I replaced my EVGA GTX 1060 6GB with an EVGA GTX 1080 Ti and removed the old drivers with a driver-cleanup utility before installing fresh NVIDIA drivers. Since then, several games that previously ran acceptably now crash shortly after the title screen or within a few seconds of gameplay. ARC Raiders runs normally, but Palworld, Helldivers 2, and Pragmata are affected.
Lowering the card to 80% power and reducing the core clock by 100 MHz makes the games last a little longer, but they still crash—and they crash almost immediately when I return to normal settings. My system has an Intel i5-6500, 16GB of RAM, Windows 10, a GTX 1080 Ti, and an EVGA 600W power supply. A power calculator estimates around 436W of usage.
Could the 600W PSU be failing or insufficient, or is the graphics card itself likely defective? Would an 800W or 850W PSU be a sensible replacement, and what should I check for compatibility besides the required PCIe cables? Event Viewer also reports an NVIDIA display-driver timeout (Event ID 153 involving nvlddmkm) when the games crash. Does that point more toward a power problem or a failing GPU?
3 Answers
The fact that reducing the GPU power limit and clock lets the games run longer is suspicious for a failing card, unstable VRAM, or a power-delivery problem. A damaged used GPU is definitely possible. Test the card with a GPU stress test and a VRAM-focused test, watch temperatures and power behavior, and, if possible, install it in another known-good system. If it fails there too, the card is the likely culprit.
A nvlddmkm timeout means the NVIDIA driver stopped responding and Windows reset it. It can be caused by a defective GPU or VRAM, unstable clocks, overheating, bad drivers, or an inadequate/failing PSU; the event alone does not identify which one. Try a known-stable driver version, confirm the card is seated properly, check all power connectors, monitor temperatures, and test at stock settings. If it only becomes stable with a large power or clock reduction, that strongly increases suspicion of hardware trouble.
The wattage on its own should generally be enough for a 1080 Ti and an i5-6500, assuming the PSU is healthy and has the proper PCIe connectors. However, some older budget EVGA units—especially the W1 series—are poor-quality designs, so the exact model and age matter more than the 600W label. A reputable 650W–750W replacement would be reasonable; 800W or 850W is optional headroom, not automatically required. Check that it is a reputable modern unit with the correct PCIe cables and protections.
The timeout appeared in Event Viewer at the same time Palworld crashed, so I’ll check the card’s temperatures and run a GPU/VRAM test. I’ll also verify the exact PSU model before replacing anything.