You boot up a game, the case lights come on, your RGB software launches, and everything seems fine — until you glance inside the case and notice the graphics card fans are completely still. Or maybe they spin for a second at power-on, stop, and never move again, even while you’re mid-match in a demanding title. This catches a lot of builders off guard because the card often appears to work normally: the monitor has a signal, drivers load, and frame rates look reasonable. The symptom can show up in a few ways — fans that never spin at all, fans that spin only during startup, fans that twitch and stall, or fans that spin constantly on a card that used to be silent.
Here’s the reassuring part first: on virtually every modern GPU, zero-spin (or “semi-passive”) cooling is a designed feature, not a fault. Most cards from the last several years keep their fans off entirely until the GPU core reaches roughly 50–60°C, which means idle fans are usually exactly what should happen. The situation becomes urgent only when fans stay motionless while temperatures climb past 70–75°C under load, or when you see thermal throttling — sudden frame drops, artifacting, or driver timeouts. That’s a real risk of long-term damage, so it’s worth fixing promptly. Everything else on this list is safe to troubleshoot at your own pace with basic tools.
Likely Causes
| Cause | How to confirm | Fix difficulty |
|---|---|---|
| Zero-RPM mode is active by design | GPU temp is below 50°C in monitoring software while fans sit still | None — behavior, not a fault |
| Fan curve set to silent or 0% minimum in software | Fan speed reads 0% in MSI Afterburner or the card’s own utility under load | Easy |
| Card is starved of PCIe power or seated poorly | Both 8-pin/12V-2×6 connectors fully clicked in; card retested after reseating | Easy |
| Fan header cable inside the shroud is loose or unplugged | Visible after removing the shroud; fans worked before a recent teardown | Moderate |
| Fan bearings are worn or a blade is physically blocked | Card powered off, fans nudged by finger feel gritty, wobbly, or snag | Moderate |
| Failed fan controller or dead fan motor | 12V confirmed at the fan header, fan spins when fed from another source | Hard — warranty/RMA territory |
Fix 1: Verify It’s Actually Zero-RPM Mode
What causes it
Nearly all modern cards use semi-passive cooling. The BIOS on the card holds fans at 0% until the GPU core crosses a threshold, typically 50–60°C, to eliminate noise at idle and during light desktop use.
How to check
Open GPU-Z (Sensors tab) or MSI Afterburner’s monitoring panel and watch the core temperature and fan speed together. Run a light load — a browser video or a short benchmark — and watch the temperature climb. If the fans kick in once the core passes roughly 55°C, nothing is wrong.
What to do
Do nothing, but reframe your test: judge fan behavior under load, not at idle. If you want a proper load test, run a 10–15 minute session of a demanding game or a stress loop and watch whether fans engage before the core hits 70°C.
How to undo it
Not applicable — this is factory behavior. The only way to change it is a custom fan curve (see Fix 2).
Fix 2: Reset or Correct the Fan Curve
What causes it
A custom fan curve saved in MSI Afterburner, EVGA Precision X1, or the card vendor’s utility (ASUS GPU Tweak III, Gigabyte Control Center, MSI Center) can pin minimum fan speed to 0% or use a user mode that never ramps. A driver update can also silently reset the profile to “User Define” with a bad curve.
How to check
Open your fan utility and look at the fan tab. In MSI Afterburner, click the gear icon, open the Fan tab, and check whether “Enable user defined software automatic fan control” is ticked and where the curve’s leftmost points sit. Also check whether the manual fan slider (%) is parked at 0.
What to do
In Afterburner: open the Fan tab, drag the curve so 40% speed corresponds to about 50°C and 70–80% corresponds to 75°C, click Apply (the checkmark), and confirm fan RPM rises in the hardware monitor. In vendor utilities, switch from “User Define” to “Auto” first to test factory behavior, then rebuild a sane curve if you want customization. Click Reset (the circular arrow) if you suspect a corrupted profile.
How to undo it
Click Reset in Afterburner or switch back to Auto mode in your vendor utility. Custom profiles live in the app, not on the card, so a reset always restores default behavior.
Fix 3: Reseat the Card and Verify PCIe Power Cables
What causes it
A partially seated card or a loose 8-pin / 12V-2×6 connector can leave the card running in a low-power fallback state where fan control behaves erratically or the card limits itself before fans ever trigger.
How to check
Power off, flip the PSU switch off, and press the case power button for 5 seconds to drain residual charge. Check every power connector on the card’s top edge — give each a firm tug-test. A properly seated connector should not pull out by hand. Visually confirm the card’s gold PCIe edge connector is fully in the slot and the retention bracket is latched.
What to do
Release the PCIe slot latch, remove the card, inspect the edge connector for debris, and reinsert it squarely until the latch clicks. Reconnect power cables until each clicks. If your PSU allows it, use separate native PCIe cables per connector rather than a single daisy-chained cable on high-draw cards.
How to undo it
Nothing to undo — reseating restores the intended configuration. If you swapped cable arrangements and want the old one back, just reconnect as before; there’s no software state involved.
Fix 4: Update or Clean-Reinstall the GPU Driver
What causes it
Corrupted driver installs can break fan-control communication between Windows and the card, leaving the fan controller at a stale 0% command even under load.
How to check
Open Device Manager (Win+X, then M), expand Display adapters, and check for a warning icon. Note the driver version under the Driver tab. If fans stopped working immediately after a driver update, that timing is your clue.
What to do
Download the latest driver from your GPU vendor’s site, then run a clean installation: in the NVIDIA app or AMD Software, choose the custom/clean-install option that resets settings. For stubborn cases, run Display Driver Uninstaller (DDU) in Safe Mode (Shift+Restart from the Start menu power icon, then Troubleshoot → Advanced options → Startup Settings → Safe Mode), remove the driver, and reinstall fresh. Afterward, recheck that fan control software regains RPM readouts.
How to undo it
Roll back via Device Manager → Display adapters → Properties → Driver tab → Roll Back Driver, or reinstall the previous driver version from the vendor’s archive. Re-apply your fan curve afterward, since clean installs wipe saved profiles.
Fix 5: Check the Fan Headers Inside the Shroud
What causes it
On cards with two or more fans, each fan plugs into a small header on the card’s PCB. After a teardown, cleaning, or thermal pad replacement, a header can be left half-seated, killing one or both fans.
How to check
If fans died right after the card was opened, this is almost certainly it. Otherwise, remove the card and take off the shroud screws (usually Torx T6 or Phillips PH0, 4–8 screws). Confirm each fan’s small 2–3 pin plug is fully seated on its PCB header.
What to do
Push each connector fully home — they seat with a light click and sit flush. While you’re inside, briefly spin each fan by hand to check for grinding or snags, and clear dust from the blades with compressed air held short bursts.
How to undo it
Nothing to undo, but note that opening the shroud may affect your warranty — check your manufacturer’s terms before disassembling. Reassemble in reverse order and torque screws just snug, not tight.
Fix 6: Free a Physically Blocked or Seized Fan
What causes it
Cable routing pressed against blades, dust clumps, peeled case foam, or worn bearings can stop a fan mechanically. Fans that twitch but won’t turn are classic bearing failure.
How to check
With the PC powered off and the PSU switched off, spin each fan gently with a finger. It should coast smoothly and silently. Grittiness, scraping, side-to-side wobble, or hard stops indicate mechanical failure. Also inspect the blade path for stray cables (especially front-panel or RGB wires) and dust bricks.
What to do
Clear any obstruction and reroute cables with zip ties or velcro so nothing sits within a centimeter of the blades. Blow out dust with compressed air in short bursts while holding the fan still so it doesn’t over-spin. If the bearing itself is worn, the fan needs replacement (see Fix 7) — lubricating bearings is a temporary stopgap at best.
How to undo it
Not applicable — clearing an obstruction is strictly beneficial. If you rerouted cables and prefer the old layout, just ensure clearance before restoring it.
Fix 7: Test the Fan Electrically and Replace It
What causes it
A dead fan motor or a failed fan controller on the card. Individual GPU fans are usually 4-pin, 5V or 12V units with proprietary mounting and connector pinouts.
How to check
If one fan spins normally and an identical twin on the same card doesn’t, the dead fan (or its header) is at fault, not the controller. With a multimeter, you can confirm ~12V at the header pins under load — voltage present plus a motionless, unobstructed fan means the motor is dead. On some cards you can swap the two fans’ connectors: if the fault follows the fan, the fan is bad.
What to do
First, check warranty status — most GPU makers (ASUS, MSI, GIGABYTE, Sapphire, etc.) cover fan failures for 2–3 years, and an RMA is free. If you’re out of warranty, source a matching replacement fan by the card model and revision, photograph the original connector pinout before unplugging, and swap the unit. Interim safety measure: set a temporary aggressive fan curve on the working fan(s) and cap clocks with a utility to keep temps under 80°C.
How to undo it
Reinstalling the original fan reverses the change. There’s no software state to revert, but recheck your fan curve afterward, since long downtime may have reset saved profiles.
If Nothing Works
If you’ve confirmed the card reaches load temperatures above 70°C with fans verified unobstructed, curves reset to auto, drivers clean-installed, and power firmly connected — and the fans still won’t move — the fault is almost certainly on the card’s PCB: a failed fan controller IC, a blown fan supply rail, or board-level damage. At that point, stop troubleshooting. Run an RMA with the manufacturer if the warranty window allows it, including a short video of the symptom and the troubleshooting steps you’ve taken; support teams approve claims faster when the diagnosis is documented. Out of warranty, a reputable board-repair shop can often replace a fan controller for far less than a new card, though weigh that against repair costs. As a last-resort workaround only, external controllers or carefully spliced adapters can force fans to run at constant speed, but that sacrifices the card’s temperature-based control and should never be a long-term arrangement.
FAQ
Is it bad if my GPU fans only spin under load?
No — that’s the intended zero-RPM design. Fans idling at 0% below roughly 50–60°C reduces noise and dust buildup, and the fans engaging once the core warms up is exactly what should happen.
How hot is too hot before I should worry about dead fans?
Sustained core temperatures above 83–85°C in a well-ventilated case, or any temperature climbing past 90°C, warrants action. Most cards throttle around 83–90°C, which shows up as sudden frame drops before outright damage occurs.
Can I force my GPU fans to spin all the time?
Yes. In MSI Afterburner, enable user-defined fan control and set the leftmost curve point to 30–40%, or drag the manual fan slider to a fixed percentage and click Apply. Be aware this adds constant noise and dust intake for no cooling benefit at idle.
Will a GPU still work if one fan is dead?
Often yes, at reduced performance. The remaining fan and the large heatsink may hold temperatures in check at moderate loads, but sustained gaming will push the card into thermal throttling. Treat a single dead fan as a fix-soon problem, not a stop-using-the-PC emergency, and avoid heavy workloads until it’s repaired.
