I'm troubleshooting a recurring black-screen problem on a fairly new PC. The system has a ZOTAC RTX 5070 Ti AMP Extreme Infinity 16GB, Ryzen 7 9800X3D, MSI X870E PRO motherboard, 32GB DDR5-6000, and an ASUS 850W Gold PSU using a native 12V-2x6 cable. Both monitors are connected: an ASUS 1080p 180Hz VRR display and a ViewSonic 1080p 60Hz display.
The first failure happened while playing GTA V Enhanced. Both monitors lost signal, but the PC remained powered, audio continued working, and voice chat was still active. The GPU was only around 53°C. I had to force the system off, and it initially would not display properly after restarting. Disconnecting AC power from the PSU and discharging the system allowed it to boot normally again. Reliability Monitor recorded LiveKernelEvent 141 followed by 1A8, apparently involving a graphics engine timeout and a DXGKRNL display-path failure. The NVIDIA driver at that point was 596.21.
I updated to driver 610.88 WHQL. Superposition and Heaven completed normally at roughly 286–297W GPU power, with temperatures in the low-to-mid 60s. I also played GTA V for about two hours without another failure. GPU-Z shows the PCIe link correctly switching from x16 1.1 at idle to x16 5.0 under load. Hardware monitoring reports no lane, replay, WHEA, or other obvious PCIe errors, although the PCIe PEX Errors Recovery counter seems to increase mostly while the GPU is idle.
The second failure happened several days later only about five minutes after boot, while the system was idle. Both monitors lost signal again and the GPU fans appeared to ramp to 100%. There was no new LiveKernelEvent dump or obvious NVIDIA, Display, WHEA, or PCIe error; only the expected Kernel-Power and improper-shutdown entries appeared after I forced the PC off. The 12V-2x6 connector looks fully seated with no visible damage, and the GPU has not been manually overclocked or undervolted.
I have disabled Windows PCIe Link State Power Management and set NVIDIA's power-management mode to Prefer maximum performance because the problem seems connected to idle power states. Windows component checks are now clean after SFC repaired files previously. Has anyone experienced this behavior with an RTX 50-series card, and does it point more toward a driver or Windows graphics issue, PCIe power management or Gen5 compatibility, GPU firmware, the card itself, or the PSU/cable?
1 Answer
I’m seeing very similar random black screens with the same CPU and GPU combination. Stress tests pass normally, which makes it especially confusing. In my case it can happen after a week, then disappear for a month, so a short benchmark run doesn’t seem to rule out the problem. My suspicion is currently software-related—possibly Windows or the NVIDIA driver—rather than simple GPU overheating, but I don’t have a confirmed fix yet.
It may be worth checking the installed Windows update version, including KB5121003, and recording the exact NVIDIA driver version when the failure occurs. That could help identify whether the issue follows a particular Windows or driver release.
That lines up with what I’m seeing. I’ve tested the card under heavy load without problems, so the idle failure is the part that makes me suspect power-state or software behavior.