Benchmark, then boost
Now that we’ve discussed the basic principles of overclocking, I’m sure you’re wondering what settings to try for your own PC! As every processor is unique, there are no standard overclock settings that are guaranteed to work on every PC. It’s time to experiment!
Let’s recap on the goal. We want to attain the fastest processor speed on a PC that remains cool(ish) and stable. By stable, I mean that it POSTs (completes the Power On Self Test), it boots into an operating system, and then successfully completes a benchmark stress test.
My advice here is simple. Benchmark, then boost. Then benchmark again, then boost. And so on. Benchmark your PC in a non-overclocked state, to understand your baseline voltage, temperatures, and clock speeds. Then, increase the base clock frequency and/or multiplier to target a small boost (say, 0.2GHz). Reboot the PC and run your benchmark again to see if it’s stable. If it is, repeat the boost and benchmark. If not, dial back the target settings a little (or boost voltage) and try again.
Overclocking can be a laborious process, but this safe, incremental approach helps you gain a good understanding of your processor’s capabilities.
I’ll use Geekbench 4 today, alongside a great monitoring app, CPU-Z, which provides all the detail I need on voltages and speeds. For temperature monitoring, I use NZXT’s CAM application but you may prefer an alternative such as SpeedFan or another app bundled with your motherboard.
Here’s a snapshot from CPU-Z before overclocking the processor. The figures look static in the screenshot, but continually update depending on processor load.
Pay attention to Core Voltage and Clocks section. The latter displays your overall CPU speed and multiplier. In terms of CPU temperature, at idle, my processor is running at a reasonably cool 32 degrees C.
Open Geekbench 4 (or whichever app you choose) and run your baseline CPU benchmark.
For accurate results, leave the PC alone while the benchmark runs, but keep an eye on the CPU-Z window and temperature monitor to see how speeds, voltage, and temperatures change under load. Once the benchmark completes, results can be viewed in your web browser. Geekbench provides both a single-core and multi-core score, which you should note for comparison with subsequent tests. I noticed a peak CPU temperature of 54 C during the test.
Reboot the PC and enter the UEFI BIOS once again. As we know the CPU can handle 4.2GHz (the default turbo mode speed), for my first overclocking attempt, I’ll adjust the Core Ratio Limit to 44, which targets a maximum speed of 4.4GHz. I haven’t manually adjusted the voltage at this point, but can do so if these settings are unstable.
Reboot the PC and open CPU-Z once again to check core speeds and voltage. With the new settings in place, we can validate a core speed of 4.4GHz and voltage of 1.44 V.
Repeat the benchmark test to see if the PC remains stable with the CPU under load. This time, I noticed CPU temperatures spiking at 71 C with noticeably more fan noise generated by the CPU cooler’s pump and fans. However, the PC remained stable.
CPU temperatures in the low- to mid-70s are about as high as you want to go for safe overclocking. Experience sustained temperatures in the 80s and you’ll certainly be taking some risks with your hardware. But, in the spirit of adventure, I pressed on, incrementing target speeds by just 0.1GHz with each run.
Faster doesn’t always lead to better performance
I achieved an overclock of 4.8GHz before my PC became unstable. At 4.9GHz, it simply locked up when booting into Windows. At this speed, the peak temperature of the hottest core under load hit 87 C, which is a little uncomfortable. I’d be tempted to dial back for safety, but the results show what’s possible—at least for this processor.
One final tip: Faster processor speeds won’t always lead to better performance. The additional heat generated by overclocking can sometimes cause your processor to throttle down, reducing performance. Look at my benchmark results to see the proof:
You can see there’s a steady improvement in the single-core benchmark score as I increased the maximum CPU speed. However, the multi-core benchmark score is more erratic, peaking at 4.5GHz and then decaying as target speed (and peak temperatures) increase. There’s a sweet spot to be found—the best balance of speed, heat, and performance. Its another great reason to test CPU performance in small increments. Benchmark, then boost.
Remember, your results may well be different than mine, courtesy of that silicon lottery. But armed with a grip on the basics, take your time to tweak your own CPU settings and find its sweet spot. A 10 or 15 percent speed boost is yours for the taking, at no additional cost.