Graphics card

How to test a GPU before buying — full guide

Testing a GPU before buying is critical, especially for used cards that may have been used for mining or 24/7 gaming. The GPU is one of the most expensive and at the same time most vulnerable components: overheating, fan wear, video memory chip degradation and broken core transistors may only show up under load. This checklist will help you test a graphics card step by step, from a visual inspection to a full stress test.

The quick check takes 5–10 minutes with no software. Inspect the GPU physically: check the warranty seals on the screws (missing seals mean the card was opened, possibly repasted or repaired). Spin the fans with your finger — they should spin freely, with no wobble or squeak. Inspect the back of the PCB: darkening, burn marks and bulging capacitors indicate heavy use. Open Device Manager → Display adapters — the card should be detected without yellow triangles. Test all video outputs: connect a monitor to each HDMI, DisplayPort or DVI. Run the online GPU test on FullTestPC — it will show FPS, stability and possible dips under load.

The medium check takes 10–15 minutes with monitoring. Install GPU-Z — a free tool to read GPU parameters. Match the GPU model, shader count, core and memory frequencies, memory type and size to what is claimed. Install HWMonitor or MSI Afterburner for temperature monitoring. At idle, GPU temperature should be 30–45°C, up to 55°C on laptops. Run a demanding game or 4K YouTube video for 5 minutes. Watch the temperature: if it exceeds 85°C and the GPU starts dropping frequencies — that is throttling. Watch for artifacts: colored squares, stripes, flickering, color distortion — all are signs of video memory or core issues.

The full check takes 30–45 minutes and requires specialised tools. Run FurMark or Unigine Superposition Benchmark — they create maximum graphics load. Run the test for 10–15 minutes and watch the temperature and FPS graph: it should be flat, with no sharp drops. If FPS drops after 1–2 minutes — it is overheating. If artifacts appear — the video memory or core is faulty. For Nvidia cards, run MATS (Memory Testing Software) — it tests each video memory chip separately and finds broken blocks. For a full driver cleanup, use DDU (Display Driver Uninstaller) — remove old drivers and install fresh ones from the Nvidia or AMD website.

Common GPU issues to watch for. GPU core failure shows up as artifacts and crashes, fixed only by replacing the card or reballing (expensive). Bad video memory shows up as colored squares and flickering, especially noticeable at high resolutions. Fan wear — squeak, wobble, stopping under load — is fixed by replacing the fans. Burned power circuits on the board show up as instability and require card replacement. Cards after mining often have worn fans and degraded thermal paste, but the core itself may be fine — such cards can be taken with a 20–30% discount, but only with a full stress test. If all tests pass without artifacts or dips — the GPU is healthy.