A new graphics card, whether fresh from the store or bought used, deserves a proper shakedown before you trust it with a long gaming session or an important render. Stress testing reveals instability, overheating, coil whine, and the artifacting that signals a defective or damaged card, all while it is still returnable. Done sensibly, it is completely safe and takes under an hour.
Why stress testing matters
Factory defects and shipping damage do exist, and used cards may have been thrashed by mining or heavy overclocking. A stress test pushes the GPU to sustained full load so any weakness shows up now, not three weeks later when the return window has closed. You are watching for crashes, visual glitches, excessive temperatures, and unusual noise.
What to monitor and safe limits
| Metric | Healthy range | Warning sign |
|---|---|---|
| GPU core temp | Below ~83°C | Sustained 87°C+ |
| Hotspot temp | Below ~95°C | 100°C+ or large gap to core |
| Memory temp | Below ~95°C | Approaching 100°C+ |
| Visuals | Clean image | Flickering dots, colored artifacts |
How to run the test
Install a hardware monitor to watch temperatures, clocks, and fan speed in real time. Then run a graphics stress tool or a demanding benchmark on loop, or simply play a graphically heavy game for 30-60 minutes. Keep the monitoring overlay visible. Temperatures should climb quickly then plateau, clocks should stay steady, and the image must remain clean. A card that crashes, shows artifacts, or throttles hard is telling you something is wrong.
Check for artifacts and noise
Artifacts are the clearest sign of a failing GPU: stray colored pixels, flickering geometry, or textures that break apart under load. Some coil whine, a high-pitched electrical buzz under heavy frame rates, is normal and harmless, though excessive whine may bother you. If you see artifacts on a new card, stop and arrange a return or warranty claim.
Testing a used or secondhand card
The stakes are higher with a used card, so the testing should be more thorough. Cards that spent a life mining cryptocurrency or running heavy overclocks can have degraded fans, dried thermal paste, or worn memory, none of which is obvious from a photo. Beyond the standard stress test, run several different demanding games across a couple of hours, watch memory temperatures closely since they are a common weak point on heavily used cards, and listen for fans that rattle or fail to spin up smoothly.
Combine the stress test with a visual and physical inspection. Look for excessive dust, damaged or missing screws, bent connectors, and any sign the card has been opened and repasted poorly. A card that stays cool, quiet, and artifact-free through extended real-world gaming is very likely healthy, while one that throttles hard, shows any artifacting, or runs alarmingly hot on the memory should be returned or avoided. Doing this work inside the return or dispute window is what protects you, so never skip it just because the card seems fine in the first few minutes.
Frequently asked questions
How long should I stress test a new GPU?
Thirty minutes to an hour of sustained full load is enough to expose most defects and confirm the card holds stable clocks and safe temperatures. For a used card you are unsure about, a longer session of a couple of hours across different games gives extra confidence before the return window matters.
Is it safe to run the GPU at 100 percent load?
Yes. Graphics cards are designed to run at full load for extended periods; that is exactly what gaming and rendering do. As long as temperatures stay within the safe ranges and the card is stable, sustained full load causes no harm. Modern cards also protect themselves by throttling or shutting down if they ever get too hot.
Bottom line: Stress test every new or used GPU while it is still returnable. Monitor core, hotspot, and memory temperatures, watch for artifacts and instability over 30-60 minutes of full load, and treat any glitching or crashing as a reason to return the card before the window closes.
