I built this PC about two months ago, and it was initially stable with a per-core CPU undervolt and GPU undervolt after many overnight stress tests. The problems began after I installed a TOPPING E2x2 OTG audio interface along with its drivers and control software. Since then, I have experienced black screens during Valorant and occasional crashes while the system was idle. After some black screens, the GPU fans ramped up to maximum and I had to force a restart.
Crash analysis showed UNEXPECTED_KERNEL_MODE_TRAP (0x7F) errors involving a double fault or stack overflow. The dumps named nvlddmkm.sys, with FanControl.Service appearing in two dumps and MSI Afterburner in another. I also found VIDEO_ENGINE_TIMEOUT_DETECTED (0x141) live dumps with nvlddmkm.sys and the dxgkrnl/dxgmms2 TDR engine-reset stack.
I suspected FanControl, MSI Afterburner, RTSS, GPU monitoring tools, or the audio-interface software, so I removed those utilities, disconnected the interface, and clean-installed an older NVIDIA driver. The crashes continued, although the pattern changed: the PC generally black-screens once within the first 30 minutes after a cold boot, sometimes with the fans becoming extremely loud, and then runs normally for hours after a forced restart, including during gaming.
I am trying to determine whether this is a driver or software conflict, a PCIe link issue, unstable undervolting, a power-connection problem, or failing hardware. I am especially concerned about the RTX 5090 Founders Edition or the Gen 5 PCIe riser.
System: FormD T1 v2.1 with a Gen 5 riser, ASUS B850-I, Ryzen 7 9800X3D, RTX 5090 Founders Edition, 64 GB Klevv DDR5-6000 CL30, and Corsair SF1000.
3 Answers
The cold-boot-only pattern is worth treating as a hardware or link-training clue. Test the system at default BIOS settings, try forcing the PCIe slot to Gen 4 or Gen 3 temporarily, and compare behavior with and without the riser. If it becomes stable when the card is direct-mounted or the link speed is reduced, the riser, slot signal integrity, or PCIe settings become much more likely suspects.
The first isolation test should be running the graphics card directly in the motherboard without the PCIe riser. A marginal Gen 5 riser or connector can cause display-driver timeouts that look like an NVIDIA driver failure. While doing that, inspect and reseat the 12V power connector carefully, checking that it is fully inserted and showing no heat damage or melting.
Use a proper display-driver cleanup tool in Safe Mode rather than relying only on a third-party uninstaller. Then install a known-stable NVIDIA driver with a minimal installation and leave FanControl, Afterburner, RTSS, and other hardware-monitoring tools off. If the issue continues with the audio interface disconnected, it may have been a trigger or coincidence rather than the root cause.
I already disconnected the interface and removed the related software, but the crashes still happen occasionally, so it may have exposed an existing problem rather than directly causing it.
That is a good point. I have not seen any melting, but I will test without the riser and inspect the connector again.