Odd GPU issue
Odd GPU issue
My setup is a preassembled machine I acquired from NZXT in 2019. Back then, I had limited knowledge about computers, so I followed the suggestions on the website. It wasn't too bad at the time—it still works fairly well today, with a 2070 and Ryzen 7 2700. A few months ago, I began experiencing odd issues while playing more demanding games. Sometimes the display would drop out, and the GPU fans would spin up loudly. The only fix was restarting the power supply unit. This happened across various titles, including Armored Core 6, DMC5, Destiny 2, and tasks needing heavy GPU power like AI image creation. I thought drivers were the culprit, but testing both new and older versions didn’t confirm that. I cleaned the PCI slot and heatsink, yet problems persisted. During a test where the fan cooled the system, temperatures stayed under 78°C, so overheating shouldn’t have been an issue. There were no crashes during artificial load tests, even after running for over 15 minutes. VRAM checks also came back clean. My guess is it might be related to the GPU triggering its overcurrent protection, though that seems unlikely since neither the PSU nor GPU shut down completely. My build should have sufficient power, even with a budget 620W PSU. I attempted to cap the power draw in Afterburner, which seemed to help slightly, but I’m unsure if it made a real difference. Any advice would be greatly appreciated. If you need, here’s a summary of my system specs: CPU: Ryzen 7 2700 GPU: Zotac RTX 2070 AMP Extreme Motherboard: MSI X470 gaming plus PSU: SeaSonic M12II 620W (80+ bronze) RAM: DDR4 3200MHz, two 16GB sticks by teamgroup, timing 16-20-20-40. All other parts seem irrelevant.
It might be that your budget Seasonic PSU is malfunctioning. I’d swap it out. The improvement in limiting power consumption suggests the unit is struggling. Continuing to use it could harm your gear. If you replace it with a different one and it doesn’t fix the problems, you can return it if needed.
I checked several forums for similar issues and noticed many different views. Some suggested PSU problems, others pointed to a major GPU component failure. There was even a claim that NVIDIA was deliberately restricting performance by sabotaging cards with faulty drivers. The most likely explanation seems to be a GPU issue, as I saw the same problem on a 3080 with a 1000W PSU. I’m not sure if it’s a thermal concern since the affected area had no sensors, and I’ve already tried reusing the board. I’ll check the PCB for any faults, but I’m not very optimistic about this case.
The guesswork won’t get you very far... it’s wiser to test parts systematically. Because regular users often lack the right tools, most fixes involve swapping components—probably a PSU or GPU. If you’re lucky, you might only need to replace the GPU by repasting it.
I recreated the problem by disconnecting one of the power connectors from the GPU while it was under load. I’ve purchased a replacement PSU and am currently running the graphics card at 75% power. I’ll share more updates once it works.