CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Learn the pros and cons of overclocking your CPU, the physics behind changing your clock rate and the basics of overclocking an Intel or AMD processor.

Overclocking was once the domain of enthusiasts with higher-than-average hardware know-how and a bit of derring-do. The community was made up of benchmark competitors keen on pushing CPU frequency envelopes, gamers trying to squeeze the last drop of performance out of an aging rig, or simply power users who wanted to chart the undocumented and unadvertised limits of their system. A classic example of the sort of spirit around overclocking/modding was the “DIY Cooking Oil PC” we showcased here at Tom’s Hardware in 2006.

DIY Oil PCDIY Oil PC

Times have changed. As with many other niche “tweaks” to system performance like liquid cooling, vendors have embraced overclocking, avidly promoting the capabilities of their hardware, providing software and firmware tools to make overclocking far easier, and, for a premium price, providing pre-built, overclocked systems with specifications that would have made many of us swoon in the 2000s.

Liquid-Cooled build, a necessity for many overclocked systemsLiquid-Cooled build, a necessity for many overclocked systems

The mainstream adoption of overclocking has also been pushed by a plethora of new applications; not just gaming, but currency mining and decentralized scientific computing like BOINC and protein folding. And while the actual process of overclocking has been vastly simplified in recent years, it is not a blind one. Often, overclocking is a matter of looking at a system build as a whole and eliminating bottlenecks, not just pushing one component to its limit.

For example, you can run some Core i7-3770Ks at more than 5.1 GHz (expect a voltage setting of around ~1.45V), but if the system is being used for scientific computing (or any other application that requires manipulating large data sets), memory data rates may become your performance bottleneck.

We focus on CPUs in this article, but both system memory and graphics processors are also overclockable. And many beginner-enthusiasts may realize that instead of overclocking the processor for better performance, they can simply upgrade the system’s cooling to prevent the processor’s built-in thermal throttling from kicking in under heavy load.

Despite recent shifts in hardware capabilities, the core concept behind overclocking remains the same. Components with a clock—an oscillator—have a margin of performance (frequency) that we refer to as headroom, which is available above and beyond the advertised default settings. Some of the headroom is there because of engineered safety margins for the hardware, based on the thermal performance and available voltage constraints of a nominal system. That is, a mass-market component must not put out so much heat that only the top 5% of PC builds have the cooling capability to handle it. This is called an “intentional guardband.” Extreme overclocking eats into the guardband as well as the conservatism in the design of the hardware and silicon fabrication process.

Another portion of the headroom exists because the stock value is a stable set-point determined during manufacturer testing. For example, a given CPU and system configuration may crash least often when it is operated at 2.5 GHz below the available maximum.

Finally, manufacturers are loath to hand out what is essentially a free performance increase to overclockers without charging for it; Intel’s multiplier-locked and -unlocked CPUs are great examples, where identical chips are sold with and without artificial frequency caps, with a premium charged for the ability to overclock.

A component's frequency can be increased through various means, and higher component voltage is often used to provide the stronger signal needed at that higher frequency. This guide doesn’t discuss which processor is “better” for overclocking; every overclocker has a preference, and biases come into play when recommending CPUs. We do have an older AMD FX-8350 vs. Intel Core i7-3770K story that discusses bottlenecks, in case you want to compare the state of the top processors a few years ago, though.

Heat, Stability, Damage, and Warranty Considerations

The performance-over-stock achieved by overclocking a system is not just a minor tweak; HWBot member just_nuke_em overclocked the relatively inexpensive quad-module AMD FX-8120 with a stock clock rate of 3.1 GHz to 8.3 GHz, a more than 250% increase over AMD’s published specifications.

AMD FX-8120; the AMD FX series has set most of the world records for overclocked CPU frequencies, but Intel’s i7 series can take higher core temperatures and has newer processors in market.AMD FX-8120; the AMD FX series has set most of the world records for overclocked CPU frequencies, but Intel’s i7 series can take higher core temperatures and has newer processors in market.

While most overclocks will be more modest than this world record, any performance boost comes with some limits imposed by physics. As a chip’s clock rate and voltage are increased, the waste heat output of the system also increases quickly, and this heat has to be removed somehow. Often, the cooling capacity of a build will max out long before the component’s theoretical maximum frequency. And CPU cooling will continue to become less efficient, not more, as time passes; each generation of CPU has a higher transistor density than the one before. Intel went from the 45nm Nehalem die in 2008 to a 14nm die in Skylake in 2015, and Cannonlake (slated for release in 2017) is going to be built on a 10nm process. AMD follows a similar progression.

While the transistor count tends to increase with each new architecture, die size does not, making it much more difficult for conventional cooling solutions to keep up with the rate at which thermal energy is generated. In fact, as dies get smaller, the total surface area contact between the CPU and its heat spreader drops, making cooling less efficient. All of this contributes to a much higher predilection for “hot spots” in current chips. Of course, as you increase voltage to (hopefully) stabilize more aggressive overclocks, power consumption rises very quickly. Core temperatures tend to jump for small, incremental frequency boosts.

Intel 22nm vs 14nm process, the reduction in die size provides a number of speed advantages, but makes thermal problems worse on the whole. (Photo Credit: Intel)Intel 22nm vs 14nm process, the reduction in die size provides a number of speed advantages, but makes thermal problems worse on the whole. (Photo Credit: Intel)

System stability is often another victim of overclocking. Enthusiasts sometimes have to live with more system crashes and less consistent performance. That's not to say every overclocked system is less stable than stock; many overclockers have reported finding new and better operating points at higher-than-stock clock rates. However, CPUs operating beyond their factory specifications are more susceptible to shorter lifespans due to the stresses imposed on them.

Potential damage sites for an Intel processor—it is unlikely that normal overclocking (i.e. overclocking via the BIOS and software options) will have such a detrimental effect on a system. (Photo Credit: Intel)Potential damage sites for an Intel processor—it is unlikely that normal overclocking (i.e. overclocking via the BIOS and software options) will have such a detrimental effect on a system. (Photo Credit: Intel)

Causing damage and voiding warranties are two oft-cited reasons that people are hesitant about overclocking. Damaging components due to heat or voltage overloads was easy in the old days, and it’s still possible now. But manufacturers incorporate a number of fail-safes, including thermal throttling, and the truth is that it is far more likely that a system will become unstable and crash before there is any permanent damage caused by a short-term test run.

Overclocking will, however, reduce the lifespan of system components; not just the processor, but motherboard, memory, and other parts that were stressed beyond their designed operating points alongside the overclocked processor. In electronics, the biggest source of wear is a phenomenon known as electromigration, whereby ions are slowly transferred from a structure to the adjacent structure under the force of electrical current. Major contributing factors include increased heat and voltage, but the limits of heat and voltage vary with different materials, different production technologies, and expected component lifespan. Thermal loads in particular tend to accelerate electromigration in ICs.

In terms of warranty, while GPU and motherboard manufacturers have become more overclocking-friendly lately, both Intel and AMD warranties are voided if the clock rate of their CPUs is changed at all. Intel does offer a “Performance Tuning Protection Plan” that will replace an eligible processor that falls outside of Intel’s specifications, but AMD will not cover a processor that was operated outside published specifications, even when AMD’s own Overdrive software is used for overclocking.

MORE: Best PC Builds

MORE: How To Build A PC

MORE: All PC Builds Content

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
34 comments
Comment from the forums
    Your comment
  • rush21hit
    Look at the voltage setting on those world records. Lots of LN2 involved here.

    Speaking as a former user of Q6600 @3.0 (Corsair Hydro series were non-existent back then) and i7 920 @2.8. Small OC, I know. Since I want my motherboard not get fried.

    Motherboard gets the most hammering in OC. Even the best of the best sometimes failed. Those capacitors can only hold so much. I just want to ensure it also last. Heck, even in normal use, motherboard tends to fail long before any other parts does.
    1
  • blackmagnum
    Careful not to void your warranty. Leave extreme overclocking to the pros.
    1
  • Calculatron
    Ever since the articles on the "E" series of the FX-line-up, I've tried to take efficiency into account. I run a mild overclock of 4.0ghz with just under 1.2V on my FX-8320. I can achieve a 4.7ghz overclock, but I just don't think that the performance gains are worth all the extra heat and stress on the components.
    1
  • Nuckles_56
    There's a typo in the table of record speeds for the CPUs and the respective voltages, as there is no way that the i7 860 got to that speed at 1.152V
    0
  • SinxarKnights
    Anonymous said:
    There's a typo in the table of record speeds for the CPUs and the respective voltages, as there is no way that the i7 860 got to that speed at 1.152V


    http://valid.x86.fr/xuxjnn look for yourself. I suspect there was an error reading the voltage but I do not know, CPU-Z calls it valid though.
    0
  • TexelTechnologies
    Its not that complicated but then again I am always surprised by how dumb people are.
    0
  • Murdock4321
    Overclocking used to be pretty complicated and take some trial and error. With all these new processors and new bios', its really pretty easy. I'm still rocking an i7 930 @ 2.85ghz which is rock solid and has lasted me since 2010
    1
  • leo2kp
    i7 970 @ 4GHz here with overvolt. 6 years old and still going strong.
    0
  • jtd871
    How about an underclocking/undervolting guide for those that want efficiency rather than performance?
    1
  • anbello262
    Well, isn't the procedure exactly the same, just lowering mult and volt instead of rising it? The stability concern (freq/volts) and the iterative methodology is the same, right?
    0
  • redgarl
    I have a 4770k and my chip can barely OC to the boost clock. Also, they run soo hot, it's retarded.

    My 2 cents, OC is nice, but you are better investing, if anything more than 30$ in coolers, in a better cpu than landing 100$ on a Corsair something. You don't even know if the chip is going to be stable anyway.
    0
  • Eximo
    Well the i7-4770k is notorious for having really bad samples that barely meet the qualifications of being an i7-4770. I can get mine to 4.5GHz, but I have to go thermonuclear at 1.35 volts to do it, managed to bench it for 3 minutes at 103C before deciding to back it down.

    Proof is in the i7-4790k.
    0
  • 10tacle
    I've been overclocking since the late 90s with the infamous first monster overclocker that everyone could get their hands on, the Celeron 333MHz->450MHz. Over the past nearly 20 years (holy cow has it been that long?) slowly evolving from a 1280x1024 monitor to a 2560x1440 monitor, I have noticed that overclocking the CPU offers diminished returns in gaming performance.

    Today it's more about GPU performance, especially overclocking GPU performance. Like someone said above, it's no longer worth the heat and power increase that it used to be. Overclocking a GPU is where the real game performance increase benefit is. On the other hand, productivity applications like video rendering can benefit from a CPU overclock quite well...about the only thing I ever find myself cranking my 4690K up to 4.7GHz for these days to speed rendering time up.
    0
  • SinxarKnights
    Anonymous said:
    I've been overclocking since the late 90s with the infamous first monster overclocker that everyone could get their hands on, the Celeron 333MHz->450MHz. Over the past nearly 20 years (holy cow has it been that long?) slowly evolving from a 1280x1024 monitor to a 2560x1440 monitor, I have noticed that overclocking the CPU offers diminished returns in gaming performance.

    Today it's more about GPU performance, especially overclocking GPU performance. Like someone said above, it's no longer worth the heat and power increase that it used to be. Overclocking a GPU is where the real game performance increase benefit is. On the other hand, productivity applications like video rendering can benefit from a CPU overclock quite well...about the only thing I ever find myself cranking my 4690K up to 4.7GHz for these days to speed rendering time up.


    I agree for the most part. However we are at a point where Intel is really the only option for a high end gaming PC. Sadly even AMD's best CPU offerings can't effectively use their own top end GPUs.

    In other words, that is true for Intel, not so much for AMD users. Unfortunately due to the cooling requirements of the flagship AMD chips, it is often impossible for a person to overclock it much, but even a modest OC to the CPU can give significant FPS gains depending on how unbalanced the PC is (i.e FX-8350 with a 980ti/1070).
    0
  • mitch074
    Well, the Celeron 333 wasn't the infamous monster - it was the 300A (4.5 x 66 MHz) that could reach 450 with a mere bus switch to 100MHz and a proper mobo or at worst shorting out the B21 pin; and even then, performance improvements mainly came from overclocking the graphics accelerators of the time (like the 3dfx Voodoo or the Nvidia TnT).

    As for me, I started overclocking with the Pentium 75 (mine could reach 112.5), a Cyrix P150+@133 and a Pentium 133@166 (2x83MHz and 512 Kb of L2 cache, it could kick a P200's butt). While the Cely 300A@464 I had was the biggest overclock (in proportion) I ever got for myself, I did push several (as in, 4 or 5) Sempron64 from 1.6 to 2.4 GHz without any other tinkering than raising the (cheap) mobo's reference clock and using the stock cooler; I don't doubt for a minute that, had I improved the cooling or used better mobos than bargain bin stuff and spent more time adjusting voltages, a couple of those would have reached 2.6 or 2.7.

    Because yes, there was a time when owning an AMD entry-level chip was the veryh best bang for bucks you could get; I still have an Athlon II X4 620 (cheapest quad core, $99) with a base clock speed of 2.6 GHz; with a nice aftermarket cooler and a good mobo, I pushed that one to 3.4GHz stable enough in 2009 to convert DVDs and BRs all night long (I bought a NAS then and backed up all my disks). I must admit that after a week of feeling it heating up my office, I decided to retire it to the living room and replaced it with a Haswell i5 4670K. That one is disappointing though: with a factory boost speed of 3.8GHz, it just can't hold stable past 4.2 GHz. Since I don't feel like cracking the heat spreader open to cool it down more, I make do with a barely-more-than-10% overclock. Meh - still good enough for today's games.
    0
  • Eximo
    I recall having a Duron 750Mhz that would run 1.2Ghz without too much trouble. You make me want to dust off my Voodoo collection and see if they all still work. Used to have SLI Voodoo 2 12MB (GPUs had three times the memory as my computer) with a Pentium @ 83Mhz. When is Nvidia going to dust off the Voodoo brand name, anyway? Should have been Voodoo, Voodoo X instead of Titan.
    0
  • mapesdhs
    "... you can run some Core i7-3770Ks at more than 5.1 GHz"

    Not without delidding you can't, or some really crazy cooling (bad idea to allow casual readers to think a normal oc for a 3770K could ever be that high). IB runs way too hot, deliberately so re Intel's use of a naff interface material (check the toms and other articles from back then, and the YT vids on delidding showing load temp drops of 30+C).

    It's far easier to oc a SB 2700K to 5GHz, works just fine with a simple TRUE and one fan, takes 5 mins to do this on an ASUS M4E. I've built seven such systems so far, about to do another as a charity build though with a different mbd.

    Btw, I've still yet to see a general oc'ing guide that's better than Miahallen's information:

    http://www.overclockers.com/3-step-guide-overclock-core-i3-i5-i7/
    http://www.techreaction.net/2011/01/04/3-step-overclocking-guide-%E2%80%93-sandy-bridge-v0-1beta/
    http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-lynnfield/
    http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-bloomfield-and-gulftown/

    The general principles apply to any system.

    Ian.
    1
  • 10tacle
    Anonymous said:
    You make me want to dust off my Voodoo collection and see if they all still work. Used to have SLI Voodoo 2 12MB (GPUs had three times the memory as my computer) with a Pentium @ 83Mhz. When is Nvidia going to dust off the Voodoo brand name, anyway? Should have been Voodoo, Voodoo X instead of Titan.


    Hah I still have my two 8MB Voodoo 2 cards somewhere. They were my first SLI cards and lasted from '98 to I believe '02 when upgrading to a Ti4600 when moving up to a more GPU hungry 1600x1200 CRT. Didn't go SLI again until 2009 and GTX 275s. And yes, bringing back the Voodoo name would be awesome. A whole new generation of PC builders have come around since then.
    0
  • Eximo
    My pride and joy is an operational Voodoo 5 5500 I got from ebay. I have it in a ludicrous Poweredge 4200 with dual Pentium II 333Mhz and 256/512MB of ECC (Can't seem to get the BIOS to update to see the 512MB), Soundblaster Awe 32 with Midi daughterboard, Two hot swappable 10GB SCSI drives, a pair of 700W Dell hot swappable power supplies. It plays a mean UT99.

    I bought it at a swap meet, no idea the whole computer was intact and working except for not having any hard drives. So I threw a few hundred at it to make it do something.
    0
  • kunstderfugue
    Is there a reason not to overclock in the following manner?

    Slowly push the CPU multiplier until you can't boot, then dial back a notch and bench, if you hang while bench dial it down until you cemplete a bench. Once you hit that point at a voltage that's comfortable in temps, dial it down another notch and do a long stress test
    0