My gaming PC began crashing about a week ago in Valorant, Counter-Strike, and Red Dead Redemption 2, sometimes while loading. The monitor loses signal, audio becomes distorted or stops, and the GPU fans may ramp to full speed. The display never returns, so I usually have to force the system off; once it restarted by itself.
The system runs Windows 11 Pro with a Ryzen 7 5700X, 32 GB of RAM, an ASUS PRIME A520M-A II motherboard with BIOS 3621, an ASUS RTX 5070 12 GB, and an LC Power 700W PSU rated at 58.3A on the 12V rail. The monitor is an ASUS TUF VG27AQL5A.
Event logs repeatedly show LiveKernelEvent 141 and bug check 0x116, indicating that Windows lost the GPU or could not recover its driver. The latest test showed a 73°C core temperature, but telemetry froze at exactly 64°C, 41% utilization, and 149.27W around the failure, so the logged readings may have become stale. Hotspot, memory temperature, and transient power behavior have not been recorded.
I have clean-installed several NVIDIA drivers, disabled the overlay and automatic tuning, restored Control Panel defaults, tested with zero GPU and VRAM offsets, limited the card to 80% power and 200W, changed from DisplayPort to HDMI, and forced PCIe Gen 3 instead of Auto. I also cleaned the system and visually checked the power connector, but I have not swapped the GPU, PSU, adapter, or motherboard. The PC was purchased new in November 2025 and is still covered by the seller's warranty.
What would be the most useful next isolation step? What additional evidence could distinguish a defective GPU or overheating hotspot from a PSU, 12V-2x6 adapter, motherboard, or software problem?
3 Answers
Inspect the GPU’s 12V-2x6 power connection carefully. A connector that is not fully seated, has a recessed or loose pin, or is bent or heat-damaged can cause a black screen followed by full-speed fans because power delivery becomes unstable. Make sure the plug is completely inserted and the cable is not sharply bent right at the connector. If the PSU uses separate PCIe leads or an adapter, verify that each connection is secure and that the PSU model is known to support the card properly.
The 141 and 0x116 events are consistent with a GPU timeout, but they do not identify the failed part by themselves. Preserve the minidumps and LiveKernelEvent details, then test methodically if possible: try the card in another known-good system, or test another known-good GPU in this system. A known-good PSU with suitable native cabling would also be a strong test, especially because the exact LC Power model is unknown. If no replacement parts are available, the repeatable crashes across several drivers, inputs, games, and PCIe settings are already useful warranty evidence, so avoid opening the card and consider an RMA.
Log more than the core temperature with HWiNFO or similar software while gaming: GPU core, hotspot, memory temperature, power, clocks, and fan speed. A core temperature around 73°C does not rule out a very high hotspot or memory temperature. If the hotspot is approaching its protection limit, the card may shut down and leave the fans running. Since the readings appear to freeze before the crash, the log may also help show whether the GPU stopped responding rather than simply overheating.
A 5070 Ti owner reported a core temperature in the high 80s accompanied by a hotspot close to 110°C, which pointed toward a cooling or thermal-contact issue. If your hotspot is similarly high, warranty service is safer than attempting to repaste the card.