Clear all

[Sticky] GPU Overclocking Guide

Member Moderator

Some people are unsure on how to overclock the GPU. Quite frankly, the GPU is the easiest thing to overclock in a system. Lets start from the beginning.

The first thing I recommend is to grab some useful programs and the latest drivers. The programs I recommend are Rivatuner, MSI Afterburner, or EVGA Precision (for the overclocking/monitoring), GPU-Z (monitoring/GPU info), and OCCT, Furmark, MSI Kombustor, or EVGA OC Scanner for stability testing.

Proper uninstallation/installation of video drivers (Win 7/Vista, in XP it should be somewhat similar):

  • Download but do not install desired driver.
  • Download and install Driver Sweeper or something similar.
  • Right click on computer, properties, click Device Manager (on left).
  • Select your display device (video card), uninstall video device software.
  • Delete driver folder (located in C:NVIDIA).
  • Reboot in safe mode (repeatedly hit F8 during boot sequence).
  • Run Driver Sweeper or your equivalent. Only clean Display and PhysX driver.
  • Reboot.
  • Install your desired driver.
  • Reboot.

When installing new drivers, make sure all anti-virus software is disabled.

MSI Afterburner (also includes latest version of MSI Kombustor)

EVGA Precision

RivaTuner 2.24c



Driver Sweeper

EVGA OC Scanner

For those of you who don't know, artifacts occur when your GPU is overclocked too high and/or your temperatures are too high (but not high enough for a shut down). If you get artifacting right out of the box with good temperatures and you did the proper method of installing new (and different) NVIDIA WHQL drivers, I suggest you RMA your GPU. Make sure your temperatures stay below 105 degrees Celsius, that is the max most cards can handle. It is best to stay below 90 degrees Celsius, artifacts can occur before 105.

For this guide I'll be using my primary video card for example, an EVGA GTX 480. This is the vanilla (default) version which came at the factory clocks.

Factory Settings:

  • Core Clock: 700.5 (listed as 700 or 701 depending on the program) mhz
  • Processor Clock (AKA Shader Clock): 1401 mhz
  • Memory Clock: 925 mhz (3700 mhz effective data rate)
  • Fan Speed: Auto (44%)

Uh oh, we ran into another confusing term here.

Effective data rate? What does that mean? My card was supposed to come with 3696 mhz memory but it is shown as 924 mhz!

High end video cards used to use double data rate memory (GDDR3 or in rare occasions GDDR4). They now use GDDR5 memory which is quadruple data rate. For GDDR/GDDR2/GDDR3/GDDR4 memory, multiply the speed shown in the above programs by 2, and this gives you your effective data rate. For GDDR5 memory, multiply the speed shown in GPU-Z by 4. The effective data rate is usually what is advertised.

924 mhz x 4 = 3696 mhz effective data rate

  • Raise core clock or processor clock/shader clock by no more than 25 mhz. The core clock should never run at over 1/2 the processor clock/shader clock speed.
  • Stress test your video card for at least one hour at a large resolution. Ensure your GPU temperatures do not exceed 100 degrees Celsius.
  • If stable (no artifacts detected), repeat the first two steps in that order. If not stable, you might have to move on to voltage adjustment. More on this later.
  • Once you have found your max stable core/shader clock, move onto memory overclocking. Raise the memory clock by 25 mhz, stress test for an hour, and repeat if successful.
  • Several overclocking programs support voltage tuning. Voltage adjustment can be dangerous. Only adjust voltages after finding your max overclock without voltage adjustments. Follow these steps on how to overclock using voltage adjustments.

    • Once you have found your max overclock without voltage adjustments, raise the voltage to the next increment.
    • Raise core clock or processor clock/shader clock speed by 25 mhz.
    • Stress test for one hour.
    • If you pass the stress test with no artifacts, repeat steps 2 and 3 only. If not stable, raise voltage to the next increment but do not increase clock speeds. Stress test again. You must find out your max overclock for every voltage increment before moving on.

    Here is an example of an overclocking scenario, using my GTX 480.

    Lets get started. Unlink the core/shaders if you have a last gen NVIDIA card, and set fan speed to manual and raise it a little. You can push the sliders up, or enter a number and it will give you your desired clock speeds in most overclocking programs.

    I increased the processor clock speed from 1401 mhz to 1425 mhz after raising the fan speed to 80%. This makes my core clock go to 725 mhz automatically. I ran the EVGA OC Scaner to test for artifacts. For stability testing, use all of the features included in the program you're using (MSAA, unlock power draw, etc.) and run it at a large resolution. Stress test for at least an hour. Ensure your GPU temperatures remain below 100 degrees Celsius. I didn't stress test for an hour in these images since I've already done all of this.

    If your card is stable, move on. If not, tough luck. But virtually any card can do such a small overclock. If it can't, then you should probably raise the voltage.

    Raised processor clock to 1500 mhz, no errors. But it turns out this is my max stable overclock without voltage adjustments. Now its time to adjust the voltage. I raised the voltage from 987 mV to 1000 mV, and raised processor clock to 1525 mhz. It works!

    So I repeated what is in the list above, and got to 1725 mhz processor clock. I had to max out the voltage for this overclock to be stable.

    Now that I have found my max processor clock (therefore max shader clock), I can adjust my memory clock. See the list above again.

    Does overclocking void my warranty?
    For EVGA cards no, I can't speak for anyone else sorry.

    I ran into an error. What do I do?
    Lower the speeds a little bit. If you get the nvlddmkm stopped responding and has recovered error, try lowering the memory clock first. If it still persists, lower the core. If it still persists, lower the shader clock too. Lower them to the last successful clock you had them at, and move forward again a little bit more. Sometimes you can get a BSOD from the same error. These blue screens are pretty detailed and usually say nvlddmkm stopped responding and has failed to recover or something along the lines of that. Follow the same procedure.

    I can't overclock as high as others who have the same card. Why is this?
    Not all cards are the same, some overclock further than others. Water cooling your GPU can help you reach higher speeds.

    How much does GPU overclocking affect performance?
    In games, the difference isn't that noticeable unless you reach a very high overclock. In GPU heavy benchmarks such as 3DMark Vantage you'll get a nice increase in your GPU score and even your overall score if you've achieved a decent overclock.

    EVGA offers SC, SSC, and FTW models. What is the difference?
    These are factory overclocked, guaranteed to work above the default speeds. In my opinion the SC/SSC models aren't worth it when there is a FTW model for that card, from what I've read these are the parts that fail to reach the FTW spec but I'm not too sure on this. The FTW cards are priced very high but they also have very nice overclocks on them. Same goes for the SSC on cards that have no FTW model. I passed the FTW speed on my GTX 260, but failed to reach the EVGA SSC speed on my PNY 9600GT (9600GT SSC comes at 740/1836/975).

    What is the max safe voltage I can use?
    Whatever is stable. MSI Afterburner uses the max setting allowed in the BIOS for your card, so any voltage setting in MSI Afterburner is safe unless your temperatures get too high.

    When you reach your max overclock, you can upload a validation file on GPU-Z for bragging rights 😉

    This isn't even my max overclock. I can go higher if it wasn't for heat (ran 1150 mV and 405 KHz PWM frequency for these speeds) and the program I used to use such voltages is buggy.

    Final Settings

    • Core Clock: 862.5 mhz
    • Processor Clock: 1725 mhz
    • Memory Clock: 1050 mhz (4200 mhz effective data rate)
    Topic starter Posted : 13/04/2009 2:52 pm
    Member Moderator

    Excellent work Jester! 😀

    Nicely outlined and easy to understand!


    Topic starter Posted : 13/04/2009 2:52 pm

    Thank's a lot for this useful guidelines man. I have recently been trying to get into the world of OC, and this will be my first attempt ever with anything related to overclocking. And starting with the GPU, as you mentioned is one of the easiest procedures to OC.

    So I tried to play around with a few settings in EVGA Precision, using the OC Scanner and MSI Afterburner for Voltage, after reading this guide after my understanding. And I would really appreciate some approval/guidance on how to move on with this within reasonable increases. First of all because it's fun and exciting, as I am entering the world of OC and see how much I really can squeeze out starting with the GPU.. 🙂

    Ok, so first of all I really don't mind the fan noise coming from my computer as if it's about to levitate, that's what I have the headset for!

    1. I popped up EVGA Precision [ Default Settings]

    Core Clock: 607
    Shader Clock: 1215
    Memory Clock: 1674
    Fan Speed: Default 40 - Usually tune it up to 85-95 while playing heavy games.
    Voltage: 1000 (MSI Afterburner)

    With this settings I usually have about 50-62 Celsius GPU temp, under very heavy load/maxed out settings, or almost. Fan Speed 80-90
    And on Idle everything from 25-35 temp - Fan Speed 40-50

    Keep in mind that the airflow in my case is pretty decent to be honest, and I rely on that as I only have stock fans attached on my GPU/CPU.

    I have been playing around and been observing the temp's using Aida64/EVGA P/OC Scanner and GPU-Z. And have been writing down my clock's until I found my max overclock without any voltage increase.

    This where I currently stand without voltage increase.

    Core Clock - 788
    Shader Clock - 1576
    Memory Clock - 1746
    Voltage - 1000

    I ran OC Scanner for about one hour and thirty minutes, with one artifact detected. And then I tried to tune up the voltage to about 1020 and ran the test again, without any artifacts detected. But as I was playing the Witcher 2 for about 30 minutes I got this graphical explosion on my screen, and I have carefully been monitoring the temp's using GPU-Z and OC-Scanner and in game. And it would not exceed over 75 C with the voltage increased to 1020, and the temp without me increasing the voltage ( 1000 ) I have no problems whatsoever. And I have set my temperature to auto lock-down to 90 Celsius, just to be on the safe side. What could be the cause of this? or what do you guys suggest I do?

    I apologize if anythings seems unclear or messy, let me know if you need any other information or if anything seems to unclarified. Thank's a lot on for hand guys!

    Posted : 14/07/2011 9:16 am

    Have you tried any other games? Maybe it's because of the witcher itself?

    Posted : 14/07/2011 10:00 am
    Member Moderator

    Try putting the memory clock back to default speeds - the OC Scanner is a poor GPU memory tester. The best you're gonna get for GPU memory testing is probably 3DMark 11 or Unigine Heaven.

    Topic starter Posted : 14/07/2011 10:09 am

    iGeForce, thank's for the reply. I very much doubt it's the game itself as the extreme freeze up seemed to be pretty much GPU related and I have playing the game for weeks now and had no problems. So I can only assume this is due the fiddling with the GPU. And no I have yet to try out other games!

    Alright man, downloading Unigine Heaven as we speak. Do you mean putting the memory clock back to default after the volt increase or at default?

    Posted : 14/07/2011 12:34 pm
    Member Moderator

    Keep volt increase but put memory back to default speed.

    Topic starter Posted : 14/07/2011 12:40 pm

    Ok, so now I have added my Memory Clock back to stock.

    Current Setup:

    Core Clock: 788
    Shader Clock: 1576
    Memory Clock 1674
    Core Voltage: 1062

    Been doing some testing meanwhile, and found out that everything below Core Voltage 1062 would make things unstable, not temperature wise though. Been running MSI Kombuster and Unigine Heaven on pretty high settings, seems stable for now. And did the final test by playing the Witcher 2, with everything maxed out without UBER Sampling ( Very GPU hungry ) and V-sync off. For over an hour or so, smoothly with 50-60 FPS with the temperature not exceeding 75 Celsius. So now what? should I just stick to this and be happy or what do you suggest? 😉

    Posted : 14/07/2011 4:16 pm
    Member Moderator

    You can stay like that, but it looks like you have some headroom left. 🙂

    Topic starter Posted : 14/07/2011 4:43 pm

    Any inquires? 🙂

    Posted : 14/07/2011 8:29 pm

    One question though, is it safe to switch between voltages regulary from low to high, MSI Afterburner profile wise?

    Posted : 17/07/2011 11:49 am
    Member Moderator

    People do it all the time without issue so I'd expect it's safe.

    Topic starter Posted : 17/07/2011 8:24 pm