It all started with the processors that represent the heart of the computer, and everything else is there to send the processor overclocking in a certain way and in the end to provide as much performance as possible. Logically, the PC processor manufacturers themselves were against overclocking at some point, and in a way tried to disable this activity.
In practice, something like this proved to be quite difficult, so enthusiasts always found a way to speed up the work of certain processors. AMD and Intel were generally only in principle strictly against overclocking, especially at times when it suited one or the other manufacturer or was even very useful. Simply put, when someone needed to boost advertising and sell their processors in a certain way, overclocking was left wide open.
When it comes to processors, they have been overclocking for a long time, so the whole story started to develop in the time of 80286 processors, and already in the age of Pentium processors, overclocking became a thing that all enthusiasts looked at.
A more detailed history is not particularly important, the only thing that may be worth mentioning is the moment when overclocking "entered the big door". Of course, it is the Pentium II era and now the mythical Celeron 300A processor.
Then users had the opportunity to easily get the same (and even better) performance of the then most powerful Intel Pentium II processor at 450 MHz, and to pay significantly less for the performance in question.
Everything that happened after that was just a logical sequence of things so that today we come to a situation where Intel and AMD offer "overclocking" models of their processors.
All this, however, was not possible without the presence of adequate motherboards, because it is a basic overclocking tool. Accordingly, the choice of the motherboard became a vital detail for the whole overclocking story.
Logically, all manufacturers very quickly began to use overclocking as an element in which they would offer users more and with which they would fight the competition so that today overclocking becomes one of the main arguments in a direct comparison of motherboards.
Today, it is quite normal to loudly advertise overclocking at all motherboard manufacturers, from automatic overclocking to various LN2 options and possibilities. In line with the development of the complete story, overclocking as a term has become a common thing used in regular IT conversations.
Even less experienced users, who are generally considered to be average computer users in terms of quantity and intensity of use (of which there are more and more today and make up an increasing percentage of users), have at least some basic idea of what overclocking is, although they do not use it themselves.
As it can be said today that if you don't overclock your desktop at least a little, you don't use what you paid for, this guide is mostly intended for those who are afraid of overclocking for some reason. There is no real reason for fear, at least when it comes to easy beginner overclocking, so gather courage, study the instructions, and take action.
The basic question that will be asked mainly by traditionally conservative users, who have been fearing overclocking for years. In a certain period, there was a real basis for the fear of overclocking due to the possibility of the processor "crashing" (mainly due to carelessness or clumsy handling), but the situation today is significantly different.
True, even today Intel and AMD state that the "operation of the prescribed framework" cancels the warranty, but in practice, it is very difficult to prove that the possible failure of the processor occurred as a result of overclocking.
However, as processor manufacturers now overclock their processors themselves through "turbo" functions and offer models intended for overclocking, why shouldn't users take advantage of that.
Besides, all modern processors have very good internal protection mechanisms, so the processor will automatically shut down if a certain temperature is reached or a value is reached that is set as a limit by the manufacturer. As almost all motherboards offer very good options for setting up and overclocking through the BIOS, along with automatic overclocking options, there is simply no reason to avoid overclocking.
When it’s already available, and when so much is being advertised and offered, why not just take advantage of the hardware that’s already paid for and get the most out of it. There are no real reasons against today, so the only question is to what extent someone is interested in getting the highest speed for their money, that is, whether they want to "mess" with it ...
If someone is interested in overclocking, they can be classified into one of three groups according to their situation and desire. The first group represents the largest number of users who simply found themselves with a certain configuration (for which there may be several reasons), so now they want to somehow speed up the computer they have.
The second group of users can again be beginners, but also more experienced users, who have a limited budget and who accordingly want to get as much performance as possible with as little investment as possible. The final group represents those who simply want the maximum possible performance of the computer, with less interest when it comes to the price of certain components.
Basics of overclocking
When we talk about overclocking, we primarily mean the speed of the processor. However, only the acceleration of the processor brings with it some other necessary details, for which the motherboard and memory are in charge. It is the "triangle" processor-board-memory that represents the heart of the computer and, accordingly, represents the main subject of overclocking.
Depending on the platform, the dependence of processor overclocking depends to a greater or lesser extent on the memory, ie the motherboard, so we will deal with the specifics of current platforms separately.
Of course, some basic concepts and ideas remain the same when it comes to all platforms: increasing processor speeds by changing the system bus or multiplier clock, correcting CPU operating voltage, adjusting memory speed and timing as needed, and a few more details.
The basic way to estimate the clock speed you can get from your processor is to compare it to more expensive models that are based on the same core. Simply, if you bought a CPU on e.g. 2.4 GHz, and the strongest model on the same core runs at 3.5 GHz, then this is the clock that should be the first target and which in most cases can be counted as achievable.
If this most powerful model also has a "turbo" function, you can also count on the clock at which the processor in question works in "turbo" mode. If you do not know the details of your processor, which core is in question, and which models are based on it, you will probably get the information most easily through the famous web encyclopedia and with the help of a small program.
The easiest way to find out information about your processor is through the CPU-Z software (free shareware program), which will very easily show you all the important details that are useful to know, as well as the label of the core itself.
After that, you can simply look through Wikipedia to see which models are based on the core in question and on which clocks these models work. Of course, certain processors are capable of running at much higher speeds, as is the case with Intel's latest generation processors.
Before any intervention on the CPU clock speed, it should first be assessed whether the cooling used is adequate for such a task. Today, with each new processor in the box, you get a cooler that the manufacturer intended to work with a certain model.
However, it should be borne in mind that this "purpose" is calculated so that the processor in question is provided with only a sufficient level of cooling, so this is one of how manufacturers to some extent "control" overclocking.
Simply put, with the cheaper models, the coolers that are obtained are in small cases in small proportions and, accordingly, are not adequate for any increase in the operating clock outside the nominal values.
True, it is possible to "squeeze" a little more speed depending on the processor model, but it is minimal. That is why you should always invest in a good cooler, which is not a problem today.
With AMD processors, you get slightly larger coolers, so in some cases, you can avoid buying another cooler with the platform (although this is again recommended). Intel only packs slightly better coolers with stronger processor models, while in all other situations, buying a new cooler is a must.
The ability of the cooler to cool the processor is largely determined by the final range for overclocking, because when the processor reaches a certain temperature, instability occurs, which is always the first sign that we are approaching the limit at which it is necessary to make some changes.
Regardless of the large number of processors that can be found on the market today, we can say that the approximate temperature limit is around 65 degrees. It is not recommended to go over 70 degrees at full CPU load, although in practice it is difficult and possible because system instability will occur.
Processor temperature is affected by the operating clock and voltage. A pure increase in the operating clock greatly affects the temperature, although the operating voltage of the processor is not increased. Not only that but with most current processors, it has been shown that higher speeds increase heating to a greater extent than voltage does.
When a combination of higher speed and voltage is made, it is clear that the heating increases drastically. When it comes to the operating voltage itself, the maximum recommended values vary from core to core, so before voltage correction, it is advisable to first check what the recommended values are for the model you own.
In principle, this is moving into the territory of more advanced and serious overclocking, but it is still not bad to get acquainted with this phenomenon, at least basically. So, it is a voltage drop that occurs in situations where the processor is running at a much higher frequency than the nominal when the processor is supplied with a voltage that is much lower than the value set in the BIOS.
Then the CPU simply runs at a lower voltage than it needs to maintain stable operation at high speed, which logically results in system instability. Although many enthusiasts have largely blamed motherboard manufacturers for this, vDroop is part of Intel's reference design when it comes to CPU power, as part of a system to reduce the temperature at maximum loads.
Of course, with the processor at the nominal operating clock, "vdroop" is not a problem, but as soon as the overclocking starts a bit stronger, there can already be problems with stability.
Therefore, you should keep in mind "vdroop" and when testing the stability of an overclocking value, you should also pay attention to the current-voltage that the processor receives. On this occasion, we will not go into details about why anything can happen to this phenomenon, but it is important to say that every board manufacturer today has a certain solution, especially with top board models.
So some have additional modes for the power unit, LLC (Load Line Calibration), and similar things that serve to reduce the "vdroop" to a minimum. If you have a board that has one of these options, you should use them, because you will enable stable operation at a higher CPU clock speed.
Bus and Multiplier
The base clock of the motherboard with which the chipset communicates with the processor is called FSB (Front Side Bus). Logically, the higher the link speed between the chipset and the processor, the better the performance of the whole system. AMD and Intel have improved their platforms over some time so that more instructions are sent over the FSB in one clock cycle, thus effectively speeding up CPU-chipset communication.
However, what interests us is that basic FSB clock, so for example, if we have an AMD processor running at 3000 MHz, that operating clock is obtained: FSB (200 MHz) * multiplier (15) = 3000 MHz. Here we come to the multiplier which is the second part of the equation in determining the clock speed of the processor.
The word itself speaks for itself, it is only a number with which the working clock of the highway is multiplied. This number is determined by the manufacturer and for the vast majority of windows it is fixed "higher". This means that the multiplier can only be changed to lower, which effectively reduces the final clock speed of the processor.
The multiplier can be changed to more only for specific processor models (eg Intel Extreme Editon models, Core i7 with "K" suffix, AMD Black Editon models), which means that the manufacturer directs these processors directly to overclocking.
Logically, the "free" multiplier models in question are more expensive, so this is one of how manufacturers "control" overclocking. If you do not have a processor with an unlocked multiplier, the only way to overclock is to increase the clock speed of the system bus (FSB).
However, with newer generations of processors, things are significantly different. First, the FSB itself as a term is no longer accurate today and is used only "out of habit", because there is still a basic clock that gives the final speed of the processor with the multiplier.
In order not to go into these details on this occasion, we will continue to use the usual term here as well. By moving the memory controller to the processor core, one of the main tasks of the chipset (northbridge) has disappeared, which is the case with Intel LGA1156 and LGA11366, as well as with the AMD platform.
With LGA1155 processors (Sandy Bridge), Intel has taken it one step further by integrating a PCI Express controller directly into the core, eliminating another vital feature of one part of the chipset (northbridge). This architecture drastically changes the rules of overclocking, so the LGA1155 platform is much more limited in this regard than it was before.
Memory and Multiplier, again
Overclocking rules have long implied memory as a very important item for the final result. Since without changing the CPU multiplier, the operating clock can only be increased by increasing the speed of the system bus, this move entails a higher clock than the memory must achieve.
For example, if we have a 333 MHz FSB, the 1: 1 memory will run at 667 MHz and by increasing the FSB clock to 400 MHz and the memory speed will automatically jump to 800 MHz. Therefore, you should always keep in mind whether the memory you have can withstand the clock speed that you require from it.
Don't be confused by "double clock", the DDR label for memory means "Double Data Rate", so through one cycle, twice as much flow is achieved, which again, in the end, means that e.g. with an FSB of 400 MHz it achieves a memory clock of 800 MHz.
The memory speed is controlled via a memory multiplier, ie the number with which the bus clock is multiplied to obtain the final memory clock. Since there are memory capacities on the market for speeds over 2000 MHz, it is clear that these multipliers are necessary to enable the user to get the most out of such memory.
What is done during overclocking is to reduce this multiplier as compensation for the increase in bus clock speed, thus keeping a certain memory in its optimal range. Of course, there are situations when a 1: 1 ratio (multiplier) is reached and when the memory can no longer send the FSB speed, so it is very useful to have the best possible (faster) memory.
However, with the latest generations of processors, it is difficult to get into this situation, so with these platforms, the issue of memory can be approached more relaxed. Of course, the memory can also be overclocked, but we will leave the question of fine-tuning the memory for another occasion.
For beginners, the best solution is to leave the memory settings in the BIOS "auto" and to correct the memory multiplier only if necessary.
How to overclock?
You enter the BIOS, set the bus clock or processor multiplier (or both), processor operating voltage as needed, set the memory to optimal mode, save the BIOS settings and that's it. It seems simple, but in fact, it is, at least when it comes to basic "beginner" overclocking.
The biggest obstacle for most is the fear that something will go wrong, as well as ignorance of basic BIOS functions. The best way to inform the BIOS is to read the instructions that come with each board in detail.
By carefully studying the relevant literature, one can see which options the BIOS offers on a particular motherboard model, and there is always a brief description of what each function in the BIOS does.
With this new knowledge, you will easily come to some of the most important options that need to be changed in the overclocking procedure. The optional thing is to "update" the BIOS with a newer version if one exists.
You can easily find any newer versions of the BIOS on the motherboard manufacturer's website, along with instructions on how to do it. Newer BIOS versions can often improve the overclocking capabilities of the board, so it is advisable to do this operation.
Of course, there are cases when the newer BIOS works worse, so then the old one must be returned. However, lately, this case is not very common.
The basic theoretical details generally apply to all platforms, but, logically, each of them has some specifics that we will now go through.
Intel LGA775 and LGA1366
Although it can be said that this is a very old platform, the fact is that the latest generations of motherboards and chipsets are still relevant to a solid number of users. To remind you, these are processors from the Core2Duo and Core2Quad generation, which are still quite fast enough for today's average needs.
For example, not so long ago, the Q6600 was in high demand in the "used" market, for the simple reason that its performance and overclocking capabilities are quite solid, and the price compared to "quadcore" alternatives is also tempting. The LGA775 platform can be said to be, at least when it comes to Intel, the last "old school" platform on which the long-known overclocking rules apply.
So, there is FSB, and almost complete overclocking is performed through it. With this generation of processors, the multiplier can only go down, which is a thing that can be used to advantage.
Since we have an FSB and a memory controller in the chipset (northbridge) here, the maximum performance is achieved through the highest possible bus clock together with the maximum possible memory speed.
As in some processor models the limit can be reached with a smaller FSB (depending on the factory multiplier), it is desirable to find the most favorable relationship between the FSB and the memory for the final performance to be maximum.
With the latest models of LGA775 processors, it can be counted at 3.4 GHz as realistically achievable without any special settings, of course with a good cooler. Take for example the E6600 which uses a 266 MHz FSB with a multiplier of 9, which gives a final clock of 2.4 GHz.
If we raise the FSB to 380 MHz, we will get a clock of 3.42 GHz, which is exactly within the recommended range. However, we are left with the question of memory, because that should be brought to the optimal value.
As the memory provides the best performance in a 1: 1 ratio (multiplier 1), you should see if it is possible (depending on the memory you have) to use this. If, for example, we have DDR2 1066, we should go to the FSB of 550 MHz, but then the multiplier 9 will give 4950 MHz, which of course will not work at all.
Then we lower the multiplier to 6 and get a clock of 3.3 GHz, so a little less than what we would like. Here you need to decide whether to go for a 0.5 higher multiplier or find some other combination that will be as close as possible to the target clock. Digitron and paper help the best here, so everything can be calculated nicely before you transfer the obtained values to the BIOS.
From personal experience, the FSB limit on better motherboards in the last couple of generations of Intel chipsets is at least 500 MHz, so there is a lot of room for computing. The P35 and X38 go around 500, while the P45 and X48 can provide FSB around 600 MHz. Even older chipsets, type 965P can transfer 500 MHz on the best board models, so definitely the LGA775 platform was very grateful for overclocking.
The best way to determine the FSB limit of a particular board is to bring the processor and memory multiplier below the declared level, and only increase the FSB. In the beginning, we can take bigger steps, and later we can slightly reduce them as we approach larger values.
Armed with apprehension, they will reach the FSB limit that a certain chipset and board can withstand. After that, you can insert the obtained data into the equation and very easily calculate the optimal ratio of the working clocks of the processor, board, and memory.
As for the operating voltage of the processor, for the Core2Duo and Core2Quad generation of processors (Allendale 65nm, Conroe 65nm, Wolfdale 45nm, Kentsfield 65nm, Yorkfield 45nm, etc.) the range was from 1.35 V to 1.5 V, and when correcting the voltage should not go over 1.7 to 1.8 V.
The new era began with processors based on the Nehalem architecture, which brought with it a new platform and slightly changed the rules of the game when it comes to overclocking. The classic FSB no longer exists, but the term BCLK (basic clock) is used to determine the final processor speed.
So here the BLCK is at 133 MHz, which results in quite high CPU multipliers. In principle, this is a suitable situation for overclockers, because the BLCK does not have to be raised too high to reach the desired operating clock.
With the LGA1366 platform, memory adjustment is somewhat freer, as it is not now that memory can be adjusted through a wider range of memory multipliers according to memory capabilities.
Of course, by raising the BLCK, the clock speed also increases the memory clock, but through the already mentioned multipliers, it is very easy to regulate if necessary. What is very important when it comes to memory is the operating voltage.
Since the memory controller is located directly in the processor, Intel has changed the specification when the memory voltage is in question, so the maximum recommended voltage is 1.65 V.
As the older top DDR3 modules used a voltage of up to 2.0 V, it is clear that such a high voltage in LGA1366 processor is by no means recommended. If you have this type of memory, you must be satisfied with the clock speed that the memory in question can achieve with a voltage of 1.65 V.
For example, we can take the Core i7 920 as the most popular LGA1366 processor, the multiplier is 20, it has a voltage of 1,375 V, and the nominal memory speed is 1066 MHz.
The maximum optimal speed for this generation of processors is around 3.6 GHz, which in the case of Core i7 920 processors means an increase in the BCLK clock to 180 MHz. This is no problem for the vast majority of motherboards, so overclocking on this platform can be considered a fairly easy operation.
Finally, it remains to adjust the memory multiplier and possibly adjust the operating voltage of the processor. For these processors, it is not recommended to go over 1.45 V, so here you should try with minimal corrections of operating voltage.
With good cooling, it is possible to achieve the specified speed without raising the processor voltage, which will help the CPU heat up less during operation.
In general, the temperature under the maximum load should be around 60 degrees, and anything over that is not recommended.
Intel LGA1156 and LGA1155
A year after the LGA1366, the LGA1156 platform also enters the scene. The main difference is working with memory in dual channel mode, while other things are very similar.
The same is true when it comes to overclocking, so all LGA1156 processors use the same system as on the LGA1366 platform. A detail to pay attention to (also with the LGA1366) is Intel Turbo Boost technology because in certain situations and on some boards it can create certain problems when overclocking.
Therefore, the simplest option is to turn off Turbo Boost if you do not need a larger multiplier (which opens when this option is turned on). The P55 chipset and the boards based on it offer the possibility of stable operation at a BCLK clock of around 200 MHz, which should be enough for the moderate overclocking ranges we are talking about here.
As far as memory is concerned, the same thing applies here when it comes to the voltage limit of 1.65 V, so it is not recommended to cross this limit here either.
Again, the situation is the same in terms of CPU operating voltage. For speeds around 3.6 GHz, no voltage correction should be necessary, and if it happens to be necessary, values of 1.45 should not be exceeded.
Here, too, it is not necessary to try too hard to achieve the highest possible BCLK clock, because it means almost nothing when it comes to the final performance. The optimal solution is to go for the lowest possible BCLK clock, along with the largest multiplier possible with your processor.
If you still want to examine the limits of the board and processor more precisely in terms of the BLCK clock, simply lower the CPU and memory multiplier and then slightly raise the BCLK clock. As soon as the first time the operating system fails to boot, you have reached the BCLK limit. However, with most high-quality motherboards, there should be no problem reaching the limit of the processor itself, before the BCLK clock becomes a problem.
Intel LGA1155 - a drastic change in the rules of the game
With the introduction of this platform, Intel has made a drastic change in the rules when it comes to overclocking. By placing the PCI Express controller in the processor core, the need for any northbridge functions disappeared.
As a result, the significant role that the chipset played in overclocking disappeared. Besides, Intel took the opportunity to eliminate overclocking by changing the BCLK clock.
Thus, on P67 / Z68 boards, BCLK is nailed at 100 MHz, and rarely does any board manage to enable stable operation at more than 105 MHz, with a maximum of 107-108 MHz in the best models.
In practice, this means that the classic overclocking is no more because you can't buy a cheap processor and get a drastically higher speed out of it.
With BCLK corrections possible, the maximum that can be counted on is an increase in processor clock speed to 100 to 200 MHz depending on the multiplier and BCLK clock.
The "Turbo" function can help here because some boards allow the maximum turbo multiplier to be used as the base, so in the end, you get a little over the nominal clock speed of the processor.
Logically, this overclocking didn't require anything other than this change to the BLCK clock, you don't even need a better cooler or anything else.
Thus, a real overclocking on the LGA1155 platform can be done with Sandy Bridge processors that have an unlocked multiplier, and these are models with the "K" suffix in their name (Core i7 2600K, Core i7 2500K).
When using one of these "K" processors, overclocking on the LGA1155 platform becomes a very simple thing.
Take a look at the example above - the multiplier is unlocked here so you just need to set it to the desired value. 4 GHz is a very easily achievable clock for these processors, and some optimum is a speed of about 4.5 GHz.
This can be achieved without correcting the operating voltage of the processor, while with more voltage you can reach the magic 5 GHz. However, it is recommended to stick to a voltage of up to 1.4 V, because even these processors heat up fiercely at such a high operating clock.
Therefore, it is very important to constantly monitor the temperature during testing, to more easily determine which clock can be maintained stably.
AMD AM2 + and AM3
Overclocking on the current variants of the AMD platform is not particularly different, because the basic principles are the same. The Phenom II generation of processors is the most current, in all its variants, from X2 and X3 to Athlon II derivatives that do not have an L3 cache.
Since all of these processors are based on the same Phenom II core, the overclocking ranges are quite similar. It can count on speeds from 3.4 to 3.5 GHz as the maximum achievable stable values, with adequate motherboard and cooling.
Anything beyond that is possible, but only with stronger Phenom II models that have slightly greater overclocking capabilities.
This primarily refers to the "Black" models, which, in addition to greater cooperation in terms of overclocking, also have an unlocked multiplier. Thus, in this case, overclocking is reduced to setting the appropriate multiplier and eventual correction of the operating voltage, while all other settings can be left at the factory values.
If you do not have a "Black Edition" processor, then the system bus setting (reference clock) comes into play. In the BIOS, this will usually be displayed through two options: HTT and NorthBridge.
On most boards, the "auto" option will do the job for you and automatically lower the HTT multiplier if the reference clock is too high. Besides, each chipset and board has its maximum when it comes to the reference (northbridge) clock.
Better board models, depending on the chipset, can offer work in the range of 300 to 400 MHz (somewhere a little more than that), while cheaper board models will struggle to work stably at speeds approaching 300 MHz.
Of course, the exceptions confirm the rule, so you can find some cheaper models that can reach 400 MHz, but these frames are what you can usually expect from AM2 + and AM3 motherboards. Since this clock is also related to the memory speed, the memory multiplier setting is always available.
The final memory clock must be within what your memory can handle, but there is a little mitigating circumstance. AMD processors have long had a memory controller in the core, but they are much less sensitive when it comes to the operating voltage of the memory, so you can experiment with the memory a little more if necessary.
In principle, it is not necessary to chase the combination of the highest possible bus reference clock, because it means very little for the final performance, but you should focus only on reaching the maximum processor speed with a certain frequency, together with the optimal memory clock.
Depending on the available CPU multiplier, the maximum reference clock of the bus can become a limiting factor, so testing this limit is also a thing to do first when overclocking the AMD processor.
Simply lower the CPU by at least a few integer values, the memory multiplier goes to the lowest value, and then slightly raise the reference (northbridge) clock speed. As soon as the operating system fails to boot for the first time, you have reached the limit of your motherboard.
After that, you only need to insert the maximum available CPU multiplier into the equation and adjust the memory multiplier so that the memory works at its optimal capabilities. Any corrections to the operating voltage can go up to a maximum of 1.5 V, while anything beyond that is not recommended.
When it comes to temperature, even with AMD processors, around 60 degrees is indeed the maximum for a processor under full load.
For the end...
Lastly, keep in mind: testing and tools
Before embarking on an overclocking action, it is useful to do some basic performance tests, so that after the action is completed, you can compare the obtained performance with what you started with. FutureMark tests like 3DMark and PCMark are free and popular, so you can start with them.
You can use some pure CPU synthetic tests that we also use when testing, and you should always insert some test with the applications that you use the most. In the end, there are games, and the ultimate FPS is always a good rapper.
The most important thing when overclocking, in addition to knowing the basic things we have already mentioned, is patience. Slightly raise the CPU and system bus clock speeds, and raise the system after each change.
You can use Prime95 for the stability test, and it is very popular software that will maximally load each CPU core, and in that way, you can check the stability of the system with a fairly high dose of security.
Let Prime95 run for at least ten minutes, if all goes well, you can continue further by raising the operating clock. We remind you once again, while Prime95 is doing its job, monitor the processor temperature.
The easiest way to do this is with CoreTemp or RealTemp programs, and you can also use the system monitoring software that came with the motherboard.
If the processor reaches a temperature above 60 degrees under this load, it can be calculated that you have reached the speed that can be achieved. After setting the limit, the stability of the system should be checked once again, first through longer work with Prime95, and then with all available test software.
If everything goes as it should, there is no BSOD (blue screen of death), cracking of applications, and similar things, you can say that the job of overclocking is finished, so now you can enjoy better performance.
When it comes to stability, the advice is to lower the obtained operating clock, which you have determined to be stable, a little more to have an additional margin of safety in stability.
Simply put, the temperature in the case also varies under the influence of ambient temperature, which can be reflected in a couple of degrees higher heating of the processor, which in turn can produce instability if the CPU is at its limit in given conditions.