To stress test a GPU, close other programs, open FurMark, set your monitor’s native resolution, and run the test for 20 to 30 minutes while watching temperature. GPU usage should reach about 99%. Temperatures of 60 to 85°C are normal, and 90°C or more means overheating. If it finishes without a crash, glitches, or a shutdown, the card is stable.
- A stress test pushes the GPU to full load to expose instability.
- Run FurMark for about 20 to 30 minutes, not hours.
- Watch temperature and usage the whole time.
- A crash, glitches, or shutdown means something needs fixing.
What a GPU Stress Test Does
A stress test makes the card work as hard as it can, so problems show up now instead of in the middle of a game. Games load the GPU unevenly, with quiet moments and heavy scenes. A stress test removes the quiet moments and holds the card at its limit, which is why weak cooling, an unstable overclock, or a marginal power supply shows up quickly.
People run one for three main reasons: to check a new build or a second-hand card, to confirm an overclock or undervolt is stable, and to see whether a cooler or new case airflow is doing its job. Corsair’s FurMark guide describes it as a burn-in that makes a card run hotter than you’d expect, so treat the numbers as a worst case, not a typical gaming temperature.
How to Run FurMark
1. Get the PC ready
Update your GPU driver first, so you’re testing the current one. Close browsers, launchers, and anything else that touches the GPU, and save your work in case a crash restarts the PC. Make sure the case fans are running and the vents aren’t clogged with dust, or you’ll be measuring a dirty case, not the card. Note your idle temperature so you have a baseline to compare against.
2. Install FurMark and a monitoring tool
Download FurMark from the official FurMark page. It’s free and small. FurMark shows some stats itself, but a monitoring tool such as HWiNFO or MSI Afterburner lets you log temperature, clock speed, fan speed, and power draw over the whole run, which is far more useful than glancing at one number.
3. Choose the settings
Pick the API (OpenGL or Vulkan) and set the resolution to your monitor’s native one. Leave anti-aliasing at the default unless you have a reason to change it. If FurMark offers an on-screen overlay, turn it on so you can watch the stats live.
4. Start the test
Click Run for the stress test, or Benchmark for a scored run with a fixed length. The stress test is the one to use for stability checking, because it keeps going until you stop it. The card will ramp up within seconds, and the fans will get loud. That’s expected.
5. Watch and stop on time
Watch temperature and GPU usage for the next 20 to 30 minutes. Temperature should climb for the first few minutes and then level off. If it keeps rising with no plateau, stop the test. Don’t leave it looping for hours, since twenty minutes or so is enough to see how the cooling copes, and longer runs only heat the card for longer.
What Numbers Are Normal
| What you see | What it means |
|---|---|
| GPU usage around 99% | The test is loading the card properly |
| Temperature levels off at 60 to 85°C | Normal under this load |
| 90°C or more | Likely overheating |
| Clock speed stays steady | No throttling |
| Clock speed drops over time | Thermal or power throttling |
| Fans ramp up as the card heats | Cooling is responding correctly |
| Passes 30 minutes cleanly | The card is stable |
| FurMark crashes | The card couldn’t handle the load |
FurMark runs hotter than real games, so expect higher numbers than you see in-game. Compare them against what GPU temperature is too high, and don’t panic over the usage figure, since 100% GPU usage is expected here. If your clock speed falls during the run while the temperature sits at its limit, the card is throttling itself to stay safe, which points to a cooling problem rather than a faulty card.
Warning: stop the test if temperatures keep climbing past 90°C, the screen shows glitches, or the PC shuts down. A stress test is meant to find problems, not to run a failing cooler into the ground.
What the Result Tells You
Visual glitches during the test, such as flashing lines or stray shapes, point to GPU artifacting, often from heat or an unstable overclock. Reset any overclock to stock and repeat the run. If the glitches stay at stock clocks with normal temperatures, the card may be failing.
A crash or black screen usually means the card, the driver, or the power supply couldn’t cope. Try a clean driver install first. If the PC shuts down completely, especially when the CPU is busy too, suspect the power supply. High temperatures with slow fans are worth checking, so see how GPU fans should behave. A card that runs hot every time it’s tested is also one to watch, as covered in how long GPUs last.
If you changed clocks or voltage, test in small steps: raise the setting a little, run the test, and move on only if it passes cleanly. FurMark isn’t the only option. OCCT and AIDA64 both include GPU stress tests too, and some settings pass one tool while failing another. Whichever you use, finish by playing a demanding game for a while, since that’s the load you actually care about.
On the CPU side: a CPU stress test.
Frequently Asked Questions
How long should I stress test my GPU?
About 20 to 30 minutes is enough for most checks. Running it for hours adds little and just heats the card for longer.
What GPU usage should I see in FurMark?
Around 99%. If it’s much lower, something else is holding the card back or the test isn’t set up correctly.
Is FurMark safe for my graphics card?
It’s a widely used test, but it draws maximum power and heat. Watch the temperatures and stop if they keep climbing or you see glitches.
What does it mean if FurMark crashes?
The card couldn’t handle the load. Check for an overclock, high temperatures, or driver problems before assuming the hardware has failed.
