overclocked graphics card.

But what is it? In the manufacture of any chip, the size of the elements of the chip plays a very important role, because the degree of integration can be different, two million transistors can be "stuffed" into one chip, in another - one hundred and two. And when the physical size of the chip chip is limited, the size of the chip elements and the distance between the elements in the crystal play a very large role. This size is called a technological process, and the smaller it is, the more elements are placed in the chip, the smaller the currents require the elements for power, the less energy the chip releases, and finally, at higher frequencies it can work.

At the moment, the vast majority of chips produce 0.13 and 0.15 microns in process technology, and 0.11 microns at the stage of active development.

As for memory, access time plays an extremely important role here. Any memory chips have a time declared by the manufacturer during which the information is read from the memory cell, and the shorter the time, the faster the memory operates, and the more its operating frequencies. The dependence of the approximate operating frequency on the time of memory access is extremely simple, and it can be described by the following formulas:

Memory frequency DDR = (1000 / access time) X 2
Memory frequency SDR = 1000 / access time

The next question is how to know the time of memory access. Typically, the access time is hidden at the end of the first line of the marking. For example, on the memory chips Samsung at the end of the first line you can find an inscription such as TC-33 or TC40. This means that the memory has access time of 3.3 and 4 nanoseconds, respectively, although in some cases, the time is not indicated by a digit, but by special marking, for example Samsung memory chips with access time 2.8 ns. Are designated as GC2A.

Do not forget also that the exact information about the memory chip can be obtained on the manufacturer's website, or simply by using the search on the line with the memory marking in the same Google.

How to effectively improve cooling?

The most important problem with overclocking is overheating. In order to avoid this, it is necessary to ensure good cooling. You need to start from the enclosure (because if there is not enough ventilation inside the case, then no additional cooling will help), put the cooler on the blow-in front, behind under the blow-out power unit, you can also put the blender under the card, to remove hot air from it . Naturally, all this applies to those who have a case closed, but with an open case everything is simpler, it's enough to put a fan perpendicular to the video card to blow it. If none of this has helped, then it's time to move on to more cardinal methods.

Memory chips also get hot, especially when overclocking on freelance frequencies. Overheating memory can lead to image defects, so-called artifacts, so cooling the memory should also not be neglected. To cool the memory chips, you can find ready-made kits from such brands as Thermaltake, or you can make radiators on your own using a heatsink from the old processor Mount the radiators in several ways, for example, on hot melt gloves or on brackets.

As a heat-conducting interface between the graphics chip and the cooler of the video card, as a rule, a sticky mixture is used, we are called in the people as a thermal carrier. This rubbish should be wiped off and replaced with a high-quality thermal paste, for example, domestic KPT-8.

On some video cards with an open crystal, a special protective frame is laid around the chip, preventing the possibility of an accidental chip break with a careless installation of the cooler. Such a framework, for example, is installed on almost all ATI RADEON 9500/9700/9800 / X800 graphics cards. The removal of the frame makes it possible to improve the contact between the base of the radiator and the chip, but significantly increases the chance to chip the chip.

Do not forget about water cooling, which by efficiency will be better than any air, but much more expensive and difficult to manufacture.

Tips for improving cooling from our overclocking expert

And now a few tips for improving the cooling, which our overclocking expert shared, Mr. Oleg Ivanov aka m @ ddog, who took a direct part in writing this material.

1) Air cooling is the cheapest, but at the same time very effective way to improve the cooling of the video card. The cooler for this must be selected copper, a small height (but not low profile). If copper you can not afford, then you can put and conventional aluminum. Choose an unpainted radiator with a qualitatively processed sole, put a fan such as your ears suffer - it's better to put it on 6-7 thousand turns, in which case you can slow it down, but it's very problematic to disperse an initially slow fan. This type of cooling requires periodic maintenance, which consists in periodically cleaning the cooler of dust. Also worth noting is the large drop in efficiency with increasing air temperature (read the onset of summer).

2) Water cooling system. Just a couple of years ago, even thinking about installing the video card "dropsy" was ridiculous - video cards simply did not require such a high-quality cooling. But time goes by, cards all increase the level of heat release, and now dropsy becomes an indispensable attribute of good overclocking of modern video cards. VCO (read, water cooling system) can be bought, but immediately get ready to shell out $ 120-150 for an efficient system. Not ready? Then it can be collected independently. A pump is not particularly powerful - it's better to give money for reliability than for unnecessary liters per minute. The water block is desirable for copper, but it is suitable aluminum or steel. Behind the radiator head to the nearest autodepo. It is not advisable to fill the water drop with water, it is better to fill it with kerosene.

4) Waterchillers. This is the most inefficient and most accessible of extreme systems. In fact, the water chiller is the same VCO, only the temperature of the working fluid is below room temperature. The easiest way - ice, reduces the temperature of the liquid to + 5 / + 7. You can use dry ice - this will bring the temperature of the liquid to -60 / -50, but in this case you need a special pump and special liquid. In the course are also freezers, phase-change systems, etc., etc.

5) Phase change system, freon. The efficiency of these systems is very high in operation, they are simple. But it's almost impossible to buy a ready-made system (~ 1000-1500 $). Collect it yourself at home is almost impossible, and still it turns out terribly expensive.

6) Liquid nitrogen. VERY effective cooling system, besides inexpensive. But when you install it you need to think a lot (preferably threesome )). However, after everything is thought out and done, it is relatively easy to use this system. However, liquid nitrogen - not for long-term use - is suitable only for short-term extreme overclocking.

Also it is necessary to take care of cooling of power elements, especially it is actual after voltmodding. Heat dissipation of power elements is small - 0.2-1 W, so you can do with plates or small radiators. By the way, this measure has not only a moral and calming significance, but also quite practical, the fact is that at high temperatures the accuracy of these details decreases. And they can burn ...

What and how to overclock?

And finally, we come to the very same main thing - than, and most importantly, how to overclock your graphics card ) It is necessary to start with the choice of the tool - after all, in many respects the result of overclocking will be determined by the program with which you will overclock your card. At the moment there is a Tueva heap of all kinds of utilities, differing from each other both in capabilities and in degree of glitchiness. We recommend using PowerStrip for overclocking video cards, and for ATI a small handy program called ATITool. For more details about various programs for overclocking graphics cards, you can read also in our article with the talking title " Software for overclocking graphics cards - what to choose? ". So, we will assume that we have already decided on the program and go directly to overclocking.

Overclocking should be done gradually, slightly raising the frequencies of the chip and memory (at first it is possible together, and at the end one can search for the optimum frequencies separately) in steps of 5-10 MHz and after each lifting test in heavy 3D applications.

After reaching the maximum frequencies, you should once again drive all the tests with the maximum quality settings, you can also throw off 5MHz for full confidence and reinsurance, the speed will not decrease, but stability will increase.

In the process of overclocking, there may be several problems, for example, hangup of the computer in games, the appearance of various image defects and broken pixels (otherwise this phenomenon is called artifacts), overheating of the video card. But let's take everything in order:

• If in the process of playing on an overclocked graphics card the computer began to hang, then with a probability of 99%, it can be said that this is a chip overclocking, and this is treated by lowering the frequency to a stable value at which the hang-up stops. As a rule, in such cases you just need to refine the chip cooling system.
• If during the game you begin to appear on the screen stripes, deletion of textures, snow and similar unpleasant phenomena, it is necessary, by analogy with the chip, simply to lower the memory frequency and once again to conduct a full testing of the video card.

In some cases, hang-ups may also be due to the insufficient power of the power supply unit, simply unable to feed the become over-hungry after overclocking the graphics card. In any case, do not forget to take care of the good cooling.

What is "autobrake" and what is it served with?

Mankind with perseverance worthy of the best application, with all its might, draws nearer the "Judgment Day" - the moment when the machines will become so intelligent that they understand how much we impede their development. Movement in this direction can be seen in everything: washing machines, food processors, cars, etc. Over time, they become more sophisticated, acquire new functions, acquire a user-friendly interface, but at the same time distance us directly from management, allowing only to observe the process from the outside. In cars, these are systems like ESP (which "help" to drive a car), and in video cards - "auto brake".

"Auto Brake", according to the manufacturer, is needed in order to protect the device from overheating by lowering the clock frequency. The pioneer was Intel, entrusting the system to choose whether to work at full capacity, or to skip a part of the bars in case of a temperature increase. Now the turn of video cards has come. Users encountered this function when overclocking the GeForce FX 5900/5950 line. When overclocking these cards, the situation often occurred when the speed was lower at the limiting frequencies than with the "default" setting. This was due to the fact that the manufacturer entrusted the video card to choose at what frequency it would work, based on the readings of the built-in thermal sensor. NVIDIA engineers are easy to understand: very many novice users, overclocking a card, rarely care about improving cooling. Because of this, cards are often broken, which the manufacturer naturally does not like. To improve fault tolerance, and in Russian "protection from a fool", and introduced this function.

But, like all good intentions, it continued to lead the same road. For some reason, the video cards began to "make their own" when overclocked. Even if the temperature was normal, they sometimes believed that they did not want to work at full strength, and lowered the frequencies. The problem was aggravated by the fact that the most common utility for overclocking nVidia cards overclocked the card data via the driver, giving the built-in AI complete freedom of action.

To note the work of "auto brake" is simple. It is enough to run during the overclocking any test that shows the speed of the video card, for example AquaMark 3 and to remove the results of the same program at nominal frequencies. In the event that you notice that performance is actually declining, you can advise several outputs:

• Try to improve the cooling of the video card
  
  
• Pick up and set for the video card such frequencies, in which "braking" does not turn on
  
  
• Dispersing the driver bypass using the PowerStrip utility

Any more specific recommendations on this issue are difficult to give, because each card is individual, but the general principles are as follows.

Instead of concluding

Well, now our material has come to an end. We hope that most of the questions you raised about overclocking were still answered, and now you can quickly and without problems increase the performance of your card in such an effective way as overclocking. However, if you have any questions, we will always be happy to answer them and solve your problem on our forum.

This material was only the first version of our FAQ on overclocking, and it will be constantly updated and supplemented, which, of course, we will certainly write )

And for this, please leave!