My Nvidia GeForce RTX 3060 Ventus 3X 12GB OC worked normally for nearly two years, but it recently started losing video output during demanding games. In Counter-Strike 2 and Fortnite, the monitor suddenly reports no signal while game audio continues. The GPU fans stop, and afterward the card sometimes refuses to initialize when I reboot. When it does start, it can run for hours during light use, but heavy loads make it fail again within minutes. The GPU core temperature never exceeds about 65°C, so I don't think the core is overheating. I cleaned and reseated the card, checked its power connection, and tried both PCIe power cables from my Corsair CX750 (2023), with no improvement. Event Viewer only shows the expected display-driver/device error. Could this be a GPU protection feature, a power issue, a hotspot-temperature problem, or something else?
3 Answers
A driver crash is still possible, so a clean driver reinstall with DDU from Safe Mode would be worth trying if you can get the card to initialize. However, the fact that the card sometimes fails to initialize at boot and its fans stop makes a hardware or power problem more suspicious than an ordinary driver issue. Testing the GPU in another known-good computer, or testing this system with another suitable PSU, should help isolate it.
Check the GPU hotspot temperature rather than only the regular core temperature. GPU-Z can show this while the card is under load. A core temperature around 65°C can still hide a hotspot near 105–110°C, especially if the thermal paste or cooler contact has deteriorated. That can cause a black screen while the game audio continues. If the hotspot is excessive, servicing the cooler and replacing the paste may help, although opening the card could affect its warranty.
Don’t rule out the power supply just because it worked for two years. A failing PSU, cable, connector, or power stage on the card can appear only when the GPU demands more power. The CX750 should normally be enough for an RTX 3060, so simply buying a much larger unit isn’t automatically the answer, but testing with a known-good PSU is useful. Make sure the PCIe connector is fully seated and use a dedicated cable rather than an unsuitable splitter.
I’ll try installing the card in another computer first. This is the only PSU I currently have, and it had been stable for almost two years, so I’d like to test the GPU separately before replacing anything.