My PC began crashing during light web browsing, possibly with brief space-invader-like artifacts, followed by a blue screen. After wiping the system drive and reinstalling Windows 10, the computer sometimes showed artifacts during POST or boot, although the display looked normal once Windows loaded. However, whenever the NVIDIA graphics driver installed or initialized, the screen went black, the system rebooted, and Windows became difficult or impossible to start.
The system uses a Ryzen 5 3600, EVGA RTX 2070 Super XC Ultra, MSI X470 Gaming Pro Carbon, Seasonic Focus Gold GX-750, 32GB of DDR4-3600 memory running at 3200MHz, and an older Samsung 850 EVO 250GB SSD. Everything is running at stock settings. I have already reassembled and cleaned the PC, reinstalled Windows several times, tried different display cables and monitors, updated the motherboard and GPU BIOS, changed power cables and outlets, tested the other PCIe slot, cleared CMOS, used DDU, tried both old and new drivers, updated Windows, and run MemTest86 for four passes without errors. I have also attempted a VRAM test, although I am not completely confident that I performed it correctly.
I have not been able to test another GPU, PSU, SSD, or computer because I do not have spare parts available. Could the graphics card be failing, and are there any worthwhile checks left before paying a repair shop?
2 Answers
Artifacts appearing before Windows loads, followed by a crash when the NVIDIA driver initializes, point much more strongly to a graphics-card hardware problem than to Windows or the SSD. The old SSD could cause corruption, but it would not normally create artifacts during POST.
Before writing off the card, try using only two matching RAM sticks, load BIOS defaults, force the PCIe slot to Gen 3 instead of Auto, connect the GPU with two separate PSU cables without extensions, and perform one clean driver installation after running DDU in Safe Mode. The most useful test is still swapping in a known-good GPU or testing this card in another computer. A repair shop might bench-test it cheaply. Replacing thermal paste or pads is unlikely to fix failing VRAM or another board-level fault, and oven or heat-gun reflow methods are risky and usually temporary.
The lack of cleaning probably did not prove the exact cause, especially since the system was not extremely dirty, but dust can contribute to higher temperatures and shorten component life. Still, the timing of the artifacts and the failure when the driver activates are much better evidence of a failing GPU than simple dust buildup. Cleaning is worthwhile, but it is unlikely to restore a card with defective memory or damaged circuitry.
The PC was not in terrible condition, although I definitely neglected maintenance longer than usual. I used to clean it monthly, so I may simply have been unlucky this time.
I had already tried the RAM, separate power cables, and DDU steps. I will try forcing PCIe Gen 3. Unfortunately, I do not have another system or graphics card, so I may need to ask a shop to test it. Do GPU repairs usually make financial sense, or could the underlying fault be too expensive to fix?