Overclocking GPUs is a fairly safe and easy process. If done right, it won’t void your warranty and can boost your gaming performance by at least 10%. Unlike CPUs, you don’t have to mess with the BIOS or worry about BSODs here. You just need a couple of free overclocking tools and a benchmarking application which can be a game or a synthetic like Unigine Heaven or Valley. Let’s get overclocking!
How to Overclock Your Graphics Card
You’ll be needing:
- MSI Afterburner/EVGA Precision X
- A GPU Intensive Game with an in-built benchmark. We prefer Deus Ex: Mankind Divided, The Division 2, or Shadow of the Tomb Raider. You can use Heaven or Valley too, but real games are the best test.
First and foremost, let’s go over the terms used in overclocking:
- Core Voltage: This is the additional voltage you’ll be feeding your GPU to extend the overclocking headroom. Although back in the days of Fermi this was a risky parameter, now, courtesy of GPU Boost, there are too many safeguards in place, so you need not worry.
- Power and Temperature Limit: The power limit is the maximum power your GPU will draw while operating at peak frequencies without throttling the clocks. The temperature limit is the same, except, the GPU will start throttling as soon as you hit 75+ degrees. This is the point when there’ll be a substantial drop in the core clocks to keep the temperatures in check.
- Throttling: This is the gradual decrease in the in-game core clocks that is a safety measure put in place by the OEMs to keep you from frying your GPU. Usually, when the GPU temps cross 75 degrees, or the TDP goes above the set limit, your GPU clocks will start throttling.
- Core Clock offset: This is the figure by which you’ll be overclocking your GPU core.
- Memory Clock offset: This is the figure by which you’ll be overclocking your GPU memory.
- Limiting factor (PerfCap-NVIDIA only): This is the reason why you can’t push your clocks past a specific limit. It can be the thermal limit, the power limit, the voltage, or simply your silicon lot in life.
How to Overclock Your GPU Core:
- Firstly enable temperature and core clock monitoring and fire up your test game to see how the temps are holding up prior to the overclock. If the average is below 75 degrees, you’re good to go. If not, then do something about your thermals. Check your case ventilation, increase your fan speed (at the expense of noise) or simply buy a custom water block or something.
- Next, max out your core-voltage, power limit and the temp limit (don’t worry as long as the temps are stable, you’ll be fine).
- Then add +100 to your core clock and hit apply. Run a game in windowed mode, and keep an instance of GPU-Z running on the side as well.
- Scroll down in the GPU-Z sensor window and you’ll see a bunch of options. Out of these, you need to keep an eye on the GPU clock, GPU temp, power consumption, and PerfCap Reason. If the game runs for 15-20 minutes without crashing and the temps and power stay below the safe limits, you just managed to overclock your GPU by +100 MHz.
- Increase the core offset by another +100MHz. Continue monitoring the GPU clock and note the FPS average and lows through the course of the benchmark run. If it’s higher than the previous run, you’re good to go.
- If you experience a crash or image artifacts, then revert to the last stable setting. This is your GPU’s highest stable core clock.
- The PerfCap Reason will tell you what’s the reason you can’t further increase your core clock. It may be voltage, power or temp, or a combination of the three. You can do something about the latter, but voltage and power can’t be tweaked using conventional means.
- On AMD GPUs where the PerfCap option isn’t available, keep an eye on the power consumption percentage (or power if you know the TGP) and temps. If you are running into a power or TDP ceiling, there will be a hard GPU clock throttle as soon as you hit a certain thermal or power limit. Keep an eye out for this.
- For power, it will be near the max power limit set up via Afterburner or Precision X.
- For temperature, it is usually around 75 degrees.
How to Overclock Your GPU Memory:
- Overclocking the memory is more straightforward. You can now close GPU-Z and increase the memory clock by 200MHz in each step. Keep going till the game crashes or you start seeing artifacts like these:
- They can be subtle or easily noticeable, and usually, start appearing right before the memory is about to become unstable. If you overclock the memory further, it’ll probably crash. So these artifacts are a good indicator of stability.
- It’s also important to keep an eye on the average FPS and the lows with each benchmark, as newer GPUs like the RTX 3080/3090 include ECC memory which prevents the game from crashing even if the memory is unstable, as the errors are automatically corrected.
- Instead, the performance will start degrading once your memory overclock becomes unstable due to the repetition of the same transfer cycles due to errors.
Tips for Overclocking Graphics Cards
Overclocking your GPU is a slow and painstaking process. Your part might not overclock at all or you might get lucky and win the silicon lottery. Regardless, here are a few tips to keep in mind while overclocking:
- It’s a good idea to keep track of the frame rates (both average and min). Often right before your GPU becomes unstable, you’ll notice abnormally low min FPS as well as relatively lower averages.
- If the GPU clock is throttling, consider increasing the voltage/power limit in case you haven’t already.
- If the GPU keeps on crashing and you can’t figure out a stable configuration, test out the GPU Core overclock and the memory overclocks separately and determine which one is the culprit. Then combine the two and see if things go smoothly. In most cases, it’ll be the core offset and not the memory.
- That’s about it. If you run into any problems, let us know in the comments section below. Happy overclocking!