Fine-tuning fan curves is one of the highest-impact, zero-cost optimizations you can make to a gaming PC. A well-configured curve keeps CPU and GPU temperatures stable during long sessions, reduces the constant speed oscillation that makes some builds sound like they are breathing, and lets you run fans quietly when idle without thermal throttling under load. The goal is not the lowest possible temperature — it is a predictable, smooth response that matches how heat actually builds up in your case. Priority order: keep the GPU under control first (it is your biggest heat source in games), then CPU, then case airflow, then acoustic polish.
This guide covers BIOS-based CPU fan curves, GPU fan curves in MSI Afterburner or AMD Adrenalin, and case fan control via the motherboard BIOS or software like Fan Control. Everything assumes a standard Windows 11 setup as of 2026.
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Our Top Picks
Top 3 picks at a glance
Quick Recommended Settings
| Setting | Recommended value | Why |
|---|---|---|
| CPU fan curve (BIOS) | 40% at 50°C, 60% at 70°C, 85% at 80°C, 100% at 90°C | Keeps most CPUs out of throttle territory while staying quiet under light load |
| Fan curve response / ramp rate | Slow or “normal” mode, 2–4 second smoothing | Prevents the annoying speed-up/slow-down cycling caused by momentary load spikes |
| GPU fan curve (Afterburner) | Flat 30% until 50°C, then 40% at 60°C, 70% at 75°C, 100% at 85°C | A flat idle section stops RPM fluctuation; GPU fans stay silent in menus and light titles |
| GPU fan stop / zero RPM | Enabled below ~50°C | Fans off at idle means a silent desktop |
| Case fans (intake) | Track GPU temperature, not CPU; ~40% below 50°C, 70% at 70°C, 100% at 80°C | The GPU heats the case air in games; GPU-linked intakes react to the real source |
| Case fans (exhaust) | Same curve as intake or 5–10% higher duty | Slightly higher exhaust preserves positive-to-neutral pressure and limits dust |
| Pump speed (AIO, if equipped) | Fixed 80–100% (or 12V / max in BIOS) | Pumps are near-silent at full speed; varying pump speed adds noise and hurts cooling |
| Hysteresis / fan stop delay | 2–5°C hysteresis or 3–5 second smoothing | Stops fans from ramping up and down with every short load burst |
Setting-by-Setting
CPU fan curve (UEFI)
Exact value: 40% duty at 50°C, 55–60% at 70°C, 80–85% at 80°C, 100% at 90°C. On most boards this lives under a tab named Q-Fan Control (ASUS), Smart Fan (MSI), Smart Fan 6 (Gigabyte), or Fan-Tastic Tuning (ASRock). Enter the UEFI with Delete or F2 during boot, open the fan control page, select the CPU_FAN or CPU_OPT header, and switch the mode from “Auto”/”Standard” to “Manual.” Enter the temperature/-duty point pairs directly or drag the graph.
Trade-off: a steeper curve (higher duty at lower temps) buys 2–4°C of headroom at the cost of a clearly audible fan under moderate load. A flatter curve is quieter but lets a hot chip like an i7 or Ryzen 9 sit closer to its thermal limit, which means sustained all-core loads will trigger thermal throttling and lose performance. Use “CPU Package” temperature as the source, not “CPU” socket sensor, if your BIOS lets you choose — Package reacts correctly to real load.
Response mode / ramp rate
Exact value: set the fan mode to “Silent” or “Normal” ramp behavior (rather than “Turbo”/”Full Speed” instant response), and enable hysteresis of 2–4°C if offered. In ASUS Q-Fan, this is the “Fan Speed Lowering Rate” and hysteresis fields; MSI labels it “Fan Stop” and temperature tolerance.
Trade-off: slower response means temperatures overshoot slightly for a few seconds after a load spike — usually harmless, since silicon tolerates brief excursions fine. Instant response removes the overshoot but produces the characteristic “whoosh-cycle” every time you open a game or tab.
GPU fan curve (MSI Afterburner or AMD Adrenalin)
Exact value: In Afterburner, click the settings gear, select the Fan tab, check “Enable user defined software automatic fan control,” then drag the graph points to: 30% at 40–50°C, 40% at 60°C, 55% at 70°C, 70% at 75°C, 100% at 85°C. In AMD Adrenalin (Performance → Tuning → Custom fan control) use the same anchor points. Make Afterburner start with Windows (Settings → General → “Start with Windows” and “Apply overclocking at system startup”) so the curve persists.
Trade-off: keeping the flat 30% section below 50°C means the fans do not spin up in menus, cutscenes, or light esports titles — the biggest acoustic win for most people. The cost is that a card idling at 49°C sits right at the cliff edge; if your card idle-tempers above 45°C (common with high-refresh multi-monitor setups), raise the flat section start to 55°C. Do not set the curve floor below about 25–30% duty on most cards — many fans stall or buzz at very low duty cycles.
Zero RPM / fan stop
Exact value: leave the vendor’s default fan-stop behavior enabled (typically fans fully off below ~50°C) rather than overriding it with a curve that forces a minimum RPM at idle.
Trade-off: zero RPM gives you a genuinely silent desktop, but if you override it with an aggressive custom curve you lose that. If you do want fans always spinning (e.g., a very hot case), set a fixed idle of 25–30% and accept the constant low hum.
Case fan curves (intake vs. exhaust)
Exact value: map front/top intakes to a GPU temperature sensor if your board supports GPU-source control via software (Fan Control app, free, or your motherboard suite), with 40% below 50°C GPU temp, 70% at 70°C, 100% at 80°C. Rear and top exhausts: same curve, or duty 5–10 points higher. Physically, aim for slightly more intake than exhaust for mild positive pressure, which reduces dust ingress through every crack in the case.
Trade-off: linking case fans to CPU temperature (the default header behavior) is wrong for gaming — the CPU is barely warm in most games while the GPU dumps 200–350W into the case. GPU-linked intakes react to the actual heat source. The downside is it requires running third-party software or a board whose firmware accepts a GPU sensor input; otherwise split the difference with a CPU-linked curve that is slightly flatter.
AIO pump speed
Exact value: fixed at 80–100% (or set the pump header to “Full Speed”/12V). Most pumps are inaudible at max; curve-controlling them trades cooling performance for noise you cannot hear anyway. Connect the pump to a dedicated PUMP header or AIO_PUMP; if your board lacks one, use CPU_FAN for the pump and CPU_OPT for the radiator fans so a pump-failure warning still works.
Trade-off: none, really — a fixed pump is nearly always correct. The only exception is a small 120mm AIO where pump noise at 100% is audible; then drop to a fixed 70–80%.
By Hardware Tier
Low-end (budget CPU, single-fan GPU, stock airflow)
Small coolers and 1–2 case fans mean heat accumulates fast. Run the CPU fan more aggressively: 50% at 60°C, 75% at 75°C, 100% at 85°C. Set case fans to a fixed 50–60% rather than a curve — with only two fans, the noise delta between 50% and 80% is small, but the temperature difference is not. Skip fan-stop on the GPU if the card shares a badly ventilated case; a constant 35% keeps hotspots down.
Mid-range (6–8 core CPU, 200W-class GPU, 3–4 case fans)
This is where the main settings table applies as written. Prioritize the GPU-linked intake curve, since the GPU dominates gaming thermals. Enable fan-stop at idle. If your board supports it, set a slight duty offset (exhaust +5%) rather than physically balancing pressure with resistors or fan swaps.
High-end (high-core-count CPU, 300W+ GPU, AIO, 6+ fans)
With a 240mm or 360mm radiator, decide radiator placement carefully: front-mounted intake rads feed the GPU warmer air; top-mounted exhaust rads feed the CPU warmer air. For gaming, top-exhaust radiators usually win because GPU air goes straight out instead of through the rad. Run rad fans on the CPU liquid-temperature sensor (not CPU package temp) with 30% below 35°C liquid, 60% at 45°C, 100% at 55°C — liquid temp changes slowly, so the curve stays smooth. Run all remaining case fans on GPU temp as in the main table. High-end fans often inaudible below 40%, so use 35–40% as the idle floor instead of 25–30%.
Common Mistakes
Fans oscillating up and down constantly
What causes it: a steep curve with no hysteresis combined with a fast response mode. Every brief load spike (loading a level, alt-tabbing) bumps the temp 3–5°C and the fans chase it.
How to check: sit at the desktop or in a game menu and listen — rhythmic ramping every few seconds confirms it. Watch RPM in HWiNFO64’s sensors panel for sawtooth behavior.
What to do: add 3–4°C hysteresis in the BIOS fan page, switch the response/ramp mode to “Normal” or “Silent,” and flatten the low-temperature section of the curve into a plateau (e.g., one duty value from 40–55°C).
How to undo: reset the fan header to “Auto/Standard” mode in the UEFI, or in Afterburner uncheck “Enable user defined software automatic fan control.”
GPU fans never spinning up during games
What causes it: zero-RPM/fan-stop threshold left active with a custom curve that never crosses it, or a curve authored against the wrong temperature scale (some cards report hotspot temp, some report edge temp).
How to check: in Afterburner’s monitoring overlay (enable “GPU temperature” in the Monitoring tab, shown in On-Screen Display), confirm the fan speed percentage tracks the curve while the card is under load.
What to do: verify the curve’s X-axis matches the temp your card reports at load; add an anchor point at 60°C/50% so there is no gap between fan-stop and your curve start. Update Afterburner if the curve behaves erratically on a newly released GPU.
How to undo: click “Reset” in Afterburner or restore “Auto” in Adrenalin’s fan settings to return to the stock vBIOS behavior.
Case fans tied to the wrong sensor
What causes it: motherboard headers default to the CPU temperature source, which barely moves during gaming while the case cooks from GPU heat.
How to check: run a game for 20 minutes, then compare case fan RPM (HWiNFO64) against GPU temperature — if RPM stays flat while GPU temp climbs, the fans are CPU-sourced.
What to do: install the free Fan Control application, add a “Mix” or “Graph” sensor bound to GPU temperature, and assign your intake/exhaust headers to it. Set it to launch with Windows. Alternatively, some UEFI fan pages (newer ASUS and MSI boards) allow selecting a GPU thermal source directly.
How to undo: delete the sensor assignments in Fan Control, or return the header to “Auto” in the UEFI.
Radiator fans controlled by CPU package temperature
What causes it: CPU package temperature spikes violently (20°C jumps in milliseconds) on modern high-core chips with fast boost behavior, even though the water loop absorbs them.
How to check: if rad fans roar briefly every time you launch or close a program while liquid temp (check “Coolant” or “Liquid” sensor in HWiNFO64) barely moves, you are controlling off package temp.
What to do: switch the rad fan headers’ source to the coolant/liquid temperature sensor (available in Fan Control, and in most AIO vendor software), and build the curve on the 30–55°C liquid range given above.
How to undo: reassign the header back to CPU source in the BIOS or remove the AIO vendor software’s custom profile.
FAQ
Should case fans be PWM or DC controlled?
PWM (4-pin), always, if your fans support it. PWM control pulses a constant 12V power line with a duty signal, so fans start reliably at low duty (5–20%) and hold low speeds smoothly. DC (3-pin) control lowers the voltage itself, and many fans stall, tick, or refuse to spin below ~40% voltage. Set each header’s mode in the UEFI fan page to match the physical connector — a 4-pin fan forced to DC mode will run erratically at low speeds.
Do I need extra software if I set curves in the BIOS?
Only for GPU-linked case fans. BIOS fan control handles CPU and case headers well, but stock UEFI firmware generally cannot read the GPU temperature sensor. If you want case fans to respond to the GPU — the correct source for gaming — you need Fan Control or your motherboard suite’s GPU-sensor option. GPU fan curves always require Afterburner or Adrenalin, since the BIOS cannot reach the graphics card.
Is it bad to run fans at 100% all the time?
Not electrically — modern fans are rated for tens of thousands of hours at full speed — but it is unnecessary and loud. Fans at 100% move air at their maximum airflow and maximum noise; you gain almost nothing thermally over an 80% setting on most models because the airflow-to-RPM curve flattens near the top. Reserve 100% for the upper end of your curve (80°C+) so it functions as an emergency tier, not a constant state.
Will a custom fan curve void my warranty or damage anything?
No. Fan curves are user-adjustable controls provided by the board and card vendors through their own software. The only realistic way to harm hardware with fan settings is disabling zero-RPM and forcing fans below their stall speed (causing buzz, not damage), or disabling fans entirely — which stock software prevents. If temps look wrong after editing a curve, simply reset the profile or the UEFI header to Auto to restore factory behavior.
