My PC suddenly loses power whenever I start a game such as CS:GO. It does not crash to desktop or show a blue screen—the display goes black and the entire system turns off. Everyday use, including browsing, video playback, and Discord, works normally.
The system has an Intel Core i7-4790, MSI H97 Gaming 3 motherboard, GTX 970 4GB, 16GB RAM, a 2TB hard drive, and a 240GB SSD. The power supply is an LC-Power LC6550 V2.2 rated at 550W.
I cleaned the system, checked all connections, reinstalled and downgraded the NVIDIA drivers, checked Event Viewer, and found Kernel-Power Event ID 41 after each shutdown. OCCT's 3D Adaptive GPU test also crashed on GPU #0. Lowering both the GPU core and memory clocks by about 140 MHz in MSI Afterburner changed the behavior: instead of shutting down immediately, the computer lasted around 10 minutes. Capping the power limit was also suggested, but the system eventually crashed even faster.
Does this point more toward a failing power supply, a defective GTX 970, overheating, or another power-delivery problem? What would be the best next test to identify the failing component?
3 Answers
Given the age of the platform and the fact that the PSU has likely been in service for many years, replacement is the most sensible next step. Event 41 only confirms that Windows detected an unexpected loss of power; it does not identify the cause. Temperatures are still worth checking, but a thermal problem usually causes throttling or a shutdown after sustained heat, while a weak PSU can turn the machine off as soon as a game starts. If the replacement PSU does not solve it, stress-test the GPU and check its temperatures and power connectors.
The first thing to check is the power supply. A 550W rating is technically enough for this hardware, but the model's age and quality matter more than the number on the label. A complete instant shutdown under 3D load, combined with normal desktop use, is very consistent with a PSU that cannot handle the sudden load or is failing internally. Try a known-good replacement PSU with the proper PCIe connectors before assuming the graphics card is dead.
The PSU is an LC-Power LC6550 V2.2 550W unit. I had not considered its age and quality being more important than the wattage rating.
The underclocking result is a useful clue. If reducing GPU clocks lets the system run longer, the problem is probably related to power draw or instability under load rather than a Windows driver issue. You can temporarily reduce the GPU power limit to around 70% as a diagnostic, but that is not a real fix. Since the system still shut down—and even crashed faster with the power limit change—I would test with another reliable PSU next, then test the GTX 970 separately if possible.
I tried lowering the power limit, but it actually crashed faster than before. I will prioritize testing with a different PSU.
I also suspect the PSU before the graphics card, so I will replace or borrow a known-good one and test again.