Short answer: a stopped GPU fan is not automatically broken. Some cards are fanless, while many supported models use a zero-RPM mode and leave their fans off at idle. Check the exact model and its official fan behavior first. If one or all fans stay stopped when the card should be cooling, the temperature keeps climbing, performance drops sharply, the computer shuts down, or you notice grinding, damage, a hot-electronics smell, smoke, or connector discoloration, stop the workload and arrange an inspection.
Stop immediately for smoke, arcing, melting, a burning or sharp electrical smell, discolored connectors, repeated hard shutdowns, or obvious fan-blade damage. Shut the computer down and disconnect wall power if that is safe. Do not run one more game or benchmark to confirm a serious warning.

First, identify the exact graphics card

Start with the manufacturer, board-partner brand, full model, and fan count. The chip name alone is not enough. Two cards built around the same GPU can use different coolers and fan-control rules.

Look up the card's official product or support page. Search that page for terms such as 0dB, Zero Frozr, Zero RPM, fan stop, or passive cooling. ASUS, MSI, and AMD all document versions of low-load fan-stop behavior on supported products. That does not create one universal start temperature. Use the rule written for your exact card.

  • The model officially supports fan stop: all fans being still at desktop idle may be normal.
  • The card is designed without a fan: there is no fan fault to diagnose.
  • The model does not list a passive mode: a stopped fan deserves a closer, non-invasive check.
  • You cannot identify the card: photograph the labels you can see without removing parts, then ask the seller, builder, or repair shop to identify it.

Record the fan pattern instead of forcing a test

With the side panel in its normal position, record what happens at startup, after Windows settles, and during an ordinary task that normally uses the graphics card. A game you already use is enough. Do not install a synthetic stress test just to make the fans spin.

Write down whether none, one, or all fans are stopped. Note the approximate temperature reported by the card manufacturer's supported software, whether performance changes, and whether the display remains stable. A short video through the case can preserve the pattern without putting a hand near moving parts.

  • All fans stop together at idle and start together during normal use: that pattern can match a supported zero-RPM design.
  • One fan stays stopped while the others run: that points away from a normal all-fan idle state. Stop if heat keeps rising or the card shows other warnings.
  • Fans start briefly at power-on, then stop: the startup movement proves they can move, but it does not prove the full control path is healthy.
  • No fan starts when official conditions say they should: end the workload and move to model-specific support or inspection.

Do not use your finger, a tool, or compressed air to spin a fan while the computer is on. Do not manually spin it while the computer is off either. A visual record is safer and more useful than forcing the bearing.

Watch temperature as a trend, not a magic number

There is no single safe fan-start temperature for every card. Room temperature, case airflow, sensor location, firmware, workload, and cooler design all matter. Compare the observed behavior with the exact model's documentation and with its own recent normal pattern.

End the test if temperature keeps climbing without stabilizing, the card throttles, the screen develops visual corruption, the driver repeatedly resets, or the PC shuts down. If colored blocks, flashing polygons, lines, or texture corruption are the main symptom, use the separate GPU artifacts guide. Those signs need a wider graphics and power diagnosis.

Check recent changes before blaming the fan

Record what changed just before the symptom. Useful clues include an official graphics-driver update, waking from sleep, a vendor control-software change, a new case, cable work, a support bracket, cleaning, transport, or an earlier repair. A timing match is a lead, not proof.

Use the card maker's supported driver and fan-control path. Avoid third-party fan utilities, modified firmware, driver-cleaner packages, and copied fan curves as a first response. They can change the evidence and introduce another control layer. Do not roll back a driver unless the manufacturer documents that action for the exact problem and model.

Do one powered-off external check

Shut Windows down, switch off the power supply if it has a switch, unplug the computer, and wait for all lights and fans to stop. Without removing the card, shroud, or heatsink, look through the case for:

  • a loose case cable, support bracket, or foreign object touching a blade
  • a dense dust mat blocking the cooler
  • a visibly cracked, warped, missing, or displaced fan blade
  • scorching, melted plastic, discoloration, or an unusual electrical smell
  • damage or movement after shipping, a drop, or recent internal work

Do not reach into the power supply. Do not disconnect fan leads, remove the shroud or heatsink, lubricate a bearing, solder a connection, or flash firmware. Those steps can create electrical risk, damage the cooler, and affect warranty coverage. ASUS specifically warns against improper disassembly in its current fan guidance.

Separate a fan problem from a wider PC problem

If the whole computer will not power on, the GPU fan is not enough to identify the failed part. Follow the computer no-power checklist and keep power-supply work out of the reader-facing test path. If the PC runs but several components overheat, inspect case airflow and broader cooling rather than treating one GPU fan as the only cause.

A technician may need to distinguish a failed fan assembly, control circuit, connector, board fault, obstruction, case-airflow problem, software-control issue, or damage from a prior repair. That diagnosis depends on the exact card and its condition.

What to save before service

  • manufacturer, full model, serial number, and fan count
  • official fan-stop feature and any model-specific temperature rule
  • whether none, one, or all fans stop at startup, idle, and ordinary load
  • temperature trend, workload, performance change, and shutdown behavior
  • a short video taken safely through the case
  • recent driver, sleep, software, cable, case, cleaning, transport, or repair changes
  • noise, physical damage, odor, discoloration, or visual artifacts
  • purchase record and current card or system warranty

That record gives a repair shop something testable. It also helps protect your warranty by showing that you stopped before dismantling the cooler.

Where Budget Repair fits

Budget Repair can inspect the graphics card's fan pattern, cooling condition, case airflow, recent changes, and wider computer symptoms before recommending a repair path. Start with our computer and laptop repair service to see what we handle.

When you are ready, request an initial assessment and include the model, fan pattern, temperature trend, warning signs, and warranty status. Inspection comes before a confirmed part, final price, turnaround, or promise that a board-level repair is practical.

Editorial sources

Checked August 23, 2026 against ASUS's current graphics-card fan troubleshooting and 0dB guidance, MSI's current Zero Frozr and no-rotation guidance, and AMD's current automatic fan control and Zero RPM guidance. Budget Repair added the model-first classifier, one-versus-all-fans check, non-invasive inspection boundary, stop-use warnings, evidence list, and repair handoff. Current owner reports were used only to confirm search language and recurrence, not as repair authority, a universal temperature threshold, or a reason to stress-test, solder, flash firmware, or use third-party fan tools.