I bought a used GTX 1660 for $50, but it only works for about 5–10 minutes after installation. Then the display goes black and the entire PC hangs. The only way to get it working again is to remove the card, boot using integrated graphics, run DDU in Safe Mode, reinstall the GPU, and repeat the process. Installing the NVIDIA driver seems to trigger the failure after a few minutes as well. The card reaches roughly 65–70°C under full load and sits around 34°C at idle, so it does not appear to be overheating. The seller used a different card in the video they sent me and has since blocked me. Is there anything else I can test, or is the GPU probably defective?
2 Answers
A card can show normal core temperatures while its memory or VRMs are failing, so 65–70°C does not completely rule out hardware trouble. You could try lowering the power limit and memory clock in a tuning utility, or install an older driver, but those are mostly diagnostic steps. If reducing the clock speeds makes it stable, the card is likely deteriorating and not reliably repairable. If it still fails immediately, replacement is probably the practical option.
The temperatures make ordinary overheating less likely, although a faulty fan, VRAM problem, power-delivery issue, or damaged GPU core could still cause the card to fail once the driver enables full functionality. Try the card in another PC if possible, test a different PCIe power cable and display cable, and check whether the problem occurs with another power supply. If it consistently black-screens only when the NVIDIA driver loads, the card is probably physically defective rather than something DDU can fix.
That makes sense. I’ll try another system and power cable before writing it off, but it sounds like the seller probably gave me a damaged card.
The fact that it hangs the whole computer instead of just losing the display makes a hardware fault more likely. I wouldn’t spend much money trying to repair a seven-year-old card.