Home › Articles › GPU Undervolting
GPU Cooling Guide

How to Undervolt Your GPU:
Lower Temps, Same Performance

Published June 2026  ·  By TempCore Editorial Team  ·  12 min read

Quick answer: GPU undervolting reduces the voltage your GPU uses at a given clock speed. Less voltage = less heat = lower temperatures, quieter fans, and in some cases better sustained performance. Most NVIDIA cards drop 5–10°C, AMD cards 8–15°C. Performance loss is typically 0–2% — often zero.

Why Undervolting Works

Modern GPUs ship with conservative voltage settings that guarantee stability across thousands of different chips, ambient temperatures, and use cases. Your specific chip may need significantly less voltage than the spec to achieve the same clock speed — but NVIDIA and AMD don't know that about your individual GPU at the factory.

This is called silicon lottery variation: two identical RTX 4080s can require meaningfully different voltages to run at the same frequency. Undervolting finds the minimum voltage your chip needs, rather than accepting the conservative maximum NVIDIA set for the worst-case unit.

What You Actually Gain

  • Lower temperatures: Typically 5–15°C at load, depending on GPU and how aggressively you undervolt.
  • Quieter fans: Lower temps mean the fan curve activates later and spins slower. This is often the most noticeable real-world benefit.
  • Better sustained clocks: Some GPUs thermal throttle under sustained load. With lower thermals, boost clocks stay higher for longer — meaning you can actually gain performance in some scenarios.
  • Lower power consumption: 10–20W reduction is common, which matters for energy costs over time and for laptops on battery.

Tools You Need

  • MSI Afterburner (free) — works on all NVIDIA and AMD cards; the standard tool for both undervolting and overclocking on Windows.
  • AMD Adrenalin (AMD only) — AMD's own software has a built-in Tuning section that can undervolt without a separate app.
  • HWiNFO64 (free) — for monitoring real-time GPU voltage, temperature, clock speeds, and power draw during testing.
  • 3DMark (the free Benchmark tier works) — for stress testing and comparing before/after results objectively.
  • FurMark — pure GPU stress test. Run for 10–15 minutes to confirm stability after undervolting.

Step-by-Step: NVIDIA GPUs (RTX 30 / 40 / 50 Series)

1. Open MSI Afterburner and the Voltage/Frequency Curve Editor

Install and open MSI Afterburner. Click the voltage/frequency curve icon (looks like a small graph, in the top-left panel area) to open the Voltage/Frequency Curve Editor. This is the core undervolting interface. You'll see a graph where the X-axis is voltage (mV) and the Y-axis is clock frequency (MHz).

2. Find Your Current Boost Point

Run a GPU-intensive game or FurMark for a few minutes while monitoring in HWiNFO64. Note two numbers:

  • Your typical boost clock under load (e.g., 2520 MHz on an RTX 4080)
  • The voltage at that clock (e.g., 1050 mV)

These are your baseline numbers. Your goal is to achieve the same (or close) clock speed at a lower voltage.

3. Apply the Undervolt

In the curve editor, find the point on the graph corresponding to your target frequency. You'll move this point to the left (lower voltage) while keeping the same or slightly lower frequency. The standard approach:

  1. Press Ctrl+F in the curve editor to select all points
  2. Use the up/down arrow keys to lower all frequencies together — this isn't what you want here, so undo that
  3. Instead, find the specific voltage point you want (e.g., 900 mV) and drag the curve so that your target frequency sits at that voltage
  4. Everything above that point gets "flat-lined" — you drag the right side of the curve down so it doesn't boost higher than your target

A common starting approach for NVIDIA RTX 40 series:

  • Find the 900–950 mV point on the curve
  • Set the frequency at that point to your target (e.g., 2505 MHz for RTX 4080, which is ~15 MHz below stock boost)
  • Flatten all points to the right of this at the same frequency

4. Click Apply and Test

Click the checkmark (Apply) in Afterburner. The curve takes effect immediately. Open HWiNFO64 and run FurMark for 10 minutes. Check:

  • Does the GPU crash or artifact? → Voltage is too low. Increase by 25 mV increments.
  • Is the clock frequency stable near your target? → Good.
  • What are the temps now compared to stock?

5. Fine-tune and Iterate

After stable FurMark: run 3DMark and compare scores to your baseline. If within 1–2%, you're done. If you want to push further, lower the voltage another 25–50 mV and repeat. Most RTX 40-series cards stabilise somewhere between 850–950 mV at full boost clocks.

6. Save to a Profile

In Afterburner, save this as Profile 1 (or another slot). Enable "Apply at startup" so it persists after reboots.

Step-by-Step: AMD GPUs (RX 6000 / 7000 Series)

AMD Adrenalin Software (Easiest Method)

  1. Open AMD Adrenalin → Performance → Tuning
  2. Select your GPU and set "Tuning Control" to Manual
  3. Under "GPU Tuning", enable Advanced Control
  4. You'll see a Voltage/Frequency slider — reduce it by 10–15% as a starting point
  5. Click Apply and run FurMark to test stability

AMD's automatic undervolting in Adrenalin (the Eco Mode button on newer cards) can reduce power by 10–15% automatically with a single click — try this first before manual tuning.

MSI Afterburner for AMD (More Control)

Afterburner also works on AMD cards. The voltage/frequency curve editor approach is the same as for NVIDIA. AMD RX 7000 series typically shows larger temperature improvements from undervolting than NVIDIA, often 10–15°C.

Safe Starting Values by GPU Family

GPU Family Start Voltage Freq Target Typical Temp Drop
RTX 50 Series 900–950 mV Stock −25–50 MHz 6–10°C
RTX 40 Series 875–950 mV Stock −25–50 MHz 5–10°C
RTX 30 Series 850–925 mV Stock −50–100 MHz 8–12°C
AMD RX 7000 1000–1050 mV Stock −50–100 MHz 10–15°C
AMD RX 6000 1050–1100 mV Stock −50–100 MHz 8–12°C

Starting values — your chip may need more or less voltage. Always test stability after each change.

Is Undervolting Safe?

Yes — undervolting only reduces voltage, which is the opposite of what damages hardware. The risks are zero in terms of hardware damage. The only failure mode is instability (crashes or driver resets), which you resolve by increasing voltage slightly. You cannot damage a GPU by giving it less voltage than it needs — it will simply crash, not fry.

It does not void your warranty in any meaningful way that manufacturers enforce — you're not increasing voltage, power limits, or temperatures beyond spec. Even if it did, simply resetting Afterburner to defaults removes all settings.

Troubleshooting

  • Game crashes after undervolting: Your target voltage is too low for that clock. Increase by 25 mV and retest.
  • Settings don't persist after reboot: Make sure "Apply at startup" is checked in Afterburner's settings.
  • No temperature improvement: The GPU may be power-limited rather than voltage-limited. Try reducing the power limit slider by 10–15% instead.
  • Clock speeds not reaching target: A different bottleneck (power limit, temp limit). Check what limiter is active in HWiNFO64's GPU metrics.

Combining Undervolt with a Fan Curve

For the best result, pair your undervolt with a custom fan curve in Afterburner. The GPU will run cooler, so you can set fans to spin slower — the combo gives you both lower temps and less noise than stock settings. A typical aggressive-quiet curve: stay at 0% until 50°C, ramp to 60% by 70°C, reach 100% at 85°C.