You don’t need a new GPU to get more frames. Overclocking your existing graphics card costs $0 and typically delivers 5–15% more performance — gains that previously required spending $300 or more on a GPU upgrade. If you’ve been leaving that performance on the table because overclocking sounded complicated or dangerous, this guide changes that.
The process is straightforward: download one free tool, adjust four sliders in a specific order, and run a stability test. Modern GPUs have built-in thermal protection so they won’t damage themselves if you follow the steps. By the end, you’ll have a saved overclock profile you can load with a single click before every session.
This guide pairs with our PC Optimisation guide for the full performance picture. If GPU settings terminology is new to you, start with Game Settings Explained first.
What GPU Overclocking Actually Does
Your GPU has two main performance levers: the core clock and the memory clock.
The core clock — measured in MHz — controls how fast the GPU’s shader processors run. These are the thousands of compute units doing the actual rendering work each frame. More MHz here translates directly to more FPS. The memory clock controls how quickly data moves in and out of your VRAM. It matters most at higher resolutions and in VRAM-heavy games; raising it gives a secondary performance boost on top of the core clock gain.
Manufacturer default clocks are conservative. GPU makers set speeds low enough to guarantee stability across all reasonable system configurations — adequate airflow, not-too-hot ambient temperatures, any compliant power supply. If your setup is average or better, your GPU almost certainly has headroom to run faster.
How much headroom depends on something called the silicon lottery: chips from the same GPU model vary in quality due to manufacturing tolerances. Some chips can run 200+ MHz over their base clock stably; others max out 50 MHz higher. You won’t know where your chip lands until you test it — which is exactly what this guide walks you through.
Realistic expectations: a conservative overclock adds 5–10% FPS. A moderate overclock — what most beginners reach on a first attempt — adds 10–15%. An aggressive overclock on a chip with good silicon headroom can approach 20%. That’s the equivalent of one GPU tier, at zero cost.
Safety Limits and Pre-Flight Checks
GPU overclocking is safe when you respect two hard limits.
Temperature limit: 83°C under sustained gaming load. At around 83–87°C, most GPUs begin thermal throttling — automatically reducing clock speeds to cool down. Running consistently above 90°C accelerates a process called electromigration: metal atoms in the transistors literally migrate under sustained heat, eventually creating micro-shorts that cause permanent degradation. A well-cooled overclock running at 78–82°C will have negligible lifespan impact compared to stock. One sustaining 92°C is genuinely shortening the card’s life.
Voltage: don’t touch it yet. Voltage is the other damage vector, and it’s more dangerous than temperature alone. Increasing voltage doesn’t increase heat linearly — it increases heat quadratically. A +50mV bump for a marginal clock gain generates far more thermal stress than that gain justifies. Beginners should leave voltage at its default for a first overclock. Advanced users can actually go the other direction — reducing voltage while maintaining or exceeding stock clocks (called undervolting) — but that’s outside beginner scope and covered separately.
Before touching any sliders, complete these three checks:
- Update GPU drivers. Old drivers can cause instability that looks like an overclock failing. Install the latest from NVIDIA or AMD’s official sites before you start.
- Check baseline temperatures. Run a demanding game for 20–30 minutes and note your peak GPU temperature. If you’re already hitting 85°C or above at stock speeds, clean your heatsink first — dust buildup is the culprit. There is no point overclocking a card that’s already thermally throttling.
- Close background processes. Streaming software, open browsers, and Discord video calls consume VRAM. Start clean so your stability tests reflect actual gaming conditions, not a cluttered desktop.
Download MSI Afterburner
MSI Afterburner is the industry-standard GPU overclocking tool. It’s free, works with all GPU brands (NVIDIA, AMD, and Intel Arc), and has been the go-to utility for years because it reliably does everything you need. Download from MSI’s official site. When the installer prompts you, also install RivaTuner Statistics Server — this is what puts real-time GPU stats (temperature, clock speed, FPS) on your screen while gaming, which you’ll need for monitoring during tests.
After installing: open Afterburner, click the gear icon to open Settings, go to the General tab, and enable both Unlock voltage control and Unlock voltage monitoring. You won’t be adjusting voltage, but unlocking these gives you visibility into what your GPU is actually doing during load — useful information for diagnosing any instability.
How to Overclock Your GPU Step by Step
Follow this order exactly. Each step builds on the previous one. Don’t skip ahead to memory clock before finishing core clock — if something goes wrong, you need to know which adjustment caused it.
Step 1 — Set Power Limit to 100%
Find the Power Limit slider (displayed as a percentage) and drag it to its maximum value, then click the checkmark to apply. By default, GPUs cap power draw to stay within their rated TDP (thermal design power). At 100%, the GPU can draw its full power budget under load. Without this step, any clock increase you apply will be throttled back immediately under gaming load — the GPU will boost briefly then pull back to fit within the original power envelope. This is the most common reason beginner overclocks produce no visible FPS gain.
Step 2 — Raise Temperature Limit to Maximum
The temperature limit slider sets the point at which the GPU begins throttling to protect itself. Raise it to the maximum available value and apply. This prevents the GPU from throttling below your target clocks at temperatures you consider acceptable (below 83°C). You’re not disabling thermal protection — the hardware kill-switch remains active — you’re just widening the operating window.
Step 3 — Core Clock: +25 MHz, then Test
This is the primary performance adjustment. Drag the Core Clock slider up by 25 MHz and apply. Now run a demanding game or benchmark for 10–15 minutes. Watch for: crashes, driver resets (the screen goes black briefly and recovers), or visual artifacts — flickering pixels, coloured static, corrupted geometry, or unusual patterns on screen.
If nothing goes wrong: apply another +25 MHz and repeat. Keep incrementing until you hit instability. When you do — a crash, driver reset, or visible artifact — note the unstable value, then reduce by 10–15 MHz and test again for 15 minutes. That stable lower value is your core clock overclock target.
Typical results: most mid-range GPUs have 50–150 MHz of headroom. An RTX 3060 might stabilise at +100 MHz; an RX 6700 XT often manages +125–150 MHz. If you’ve completed five increments with no instability, keep going — you haven’t found the ceiling yet.
Step 4 — Memory Clock: +50 MHz, then Test
With core clock finalised, move to memory. Memory overclocking has a different failure pattern: instability shows up as visual corruption — glitchy textures, colour banding in specific scenes — rather than outright crashes. Raise the Memory Clock slider by 50 MHz, apply, and test for 10–15 minutes. If stable, add another 50 MHz. Roll back 50 MHz when any artifact appears.
Memory headroom varies significantly by VRAM type. GDDR6X cards (RTX 30/40 series) typically accept 500–1,000 MHz over stock. GDDR6 cards are similar. GDDR5 (older generation) is more conservative — use 25 MHz increments instead. The performance impact of memory overclocking is most visible at 1440p and 4K, and in texture-heavy open world games.
Step 5 — Save Your Profile
Click the Save icon (floppy disk) in Afterburner, then click a numbered profile slot (1–5). Your overclock is now saved. To apply it next session, open Afterburner and click that number — or configure Afterburner to apply it automatically at Windows startup via Settings → General → Apply overclocking at system startup.
How to Test Stability Properly
Passing a 10-minute test during the overclock process is a preliminary screen, not a stability certificate. A proper stability check runs three phases.
Phase 1 — Synthetic benchmark (20 minutes)
Run Unigine Heaven or 3DMark on loop for 20 minutes with RivaTuner’s on-screen display active. These tools apply consistent, repeatable GPU loads specifically designed to expose instability. Watch for any artifact, crash, or driver reset. Monitor that your temperature stays under 83°C sustained.
Phase 2 — Real-game session (30+ minutes)
Load a demanding game — something GPU-heavy like Cyberpunk 2077, Red Dead Redemption 2, or Elden Ring at your normal settings — and play for at least 30 minutes. Real games exercise the GPU differently from synthetic benchmarks. Some instabilities only appear in specific rendering scenarios that synthetics don’t replicate. This phase is the real verdict on whether your overclock is usable.
Phase 3 — Temperature check
After Phase 2, review your peak temperature log in RivaTuner. If you peaked above 85°C sustained, your overclock is technically stable but thermally borderline. Reduce core clock by 25 MHz to bring thermals down — losing a few MHz is worth the headroom. If you peaked at 78–82°C, your overclock is in the ideal range.
If It Crashes: Recovery Protocol
Crashes during overclock testing are expected and don’t indicate hardware damage. They mean you’ve exceeded the chip’s stable limit — which is exactly what you were testing for. Recovery is straightforward:
- Hard reset if frozen. Hold the power button to force shutdown. On reboot, MSI Afterburner’s profile may not load automatically (it failed during the previous session), meaning the GPU reverts to stock speeds. This is the expected safe behaviour.
- Reduce the last adjustment by 25 MHz (or 50 MHz for memory) and test again. This is the fix in 95% of cases.
- If crashes persist after multiple reductions, click the reset button (circular arrow) in Afterburner to return all sliders to default. Test at stock settings. If still unstable at stock, the problem is your GPU drivers — download Display Driver Uninstaller (DDU) from Guru3D, boot into Windows Safe Mode, run DDU to fully remove all GPU driver traces, restart normally, then install the latest clean driver from NVIDIA or AMD.
Step 3 is rarely needed. Most overclock-related crashes resolve at step 2 within one or two adjustment cycles.
Should Beginners Touch Voltage?
No — not for a first overclock.
Voltage is the primary heat driver in GPU silicon. Power scales as the square of voltage, meaning each millivolt increase generates disproportionately more heat than the previous one. A +50mV adjustment for a marginal clock gain creates substantially more thermal stress than the performance benefit justifies. For a beginner still learning the stability-testing loop, adding voltage complexity makes debugging harder and the risk calculus worse.
The more interesting direction, once you’re comfortable with the basics, is undervolting: reducing the voltage while targeting the same or higher clock speed than stock. Less voltage means less heat, which means the GPU can sustain higher clocks without hitting thermal throttle. Done correctly, it delivers better performance and lower temperatures simultaneously. It requires mapping the voltage-frequency curve in Afterburner’s Ctrl+F overlay — not beginner territory, but worth knowing it exists. The takeaway: the goal isn’t always more voltage for more clocks. Often it’s less voltage for the same clocks.
Frequently Asked Questions
Is GPU overclocking safe?
Yes, when you stay within temperature limits. Every modern GPU has thermal throttling and hardware shutdown protection built in — it will slow itself down before overheating and cut power before permanent damage. The real risk is instability (crashes during testing), not hardware failure. The incremental testing approach in this guide is specifically designed to catch instability before it compounds.
Does overclocking reduce my GPU’s lifespan?
Minimally, if you stay under 83°C. The concern is electromigration — transistor degradation that accelerates above 90°C. A well-managed overclock at 78–82°C will have negligible lifespan impact. Chasing aggressive clocks that sustain 88–92°C is where longevity trade-offs become real. The simple rule: keep it cool and it will last.
Does GPU overclocking void my warranty?
Most GPU manufacturers don’t explicitly void warranties for overclocking, but damage caused by overclocking isn’t covered. In practice, manufacturers know overclocking happens and rarely attribute standard card failures to it. If you’re running an aggressive voltage overclock that thermally destroys a component, that’s not covered. Staying within temperature limits and leaving voltage alone largely removes this concern.
How do I know my overclock is actually stable?
When it passes all three stability phases — 20 minutes of synthetic benchmarking, 30+ minutes of real gaming, and a post-session temperature check under 83°C — with no crashes, driver resets, or visual artifacts. Running the same test across two or three separate gaming sessions gives you higher confidence that it’s truly stable under varied conditions.
Sources
- MSI. “MSI Afterburner — GPU Overclocking Tool.” MSI, accessed March 2026. https://www.msi.com/Landing/afterburner/
- Unigine. “Heaven Benchmark 4.0.” Unigine, accessed March 2026. https://benchmark.unigine.com/heaven
- UL Benchmarks. “3DMark.” UL Solutions, accessed March 2026. https://benchmarks.ul.com/3dmark
- XDA Developers. “How to Use MSI Afterburner to Overclock Your GPU.” XDA, 2025. https://www.xda-developers.com/msi-afterburner-guide/
I've been playing video games for over 20 years, spanning everything from early PC titles to modern open-world games. I started Switchblade Gaming to publish the kind of accurate, well-researched guides I always wanted to find — built on primary sources, tested in-game, and kept up to date after patches. I currently focus on Minecraft and Pokémon GO.