Decided to squeeze extra juice out of your GPU and need guidance? You’ve come to the right place – you’ll get all the instructions to overclock GPU you’ll need. You should know that the results will vary. Even if you have an identical GPU model as someone on the Internet, you might have had bad luck and gotten lower-quality silicon that makes the GPU chip. Plus, your cooling solution, air circulation inside the PC case, PSU, old thermal paste, and dust build-up can prevent you from reaching the maximum potential. Luckily, those can and should be improved, but we suggest you learn how to overclock GPU first. Here goes.
Is GPU overclocking dangerous?
The biggest worry people have is damaging their computers. We tried to dispel that notion when we spoke about why you should overclock. If you read it, you know the process is relatively safe thanks to safeguards in place. In the worst case, you’ll see on-screen artifacts and your computer will restart. After the boot, your GPU is back to default settings – no harm, no foul.
What do I need to overclock my GPU?
To overclock your GPU, you will need the following:
- An overclocking software. GPU manufacturers (AMD, Nvidia) and brands that customize and sell them (for example, NZXT, ASUS, Sapphire, EVGA) often design their own software.
- A benchmarking tool. Its purpose is to synthetically push your GPU to its limits so that you can monitor the overclock and accompanying temperature.
And to answer your next question – you can overclock the GPU in your laptop using these steps. Whether that’s a smart choice and will actually work depends on its cooling capabilities and how far it’s pushed by the manufacturer.
With that said, let’s get into the required steps to boost your GPU processing power through overclocking:
1. Run a benchmarking test
To get your baseline values, you’ll need to choose a benchmarking utility. There are plenty of choices, but we always had a great experience with 3DMark and FurMark since they’re both free and extremely easy to use. With that said, here’s how to run a benchmarking test in FurMark:
- Launch FurMark.
- You have two ways to start a FurMark benchmark process:
- Click on one of the presets (4K, 1440p, 1080p, 720p), then click the GO! button.
- Select the resolution manually from the drop-down list, then clicking the GPU stress test button.
- Optional. You can select an Anti-aliasing level (2x, 4x, 8x) to make things a bit harder for the GPU, but you don’t have to.
- That’s all! Don’t do anything on the computer for 10 to 20 minutes.
Before you stop the test, jot down the maximum temperature. If it’s around 65 to 75°C, you have plenty of headroom. If it’s higher than 80°C, you could boost the performance to some degree, but not too high.
2. Select the overclocking software
Now that you have temperature values with no overclock, it’s time to choose software. We mentioned above that there are many, but to demonstrate the process, we’ll use MSI Afterburner. There are multiple reasons for our decision:
- It works for both Nvidia and AMD graphics cards
- It has a simple graphical interface that can be customized
- The settings are easily configurable and hard to mess up
- You can save multiple overclocking profiles for a simple one-click overclock in the future
3. Increase your GPU clock
Now that you installed MSI Afterburner, run it, and take notice of the following values:
- Power Limit, the second bar from the top, is set at 100% by default.
- Temperature Limit, usually set between 83 and 93 Celsius by default.
- Core Clock (MHz) bar, set to +0 by default.
- Memory Clock (MHz), set to +0 by default.
Here are the steps to overclock GPU for the first time:
- Drag the slider on the Core Clock bar to +100.
- Optional. If you have a powerful PSU (power supply unit), you can increase Power Limit to 120%.
- Optional. Increase a Temperature Limit to slightly below the maximum. We recommend any number between 90 and 95 °C.
- Click on the checkmark button (✓) below.
4. Benchmark again
Switch over to your benchmarking tool. Use the exact configuration you used the first time, for consistency’s sake. Let it run for 10 to 20 minutes again while keeping tabs on these 3 things in MSI Afterburner:
- GPU Clock, the big number in the top left corner.
- Memory Clock, the number underneath.
- Temperature, the big number in the bottom right corner.
Noticed any instability, slowdowns, freezes, or weird changes in the FurMark image or experienced a blue screen of death? That means there’s some instability, and your GPU overclocking potential is low. Here’s what you can do:
- If you overestimated your PSU, reduce the power limit to 100% or 110%.
- If you’re sure PSU isn’t an issue, reduce Core Clock by 10 MHz (to +90) and try again until you find a stable setting. That’ll be your Core Clock maximum, unfortunately.
If you, however, saw no irregularities, proceed.
5. Find the Core Clock maximum
Using the same instructions to overclock GPU, increase the Core Clock further by +50 (to +150) in total. Run your benchmark for 10 to 20 min again. If you notice no strange on-screen glitches and your computer is holding up, increase it to +200, then run the benchmark again.
At this point, you should slowly progress upward (by +20 at a time) and benchmark in-between until you start experiencing glitches on the screen or a computer reboot. When you do, start going backward by 10 MHz from the last Core Clock number you had trouble with. When you find a stable Core Clock boost, stop, and remember the number.
Tip: Keep tabs on your temperature throughout the entire ordeal. It should never be too close to the maximum temperature, even if it’s apparently stable. High heat is detrimental to long-term overclocking and will reduce the lifespan of your GPU.
6. Find the Memory Clock maximum
Now that you know the maximum Core Clock value, set it in MSI Benchmark. Now’s the time to drag the slider on your Memory Clock bar. The process is identical, only this time the number will be higher (between 400 and 700, probably). Start with +200, then go to +400. If you see no glitches, increase to +450 until you experience any irregularities. After you do, reduce the number gradually (by 20 or 10) until you find a stable setting. Be aware that with GPU memory, the irregularity might be a decrease in the frames per second in FurMark, rather than a computer crash or visual glitch.
7. Ramp up the GPU Fan Speed manually (Optional)
By now, you’ve noticed your GPU fan(s) spinning faster as the heat rose and slowed down as the GPU cooled down. Well, you can also tinker with the Fan Speed (%) bar to choose one constant fan speed instead. This can help you reach higher Base Clock and Memory Clock numbers, but can also significantly increase the noise. If you do, enable the setting only while you’re doing GPU-intensive tasks.
8. Boost core voltage (Not Recommended)
The value of voltage delivered to your GPU can be increased manually, although not on every graphics card. Usually, if the value is displayed in mV (millivolts) rather than percentages, you can boost it. Here’s how to enable core voltage modification in MSI Afterburner:
- Click the gears (cogwheel) icon to go to Settings.
- In the General tab, put a checkmark in front of Unlock voltage control (and set it to third-party), and Unlock voltage monitoring.
- Click OK.
Now comes the tricky part. It’s recommended you research voltage values for your GPU model and manufacturer. Increase the core voltage by 10 mV for starters, then ramp up the Core Clock until it becomes unstable. At that point, you can keep increasing voltage in increments of 5 mV to 10 mV until you reach the maximum recommended value, then find the new corresponding Core Clock maximum.
9. Put the GPU overclock to the “real” test
Finally, you didn’t overclock your GPU to run benchmarks day and night, but rather for power output in tasks such as playing video games, cryptocurrency mining, or 3D modeling, etc. Well, after applying your stable overclock, you might discover it becomes unstable after 3 or 4 hours of constant GPU activity. If that happens, you’ll have to reduce the numbers slightly – you’re never quite “finished” with overclocking.
How to test GPU stability – The Right Way
You don’t necessarily have a stable overclock just because you apply an overclock and your graphics card doesn’t instantly catch fire. Make sure the new settings can withstand the stress of extended gameplay without slowing down or outright overheating.
Precisely for this, a ton of stress tests are available for free online. These aren’t games, but they appear to your GPU to have an almost identical burden, and they can be looped, so they’ll run for hours without any more input, freeing you up to go have lunch or catch up with a movie or whatever. If after some time the stress test or your PC crashes, you might have overdone the overclocking. Then just go back and decrease the clock speeds or voltage by a fraction before attempting it again.
To make sure your benchmark is accurate, we would also advise playing a few real games. Or if you want to do anything else in the meantime, you may utilize in-game benchmarks for this. Long, GPU-intensive benchmarks, like those from Shadow of the Tomb Raider or Horizon Zero Dawn, are best suited for this. Again, don’t lose hope if you crash or if performance remains unchanged from stock speeds. Just go back and try again after some more tinkering.
GPU stress test tools
- ningine
- 3DMark
- FurMark
- OCCT