How to safely overclock your GPU for mining: MSI Afterburner, Power Limit and clocks

The short answer

For mining, a card is usually tuned for efficiency rather than pushed up: you lower the power limit (Power Limit), slightly raise the memory clock (Memory Clock) and, if needed, the core clock (Core Clock). The voltage (Core Voltage) stays untouched. The card runs cooler and quieter, and the hashrate stays the same or even goes up.

If you would rather not tune by hand, FlowHash has ready-made overclock profiles: Settings → “GPU overclock”.

Is overclocking safe?

Yes, as long as you leave the voltage alone. Raising core and memory clocks without touching Core Voltage keeps the card within the limits set by its manufacturer: it has temperature and power protection, and the power limit keeps consumption in check.

If the clocks are pushed too far, you will see it right away: on-screen artifacts, a driver crash or a blue screen. That is not damage but a sign of unstable settings, and it only happens while that overclock is applied. Press “Reset” in Afterburner or go back to the previous value, and the card works as before.

What actually harms a graphics card is high voltage and prolonged overheating, especially of the memory. So leave Core Voltage alone and keep an eye on temperatures — that is the whole rule of safe overclocking.

What each slider does

• Power Limit — how much power the card may draw. If your card runs hot while mining, start here. Lowering it to 70–80% reduces heat, noise and your electricity bill; the hashrate drops only slightly — usually a few percent — while the temperature falls by 5–10 °C or more. This is the safest and most useful step.
• Core Clock — an offset to the GPU chip frequency. A typical range is +50 to +200 MHz. On many algorithms the gain is small, but with a reduced Power Limit a higher core clock helps win back some hashrate.
• Memory Clock — the video memory frequency. A typical range is +200 to +1000 MHz depending on the card and memory type. For a number of algorithms this is the main source of extra hashrate.
• Core Voltage — raising the voltage is the genuinely risky kind of overclocking that speeds up chip degradation. Mining does not need it — unless you are an experienced overclocker, leave it alone.

What you need

1. MSI Afterburner — a free tool for tuning graphics cards that works with cards from any brand. Download it only from the official MSI website (msi.com): builds from third-party sites are a common way to slip in malware or a hidden miner.
2. HWiNFO or GPU-Z — to see the hot spot and memory (Memory Junction) temperatures, which Afterburner does not show on every card.
3. FlowHash — the app window shows each card’s hashrate and temperature, and your dashboard shows how much it earns.

Starting points

These are starting points, not exact settings for your model: even two identical cards overclock differently — it all depends on the particular chip and memory. That is why values are found step by step and tested on your own card: do not start with high numbers from someone else’s table.

Step by step

1. Note your baseline. Mine for 10–15 minutes without any overclock and write down the hashrate, temperature and power draw.
2. Lower Power Limit to 80% and click “Apply” (the check mark). After 10–15 minutes compare the hashrate and temperature.
3. Raise Memory Clock by +200 MHz. Then add 100–200 MHz at a time; after each step mine for 10–15 minutes and watch the hashrate, temperature and stability.
4. Stop when the hashrate stops rising or drops. GDDR6 and GDDR6X memory has error correction: at too high a clock the card does not crash but spends time correcting errors, so the hashrate goes down. Go back to the previous value.
5. Optionally raise Core Clock by +50 MHz and then in 25–50 MHz steps, checking the same way.
6. Check stability over several hours. Only then enable “apply overclocking at Windows startup” in Afterburner and save a profile.

Change one setting at a time — otherwise you will not know what caused a gain or a crash.

If something goes wrong

Signs of an overclock that is too aggressive: on-screen artifacts (stripes, flickering, coloured dots), the driver crashing or restarting, freezes, a blue screen, miner errors and restarts in FlowHash. Such a crash does not break the card by itself and only lasts while that overclock is applied: lower the last value you changed by one step or press “Reset” in Afterburner. After a reboot the overclock is not applied unless startup apply is enabled — which is why you should not enable it until the settings are tested.

What not to do

• Do not raise the voltage (Core Voltage). Higher voltage combined with high temperature is the main cause of chip degradation.
• Do not jump to large values. Go up in steps — it is easier to find the limit without constant crashes.
• Do not let it overheat. Watch the memory in particular: on GDDR6X cards it can reach 100 °C and more. Safe temperatures are listed in “Does mining damage your GPU?”. If a card runs hot, lower Power Limit first instead of raising clocks.
• Do not overclock a laptop. Laptop cooling has little headroom, and overclocking adds heat. At most, lower the power limit if the manufacturer allows it.
• Do not stack two overclocks. If an overclock profile is enabled in FlowHash, do not keep your own Afterburner overclock at the same time — the settings will interfere with each other.

Overclock profiles in FlowHash

To skip manual tuning, open Settings → “GPU overclock” in FlowHash (for NVIDIA cards). The profiles are conservative and do not change the voltage:

• Off — FlowHash changes nothing; your own Afterburner overclock stays as it is.
• Eco — quieter and cooler at a small cost in hashrate.
• Balanced — a compromise between hashrate, temperature and noise.
• Max — the highest hashrate. If things get worse with it — crashes or a lower hashrate — switch back to Off.

The miner restarts when you change the profile — that is normal.

Warranty

Overclocking clocks without changing the voltage usually does not void the warranty: Afterburner and similar tools are distributed by the card makers themselves. Terms still depend on the brand and the shop — check them if the warranty matters to you.

Bottom line