Introduction

From the early bars of modern computing, overclocking has always been a fairly controversial topic. Is it worth doing? Is something going to break? What do we pay in exchange for that additional performance?

Whatever we do, there is always a small risk of forcing the components above the manufacturer's frequencies. So in any guide (and this one is no exception), we will see scary warnings warning of potential problems, and that all of the following are under the responsibility of the end user.

Much of the blame for this "bad reputation" of overclocking is due, simply, to a badly treated processor, to touching BIOS values ​​without common sense and without consulting or googling, to insufficient cooling, to an outrageous lack of cleaning and maintenance. and, in general, a combination of all of the above. A processor with a well-made and well-kept overclock has many more ballots to last for many years than a processor that has spent its entire life without touching frequencies, but with a tiny heatsink heating up day and day too.

Is there more wear overclocking? The answer is quick: Generally little, but yes. More consumption implies more electronic migration, and hopelessly more heat. Luckily, we have many dozen years before a modern processor says enough, and in general, it can be said that if a processor dies with a well done overclock, it would die exactly the same without it, surely a few weeks later that yes..

Another general tip, run away from the automatic overclocking options that boards from all manufacturers include. Why? Because they always put much more voltage than we would put (that is, unnecessary consumption, wear and heat), and worst of all, they do it without control, we can literally be frying our processor and not realize until it is late.

There are people who like to overclock with manufacturers' utilities with the operating system started. It is usually comfortable, and it is the fastest to test, personally I like to change the values ​​directly in the BIOS, first, because it is the safest to clearly see what we are doing, second, because we can format, change the OS, whatever we want that the overclocking will still be there, as stable as the first day.

To finish, a very common mistake (and you will see this question repeated daily in many forums and communities) is thinking: I have X processor. How much voltage do i need to do X Ghz? Answer: IT DEPENDS. Each processor is a world. There are very good batches that go up many mhz with the stock voltage, there are very bad batches that can hardly be rushed at all, and unfortunately, only luck influences here, and little can be done to solve this. As you will already suppose, the processors that break the world records are chosen among great games because they are the best ones. At the end of the day, if all the processors made more frequency than the manufacturer advertises with the same voltage, they would label them with that frequency, and obviously, they would sell them to us more expensive.

The most important thing before we start: don't be scared of the crashes and blue screens that we will have (because we will have them). Even the biggest instability due to overclocking is solved in such a simple way as loading the default values ​​of our BIOS.

Previous concepts

BCLK: The frequency of the main bus encompasses the old socket 775 FSB, but adding many other buses in the same clock generator, for example, the pciexpress. It is set to 100mhz, and unlike in previous generations, it is recommended not to change it, it barely holds a few more mhz compared to its stock frequency, and you can even read articles about low-end plates failing to raise this value. In the case of socket 2011, there is the possibility of applying a multiplier (x1.00, x1.25, x1.66) that affects only the frequency of the processor and memory. It may be interesting, but note that not all processors support these multipliers (some, however much you raise their voltage, they do not), and in general you can achieve exactly the same effect by increasing the CPU or RAM multiplier in your case.

Multiplier: It is the number of cycles of the processor for each cycle of the BCLK, the frequency of our processor is calculated by multiplying the value of the BCLK by the multiplier. In general it is the only value that we will change to achieve the desired frequency, normally changing the maximum turbo boost multiplier for all the cores (since it usually gives a better result than raising the base frequency, at the same performance, hopefully we can lower the voltage).

Here's the need for an unlocked processor. In this socket, all the processors (i7 4960X, i7 4930K, i7 4820K, i7 3960X, i7 3930K) comply with this, except for the i7 3820, in which we will have to take advantage of the BCLK multiplier mentioned above.

CPU / Vcore Voltage: The voltage that will reach our CPU. It is a "necessary evil", since it is what drastically increases consumption and heat, but raising it is what makes the system stable again after raising the frequencies. We must be especially careful at this point, since an excessive voltage is one of the few things that can cause permanent damage to our processor. There is no absolute rule for voltage, since it depends on our refrigeration, the safe margin will be greater or less, but it is recommended to stay below 1.4V in air, adjustable to 1.45V in liquid (custom loop, or sealed kits of very high range, for a sealed liquid it is more advisable to take the limits for air, which is its performance). For a first overclock, we will try to stay below 1.35V. If our temperatures are good, we will continue. The table of safe voltages, according to intel, is as follows:

The further away we are from these values, in general, the better. For example, memory kits running at 1.85V are usually extremely tight, rather loose chips. On socket 1155/1150, some limits are stricter, for example it is recommended that the ram not exceed 1.65V.

For a mild / moderate overclocking, in general we will not need to change any secondary voltage of our board. It is usually enough to know that they are there if we want to tighten something to the maximum, or we do not achieve stability at much lower frequencies than expected. The names of the voltages that regulate the same things are slightly different for each manufacturer, although easily identifiable.

Recommended programs

The adjustments will be made directly in the BIOS, that is, we will not need any overclock program as such. What we will need is to monitor the voltage and frequency of our CPU, temperatures, and lastly, stability. These are the programs that I only use, the Prime95 is as valid as the IntelBurnTest, or the Coretemp versus the HWMonitor, but these are the ones that I usually use and the ones that have given me the best result. All free, and all more than fulfill their function.

The offset is a value that is added (or in its case subtracted) to the VID of the processor at all times, allowing us to increase the voltage when necessary, but without losing the drop to save energy when the computer is turned on with little work.

1. Everything is done. We save the BIOS values, and restart. If the PC crashes before reaching windows, it is not necessary to try more, the overclock is unstable, we add the offset approximately 0.02V (to feel about it) and test again. If the PC fails to pass POST, the BIOS should load the default values ​​and give an error message after several boot attempts. We repeat the steps with a little more voltage. When we get to the SO, we continue with the next step, we check the stability of the equipment. We want something quick to be able to change values ​​in the BIOS as soon as possible (to increase frequency more if it is stable, or to increase the voltage if it is not). Normally, with about 15 passes in High mode (2048mb) of the intelburntest it is enough to get an idea (we do not know for sure if it is stable with "only" this, but we do know that it is rare that it was not). If you have ram in quantity, fewer passes with more ram usually give a better result to detect instabilities. For the final test, it is recommended to leave it for several hours, with as much RAM as possible (we put 100 passes, for example, and wait until we get tired).While we pass the test, we check the temperatures with HWMonitor. If the CPU temperature goes above 75º, you are already at the limit of what your cooling system allows, so you shouldn't keep going up. If it goes through a lot of 80ºC, we are at the top of what our processor can give, and we should not keep going up (what's more, I would recommend loosening the overclock a bit to normalize temperatures, it is better to have 100mhz than a processor with 2 less years of life). We always talk about encapsulation temperatures (the one that comes out as a dry CPU), if the cores go somewhat hotter, it doesn't matter. The ivy-e are hot, and you can tighten the limits a bit, but personally, since intel, quite conservatively, specifies a maximum Tcase of 71º, it would try not to go many degrees from there.

   If anything fails, the computer crashes, a controller that has never failed fails, we see a "XXX stopped working" screen, ultimately anything abnormal, turn up the CPU voltage 0.02V and go back to step two. Always without going over those 1.35-1.4V

   If the PC is stable, go back to step one and raise the multiplier one point, either due to high temperatures (most likely, if you have followed the guide strictly and do not have brutal cooling), or due to voltages at the limit that we have commented (1.4V) there will come a time when we have reached the limit of our processor. At this time it is best to return to the last stable value, and lower the voltage as much as possible, little by little, point by point, and test stability every time. As point 2 says, for the last test it is highly recommended to leave it, at least 4-8 hours (with some rest if necessary, so that the box cools down a bit) with all the available RAM to make sure.

WE RECOMMEND YOU Intel will launch Haswell-E at the end of the year

The screen that every user overclocking will see during this heavy stability testing process, saving personal preferences (there are those who prefer prime95 over IntelBurnTest, others the great OCCT that brings a little of everything…), should be similar to this (which is mine at the time of writing these lines):

About Loadline Calibration (LLC)

Although in general the normal value that the plates bring meets what we want to do, it is interesting to know that we have this option. Its role is simply to compensate for the processor's natural voltage drop when fully loaded. It is a good complement to offset overclocking, and in many manufacturers there are many levels to adjust to our liking.

In the case of MSI it is a very complete option, which compensates, to some extent, for the absence of offset options. There are people who use this option to overcompensate the vdrop on load and have an overclock with very low voltages at rest, personally it does not seem to me a recommended practice, first because the processor eats very ugly voltage spikes in the idle-> load step, second because if we go down we can have instabilities in that same transition and go crazy until we find the problem.

It is an option that is sometimes somewhat hidden, for example in the Rampage it is located in the advanced settings of the phases, in the section "Power Control DIGI +"

BSODs error codes (blue screenshots) and probable causes

Translated list from overclock.net

0x101 = Increase Vcore

0x124 = Increase / decrease QPI / VTT first, if not better, increase Vcore (normally the first case is in the 1st generation i7, the second in the Sandy)

0x0A = RAM / IMC unstable, increase QPI. If it doesn't improve, increase Vcore

0x1A = Memory management error. Many times it is a faulty module. Try to increase the RAM voltage a little, test the RAM with Memtest

0x1E = Increase Vcore

0x3B = Increase Vcore

0x3D = Increase Vcore

0xD1 = QPI / VTT, increase / decrease if necessary. It can also be unstable RAM, raise the RAM voltage a little

0x9C = QPI / VTT most of the time, but lack of Vcore can also be a cause

0x50 = Unstable RAM Frequency / Latencies or uncore multiplier, increase RAM voltage or adjust QPI / VTT.

0x109 = Too little or too much Voltage in RAM

0x116 = Low IOH (NB) rating, or GPU issue (common with heavily overclocked GPUs or massive multigpu setups)

0x7E = Corrupt operating system file, probably overclocked. Run sfc / scannow and chkdsk / r

Any errors that do not appear in the list (hangs, reboots without screenshot, frozen IBT…) are usually due to lack of Vcore.

Troubleshooting and additional information

Here we will list various “worst case” assumptions, and how to fix them.

It may happen that directly, the PC is left with the black screen, the fans are running but it does not even try to start. This usually happens almost always when we try to overclock the ram without relaxing latencies (the modules usually have very little margin, and are errors in which the BIOS has trouble recovering), or because they were in too much of a hurry to upload the multiplier instead of going little by little. Don't panic, all these problems are solved by loading the BIOS default values.

• First of all, we unplug the source, press the power button on the computer (to empty the capacitors). We wait a minute and try again. Many boards are "ready" and know how to load the default values ​​after a bad olverclock. If the previous step does not work, we will reset the BIOS to the default values. Many high-end boards include a button on the back for this (as each model is different, we recommend checking the manual). On more mundane boards it is usually a simple jumper that is close to the stack and has the abbreviation "clear RTC" or "clear CMOS" written on it. It is not necessary that the PC is disconnected from the power, but it does not hurt:

  

  If the previous step also fails, we do the same again, but this time also, we remove the button cell from the board and leave the jumper jumper in the erase position. We also remove the RAM modules, and leave the PC without power and without a battery for a few hours. Posts to insure, it is best to leave it for an entire night. Once done, we put back the battery, the ram, plug in and test. If all went well, the PC should work at this point.

  

Failures when restoring from sleep / hibernation: Check that the PLL Overvoltage is deactivated (and the voltage hovering around 1.8V if our board reports it, sometimes in Auto some boards decide to upload it unnecessarily).