Latest edition of 3 step guides – September 7, 2010
Updated – February 26, 2010
Originally Published – January 25, 2010
So many users are searching around the net these days looking for advice on how to overclock their new systems but don’t know where to start. To help everyone out, I decided a how-to guide was in order. Searching around forums can be confusing and intimidating. There are so many people willing to give advice, but who can you trust? It’s hard to know, and I’ve seen many users sent on wild goose chases because they are following advice that doesn’t solve or even address their specific problem. I’ve also seen too much trial and error overclocking. What I will attempt to do is create a very simple three step guide to “one-size-fits-all” overclocking.
If you want to continue searching out other opinions, please consider each suggestion with caution. Some will undoubtedly be great, some will not. There is one exception I would make for OC Forums- their “senior members”. These users are the ones with blue user names and blue stars. They have been designated seniors based on their knowledge, experience, credentials, and contributions to the OCF community. While not immune to error, these users generally know their stuff and can be trusted for good advice.
So, why should you listen to me? Well I’m not the end-all-be-all of overclocking, and I still learn new things every day. There are many people around OCF whom I highly regard for their knowledge and experience. Having said that, I do have a bit experience overclocking computers successfully. I’m a member of Team IRONMODS, and if you’d like to know more about the team and our accomplishments, please visit our website. I’m also very active with the OCF benching team where I regularly participate in team events and online competitions with The Raptor Pit.
My goal here is for this overclocking guide to be useful for anyone with a newer Intel based system, i3, i5, i7, LGA1156 or LGA1366. With the same basic principles applying to all of them, the basic process doesn’t change whether you are planning to use your system as an everyday system, gaming or if you want to push the limits for a single benchmark.
This guide is also independent of your cooling system. Whether you are using the stock Intel cooler or if you’re pushing to the extreme with liquid nitrogen, the basic steps remain the same. One thing that is far too common is errors in mounting your cooling system, specifically the application of the thermal interface material (TIM). If you don’t have much experience mounting a cooling apparatus, please refer to this excellent guide from Arctic Silver.
Determining methods for finding a stable overclock are highly controversial, and my suggestion is that we agree to disagree. Everyone has their individual definition of a stable system, but when I refer to “stable” in this guide, I am referring to the stability of your selected “stability test.” So for a power user or gamer who wants a reliable system that won’t ever crash due to an overclock pushed too far, you’d need to test with a program that will load all of the cores and threads applicable to your CPU, OCCT and Prime95 are two popular choices. For a benching team member looking to squeeze every last MHz out of their chip for a 7 second SuperPI 1M run on liquid nitrogen cooling, SuperPI 1M would be the ideal test. In my examples below, my verbiage will obviously be geared more towards those running tests like Prime95. Super PI 1M only takes a few seconds to complete, so when I say “run your stability test for five minutes” obviously you will have to tailor that instruction for your individual situation.
So with that in mind, we will attempt to isolate each portion of the system and overclock one piece at a time. This may seem time consuming at first glance, but rest assured this can potentially save you hours of troubleshooting and frustration. So go slow, and follow each step very carefully.
I also need to insert a disclaimer in here somewhere: I am not responsible for any bad things that happen to you or your computer as a result of you listening to my advice, nor is overclockers.com. My goal is for this guide to be a safe overclocking guideline, but the burden for damaged hardware lies on the user performing the overclock! Overclocking can damage hardware, the new 32nm CPUs seem to be particularly fragile. Overclocking will also void your warranties.
This guide is written for you if you can get around in your system’s BIOS (basic input/output system). I cannot write this guide to cover the variety of BIOS represented. If you are unfamiliar with your BIOS, search for more information regarding your specific motherboard. Also, do not be afraid to get into your BIOS and have a look around, if you are ever concerned that you may have changed a setting erroneously, you can always load defaults, and start over.
The CPU micro-architecture has taken a huge leap from the 65nm Core to the new generation 45 and 32nm technology, it has brought many changes not only to the CPU’s but also to the chipset and motherboard design and functioning. This is what make overclocking the i3/i5/i7 CPU’s so much different to their predecessor LGA 775 CPU’s.
The naming convention can be a bit confusing so let us look at the various CPU and their names:
First we have the Nehalem family which are all 45 nm CPU’s that included i7 1366 Bloomfield (i7-920 i7-975) and Lynnfield socket 1156 i5/i7 (i7-750 to i7-860). These are all quad cores with HT except for the i5 which has no HT.
Then the next family is Westmere which is essentially die shrink 32 nm CPU version of Nehalem and again you have the Clarkdale (socket 1156) in flavors of i5 and i3, both dual core processors with HT.
Gulftown which is the hex-core CPU’s and not available yet are also part of the Westmere family and features 6 physical cores with HT and will be socket 1366 only.
The above are all desktop chips, then to you get the Arrandale and Clarksfield CPU’s which are mobile processors and the Gainstown which is the server equivalent of Bloomfield.
So, to summarize, we have socket LGA 1366 and LGA1156 which are essentially the board platforms that carry certain 45 and 32 nm CPU variants. Both platforms are DDR3 where the 1156 is dual channel only.
Intel Core i3, i5, and i7 CPU’s
The following are the different CPU’s available today on the market, except for the Gulftown hex cores which will be available late 3rd quarter of 2009.
Core i3 (Clarkdale – i3 530, 540)
The least expensive and least powerful choice are the Core i3 chips. These are currently limited to two physical cores with both the i3 530/540 models supporting Intel® Hyper-Threading Technology. At this time, all i3 CPUs have two pieces of silicon (or dies) in the CPU. One contains the actual processing cores and all of the L1, L2, and L3 caches. The second die contains a GPU (graphic processing unit) which is capable of outputting video without the use of a dedicated video card when used with an H55 or H57 based motherboard. This second die also contains the PCIe controller and the dual channel IMC (Integrated Memory Controller). The first die is manufactured on a 32nm process, the second die is manufactured on a 45nm process. The two die are linked with the Intel QPI (Quick Path Interface). All i3 CPUs work only on LGA1156 based motherboards.
This is all very important to understand when attempting to overclock one of these CPUs, because you are actually in a constant juggle with overclocking both dies. Why? Because both dies derive their speed from the bclock (base clock) frequency. Please understand that these new 32nm CPUs are more sensitive to high voltages, and have been known to fail even when core temperatures are well within “safe” limits. Another major issue many users have been running into with overclocking these Clarkdale based CPUs is the IMC. It is not nearly as powerful as the Lynnfield based CPUs, and many users are forced to run at slower memory speeds and timings to get stable.
Core i5 (Clarkdale – i5 650, 660, 661, 670)
Dual core i5 CPUs (Clarkdale) are identical to the i3 CPUs, but also include Intel® Turbo Boost Technology. All i5 CPUs work only in LGA1156 based motherboards.
Core i5 (Lynnfield – 750)
Quad core i5 CPUs (Lynnfield) are identical to the low end i7 CPUs, the only exceptions being their lack of Intel® Hyper-Threading Technology, Intel® Virtualization Technology for Directed I/O (Intel® VT-d) and Intel® Trusted Execution Technology (Intel® TXT). All i5 CPUs work only in LGA1156 based motherboards.
Core i7 (Lynnfield – i7 860, 870)
The low end Core i7 (Lynnfield) are quad core CPUs made for LGA1156 based motherboards. These CPUs are manufactured on the 45nm process. They have dual channel IMCs and PCIe controllers built into the CPU die. All i7 CPUs include Intel® Hyper-Threading Technology and Intel® Turbo Boost Technology. Lynnfield based CPUs are unique in that they do not have a QPI. Because the PCIe and memory controllers are both integrated on the CPU die, there is no need for the QPI. They are known to have incredible IMCs and are capable of sustaining extreme memory bandwidth.
Core i7 (Bloomfield – i7 920, 940, 960, 965/975)
The mid range Core i7 (Nehalem) are quad core CPUs made for LGA1366 based motherboards. These CPUs are manufactured on the 45nm process. They have triple-channel IMCs built into the CPU die. The PCIe controller is not part of the CPU, but is built into the chipset (on the motherboard), the CPU and chipset are linked with a QPI. All i7 CPUs include Intel® Hyper-Threading Technology and Intel® Turbo Boost Technology.
Core i7 (Gulftown – i7 980x)
The high end Core i7 (Gulftown) are hex (six) core CPUs made for LGA1366 based motherboards. These CPUs are manufactured on the 32nm process. They have triple channel IMCs built into the CPU die. The PCIe controller is not part of the CPU, but is built into the chipset (on the motherboard), the CPU and chipset are linked with a QPI. All i7 CPUs include Intel® Hyper-Threading Technology and Intel® Turbo Boost Technology. Please understand that these new 32nm CPUs are more sensitive to high voltages, and have been known to fail even when core temperatures are well within “safe” limits.
Before we do anything, please start by going into your system’s BIOS, and load defaults, then save and exit. Your machine will restart with default BIOS values. Enter the BIOS again and disable all power saving features. These include, but are not limited to; EIST, C1E, and all other C-states. All other settings you can leave on auto for now. Although I would also recommend turning off any start-up slash screens, so that you can view your system’s post behavior. Also, feel free to disable any “integrated peripherals” that will not be used (i.e. NICs, extra PATA/SATA controllers, legacy devices, etc).
If your motherboard fails to post after changing certain settings, you will have to locate and reset the CMOS. Resetting the CMOS restores the BIOS to its factory settings and is a “hard” reset of these settings. Become familiar with where the CMOS jumper or button is, you may need to use it a lot. Most of the time, a jumper is located near the battery on the motherboard. Most newer motherboards have a button in the same location and many may also have a button on the rear input/output panel for easy access when your motherboard is installed in a case. If you need to reset the CMOS, please power off the system by flipping the switch on your power supply, or unplugging it completely for 10 sec, then activate the CMOS reset jumper/button for 10 sec.
Understanding CPU frequency
Before we go into how we overclock these CPU’s let us look at what determines how fast your CPU will run. The following simple equation determines the clock speed of the CPU’s cores:
CPU Frequency = Base Clocks x Multiplier.
This is a biggest change from the old LGA 775 where FSB and multiplier determined the CPU speed. The base clock is similar to the FSB but also has some key differences. The base clock, also commonly spelled bclocks or bclk in forms, is the foundation around all the other frequencies discussed below.
The CPU speed of the new generation is not the only factor that determines how fast your PC will run, we have a few more definitions such as:
QPI Frequency – QPI or Quick Path interconnect is the Intel communication path upgrade from the older chipset and front side bus (FSB) communication path, so instead of the CPU communicating with the memory via the LGA 775 Northbridge, there is now a direct link (QPI) that increases efficiency. QPI speeds are a function of base clocks, so as you increase your base clock your QPI speed will also increase, yielding an increase in not only communications speeds but also bandwidth, which leads to an increase in PC performance.
Uncore frequency – This sets the frequency of the on-die memory controller and the L3 cache. Like CPU clock speed, dram speed, and QPI frequency, uncore is a multiple of Bclk. Uncore can be set independently of those other frequencies, subject to certain stability limitations. The uncore must be at least 2:1 of the DRAM speed otherwise you will not get a stable overclock, in fact your PC will not even boot if the ratio is not honored. Increasing the uncore:dram ratio above 2:1 yields significant performance gains. However, when the ratio reaches 3:1 it is not possible to maintain full stability.
Multiplier and Turbo – As mentioned above, the multiplier is the second factor in how CPU core speed is determined. Now, not all CPU’s have the same multiplier, it is dependent on where the CPU is positioned in the price/performance curve of Intel’s range of CPUs. Most of these come with a Turbo multiplier which is available if you enable the Speedstep option under the CPU settings. Care should be taken when using the turbo as you may not be able to see the resultant frequency in the BIOS. For instance, if your default multiplier of your CPU is 20 (i7-920) and you set your baseclock to 200 and you boot up with turbo enabled, you will leave the bios at 20 x 200 = 4 GHz, as soon as you enter your Operating system your turbo kicks in so you end up with 21 x 200 = 4.2 GHz. Now if you also have C-State enabled, one CPU core will actually have access to a 22 multiplier which enable that core to run at 22 x 200 = 4.4 GHz. You set your voltages expecting to run at 4 GHz and you cannot understand why you get a BSOD when you enter Windows, well, that is the reason, so take care when using turbo and C-State and adjust voltage to accommodate for the higher multipliers.
Important Voltages when Overclocking
There a few important voltages which you will need to manipulate while overclocking, below are the main ones. Every motherboard Bios differ but all of them have the voltages as set out below.
- V-Core – Directly related to the CPU frequency. As you increase the CPU frequency you would need incrementally increase the v-core as well.
- QPI voltage/CPU Vtt – Increase in this voltage is necessary from the default as you increase your RAM speed, tighten the timings or increase QPI frequency. It also helps to stabilize your overclock at higher base clocks.
- VDIMM/DRAM – This is directly related to your RAM memory modules and increase will assist in stabilizing increase in Ram speeds. Care should be taken not to increase this voltage more that 0.5 volts above your Vtt as you could cause permanent damage to your CPU.
- IOH Core Voltage- This voltage aids when increasing base clocks above say 200. In most cases leaving it at auto works best.
- ICH Core Voltage- This voltage feeds the chip that regulates the communication from the peripherals to the CPU via the DMI. It is best to set this at auto.
Now that we have covered all the basics let us jump to what this article is all about…overclocking
Step 1) Maximize Bclock Frequency
Isolate the bclock from the CPU
First you need to isolate the bclock and find its stable limit with your chosen cooling. In order to isolate the bclock from the other components, the first thing you need to do is manually force a low multiplier for the CPU. For example; at stock speed, an i5 750 runs on a 133MHz bclock and a x20 multiplier which results in its stock speed of 2660MHz (133×20). Raising the bclock to 200 with the stock x20 multi would result in 4000MHz for the CPU, which you’re not quite ready for yet. If you are shooting for a 200MHz bclock, then a safe choice for now might be a x12 multi, which would result in a CPU speed of 2400MHz if you were successful in reaching your 200MHz target bclock. Doing this isolates the CPU from the bclock so you can focus on only bclock overclocking in this step. In some situations, x12 may not work, this is just an example though, so don’t be afraid to try other low multipliers if x12 doesn’t work.
Isolate the bclock from the memory
The fastest rated speed for memory on P55 with an i5 750 (for example) is DDR3-1333, which is a clock speed of 667MHz (dual data rate “DDR” doubles the bandwidth to 1333-like speed). Just like the CPU, the memory receives its clock from the bclock via a multiplier, in this case x5 (133×5=667). This is most often expressed in the BIOS as “2:10″. If you were to overclock the bclock to 200MHz as described before, your memory would be running at 1000MHz (DDR3-2000), and beyond the specs of all but the most extreme memory. To isolate the memory from the bclock, lower the memory multiplier to the lowest setting available, most likely 2:6. If you were to reach your goal of 200MHz bclcok frequency, your memory would only be at 600MHz (DDR3-1200) and well within the capability of all but the worst DDR3 on the market.
Isolate the bclock from the iGPU (Clarkdale only)
Clarkdale CPUs include an iGPU (integrated Graphics Processing Unit). If you are using an H55 or H57 based motherboard and the iGPU is enabled, please pay close attention to this section. Some early BIOS versions did not allow for iGPU clock speed adjustment, if you do not have this option in your BIOS, please update your BIOS to the most recent version. This platform is still very new and immature, so this information may only be relevant for a short time. However, at this current time, it appears as though the iGPU frequency setting in the BIOS is based on the default bclock frequency. This means that an iGPU frequency that is set at 900MHz in the BIOS, will only actually be 900MHz if the bclock frequency is set to 133MHz, if the bclock frequency was raised by 25% to 166MHz, the actual iGPU frequency would also go up by 25% or 1125MHz. This is a relatively simple concept to understand, except that YOU have to do the calculation, because the BIOS only reports the set frequency, not the actual frequency. What makes things worse at this time is that there is no software monitoring utility that is capable of reading the actual iGPU frequency.
For now, it’s only important to isolate it as a variable from our overclocking process. So, assuming your goal is 200MHz bclock frequency, which is a 50% overclock of the bclock frequency, you need to lower the set iGPU frequency to prevent its overclock during this process. If the stock iGPU clock speed is 900MHz and we were to overclock it 50%, that would yield a 1350MHz actual iGPU frequency. To bring 1350MHz back down to 900MHz we would need to reduce it by 33%. So reduce the set iGPU frequency by 33% to 600MHz, with our stock bclock frequency, the actual iGPU frequency would also be 600MHz. However, if you successfully reach your target 50% increase in bclock, your set iGPU frequency will yield an actual iGPU frequency of 900MHz, which is the iGPU’s stock speed.
For this step, there are only two voltages you should play with; VTT, and IOH. IOH is easy, if you are running a single PCIe card (graphics card), give the IOH 1.3V, if you are running more than one PCIe graphics card, give it 1.35V. VTT is the crucial voltage adjustment for achieving high bclock stability, which is also known as “CPU VTT”, “QPI/VTT”, or “QPI/DRAM”. This is the voltage that is fed to the IMC (Integrated Memory Controller), and also has a major impact in overclocking the bclock. CPU VTT is the crucial voltage adjustment for achieving high bclock stability. Stock values differ depending on platform and CPU, but as a rule of thumb LGA1366 likes a lot, P55 doesn’t need as much.
So, are you ready to start overclocking? After entering your BIOS and lowering the CPU & MEM multipliers, go to the voltages section and raise your IOH to 1.3-1.35V and your CPU VTT to +0.2V. Then restart your machine and go back into the BIOS, if your system fails to post and return to the BIOS, please re-read the last paragraph in the “prerequisites” section above, and start over. If you still cannot get past this step, post in the forums for some specific help.
After you’ve restarted your system with your manually configured voltages and returned to the BIOS, I always recommend going to the temp/voltage monitoring section and checking the CPU temp. If the temperature seems too high for your cooling, then shut the system down and double check that your cooling system is properly mounted, and making good contact. Moving on, almost all systems should be able to achieve 150MHz bclock stability with stock voltages, so go to the bclock adjustment and change it from 133MHz to 150MHz. Then save and exit and allow the system to reboot. This time, allow the system to boot fully into the operating system.
Testing for highest stable bclock frequency
Once the operating system has fully loaded, start up RealTemp. RealTemp should always be running while checking for stability of an overclocked system to ensure you do not overheat your CPU. RealTemp shows your CPU’s core temperatures real-time, as well as the distance to TJ Max, my advice is to never exceed TJ Max. Now start up CPU-Z, this will allow you to ensure that your overclocked settings have been properly applied, and that you are running at your desired speed. Check both the CPU tab for the expected CPU frequency (should be 1800MHz at this point), and check the memory tab to ensure your memory is running at the proper speed (CPU-Z will show the frequency of the memory, not the DDR3 speed, it should be 450MHz at this point). Now start up your selected test program, for example OCCT or Prime95. Run the test for just a short amount of time, five minutes should be plenty. Then reboot the system and return to the BIOS.
If the test ran without error, raise the bclock by 10MHz, reboot into your OS and run the test again. If the test failed, raise the CPU VTT voltage by a small increment, reboot into your OS and run the test again. You should be able to see where this is going, continue to raise bclock or CPU VTT voltage with a short test after each change, until you meet one of the following criteria:
- You reach your desired bclock and successfully pass your stability test.
- You reach your maximum safe CPU VTT voltage.
- Raising the CPU VTT does not allow for additional stability.
* Note – there is a phenomena known as “bclock holes,” which seem to be less common now, but may still create confusion and frustration during this process. But if you appear to have found your limit at a much lower speed than anticipated, please consider trying a step or two higher before continuing on. A bclock hole causes system instability a particular bclock values, and going past them may allow you to regain stability.
Maximum safe CPU VTT
What is the maximum safe CPU VTT voltage? Depends on a lot of things, but I feel like these are some basic conservative guidelines. If you’re running the stock Intel heatsink and fan, I would not advise more than +0.2V, if you are running a high end air cooler I would not advise more than +0.3V on LGA1156 platforms, and no more than +0.4V on LGA1366 systems. If you are running a high end custom water loop add another 0.05V to those values, and if you are using extreme forms of cooling then use whatever works best. I’ve used up to 1.70V on an i7 920, and up to 1.55V with my i5 750 with extreme cooling.
After you have met one of the criteria above, you should have a rough idea of your bclock limit, now it’s time to get a little more fine tuned. Next, instead of 10MHz bclock changes, shift to 2MHz changes. Then repeat the steps above and search for one of the three criteria again. Also, ensure you check my note about “bclock holes” above, the same concept can be applied to this fine tuning step as well.
After you have found your highest stable speed to within 2MHz accuracy, lower the bclock by 2MHz and run your test again. This time let it run for a full hour (for those of you testing with Super PI or similar, adjust for your situation). If it passes the test – Congratulations! – you have found your highest stable bclock frequency.
Step 2) Optimize Memory Frequency
The next step is to find the limit of your memory. To do this, first you need to look at the memory’s ratings. DDR3 does not typically have a lot of overclocking headroom, so it’s important to start with stock settings. In this example I will use some basic Crucial Ballistix PC3-12800 for my explanations. This memory is rated for DDR3-1600 (800MHz) 8-8-8 24 1T with 1.65V. Enter the BIOS and adjust your memory timings according to the manufactures rating, in this case 8-8-8 24 1T. Now, consider your maximum/desired bclock frequency, 200MHz for example. This memory has a stock speed of 800MHz, so a 2:8 ratio with a bclock of 200MHz would put us right at that stock speed of 800MHz. You could set it and leave it there, but let’s say your maximum/desired bclock is not 200MHz. For example, if you are actually trying to reach 210MHz. If that were the case, the resulting memory frequency would be 840MHz (DDR3-1680). So, similar to finding bclock stability above, we need to work our way up to the desired speed testing along the way. There is an exception to this section, and that is with Clarkdale base Core i3 and Core i5 CPUs. They tend to have a very weak IMC, and are often times not capable of running memory even at their stock speed. If you have a Clarkdale based CPU, you may have to sacrifice memory speed to attain a good CPU overclock.
Testing for highest stable memory frequency
Theoretically we should be able to run for at least an hour with the bclock at 200MHz and the memory multi at 2:8– why? Because we already found out that this bclock speed is 1 hour stable, and we are not overclocking the memory yet.
However, the integrated memory controller (IMC) is powered by the CPU VTT voltage. So under some circumstances, especially with the newer 32nm CPUs, you may not be stable with your memory even at stock speeds due to the overclock imposed on the IMC. This is particularly true if you are running 4 DIMMS (P55/H55/H57), 6 DIMMS (X58), or 4GB DIMMS (P55/H55/H57/X58). If this is the case keep the memory at stock speed, or even try dropping the memory clock multiplier to run at less than stock speed, and increase the CPU VTT voltage until you gain stability. The newer 32nm CPUs seem to have particularly weak IMCs, and often will not run at the higher multipliers even if your memory is perfectly capable.
For testing memory, it is important that you take a break from whatever stability test you’ve been running, and use memtest86+ instead. The easiest way is to download the .iso and burn it to disc. Then configure your BIOS to load from your optical drive before the hard disk drive. When you boot the system with the disc inserted, memtest86+ should start automatically, and immediately begin testing your memory.
But, our goal is to reach 210MHz bclock, which will result in 840MHz memory frequency. In the BIOS, set your bclock to 202MHz, and your memory multi to 2:8, save settings and exit. Allow memtest86+ to load and complete one entire loop. A single loop can vary in length, and can take quite a while if you have a large amount of memory installed. If the test ran without error, press Ctrl-Alt-Delete and enter your BIOS. Raise the bclock by 2MHz and then save and exit. If the test failed, raise the memory voltage by a smallest increment possible, and run the test again. You should be able to see where this is going. Continue to raise bclock or memory voltage until you meet one of the following criteria:
- You reach your desired bclock and successfully pass a single loop of memtest86+.
- You reach your maximum safe memory voltage.
- Raising the memory voltage does not allow for additional stability.
Maximum safe memory voltage
What is the maximum safe memory voltage??? This is determined by two things: first, NEVER INCREASE THE MEMORY VOLTAGE MORE THAN +0.5V OF THE CPU VTT VALUE, second, how much do you enjoy killing your memory? Throughout recent history, memory is probably the easiest component to damage with extra voltage. While there are exceptions, most newer DDR3 memory modules do not need very much voltage to reach their practical limits.
Once you have satisfied one of the three criteria above, drop the bclock down 2MHz from your last stable setting, and see if memtest86+ will run through 2 or 3 loops without error. If you wish to try to push your memory even further at this point, there is one more thing to try, and that is another bump in CPU VTT voltage. This will possibly boost the capabilities of the IMC and give you a little more room to overclock the memory. Otherwise – Congratulations! – you now have a relatively stable bclock frequency and memory frequency.
Step 3) Stabilize CPU Frequency
The last step in this guide is often the first step for users who run into problems and then troubleshoot for days afterward. Leaving it to the last step makes the task much simpler. You now have the following settings locked in; CPU VTT, IOH voltage, memory voltage, memory multiplier, and memory timings. That means when we are looking for our highest CPU frequency, there are only two variables we need to play with: bclock and CPU voltage.
Right now your CPU multiplier should be very low, and your bclock should be quite high. If we move the CPU multiplier up right now, we would undoubtedly become very unstable, and unlikely to post. The idea here is that if your bclock and memory are stable with the current settings, shifting the bclock down should not cause any instability. So, change you bclock to 140MHz, and switch your CPU multi up to its maximum. In our example i5 750, the normal maximum would be x20. Intel’s Turbo feature allows for extra multipliers, and some BIOS will even allow for the higher multipliers to be forced. It will not hurt to use this feature if you desire. So with my example of the i5 750, with some BIOS, I would be able to lock in a multi of x21.
This actually goes by a few different names, but they are all meant as a means to reduce or prevent v-droop. Most overclockers would advise you to enable this feature; I would only recommend it if you understand what it does. It does typically allow for measurably higher overclocking, but at the cost of violating Intel’s design specs, and putting more stress on the CPU. However, overclocking in its essence violates Intel’s design specs, so you’re not breaking any new ground with this feature. I do not enable load-line calibration on my daily/gaming system. But I always use it when I am trying to fine the absolute limit. For more insight on the matter, refer to this excellent explanation at anandtech.com.
That brings us to the first thing that most users want to play with after powering up their new system for the first time: CPU voltage, aka “vcore”. As you can see, this is actually one of the last things you should be changing. I would recommend starting at a nice and easy 1.3V. Surprisingly enough, many users are able to achieve very good overclocks with this modest amount of CPU voltage.
Testing for your highest stable CPU frequency
Once the operating system has fully loaded, start up RealTemp. Now start up CPU-Z and verify that your overclocked settings have been properly applied, and that you are running at your desired CPU, bclock, and memory frequencies. Now start up you selected test program, for example OCCT or Prime95. Run the test for five minutes. Then reboot and go back into the BIOS.
If the test ran without error, the raise the bclock by 10MHz, reboot into your OS and run the test again. If the test failed, raise the CPU voltage by 0.025V, reboot into your OS and run the test again. Continue to raise bclock or CPU voltage until you meet one of the following criteria:
- You reach your desired bclock and successfully pass your test.
- You reach your maximum safe CPU voltage.
- Raising the CPU voltage does not allow for additional stability.
Maximum safe CPU voltage
For there is no maximum “safe” CPU voltage in my book. My recommendation is to determine your maximum safe voltage based on your temperatures while running your stability test. With stock air cooling this could be as low as 1.3V on some i7 CPUs while running OCCT. Or it could be as high as 2.2V when attempting Super PI 1M with an i5 670 on liquid nitrogen. Personally, I don’t like to see my load temperatures exceed 90C on air or water cooling, but it’s really up to you.
Is it stable?
So, once you find your highest CPU frequency by meeting one of the criteria above, lower the bclock by 5MHz, and run your selected stability test until you are satisfied. If you are looking for a stable system as a power user or gamer, OCCT or Prime95 for six hours is more than sufficient in my testing, but you may run longer if you desire. But for a true test of stability, I always like to play Crysis while encoding a Blu-Ray movie into an mpeg4 format.
Cooling your CPU, RAM and other heat generating components are key to the success and the level of how high you overclock your CPU. Do not expect to reach the same level of CPU speeds on air or water cooling than your fellow bencher who is using liquid nitrogen.
Cooling not only plays a huge roll in reducing temperatures but is also determine how much voltages you can feed to your CPU before you damage or even kill it, especially with the 32nm Westmere CPU’s who are much more sensitive to voltages. Electronic components behave differently under extreme cooling than under conventional cooling.
Use voltages in moderation and rather be safe than sorry, especially if you cool with conventional cooling such as air or water.
Well, that about wraps it up. Believe me, there is so much more to overclocking. There are SO MANY settings you can continue to fiddle with, but this guide should get you 95% of the way in 5% off the time. If you enjoy overclocking you system, I highly recommend joining one of the teams at OCForums. If you like to push your system and you like to get competitive, look into joining one of our great folding teams, or the OCForums Benchmarking Team. You’ll learn a lot and have a blast!
Please feel free to comment, and post any questions you have in the thread linked below where the great OCF community will help you out with any problems you may encounter!
Thanks to Brolloks and others for their help with this guide.