When I play graphics-intensive games, the monitor suddenly loses signal a few seconds after the GPU starts working harder. The fans become louder, then ramp up even more, but I can still hear game audio and Discord, so the system itself appears to remain responsive. This makes me suspect a graphics-card or power-delivery issue rather than a complete system shutdown.
I used DDU and tried both an older stable driver and the latest driver. The older driver helped somewhat, but some games still crashed. I also reseated the GPU and one RAM stick. Undervolting the GPU or limiting its power to 70% prevents the problem, at least so far.
The issue did not occur when I tested with integrated graphics using OCCT, although I am not completely sure whether the dedicated GPU was involved. A PSU stress test also completed without a problem. My system has an Intel i5-13600K, an RTX 3070, and a 750W 80+ Gold power supply.
Does this behavior point to a failing or overheating GPU, a loose or faulty display cable, unstable RAM, or a PSU problem? If the PSU were failing, would I expect the entire computer to shut down rather than only lose video output?
3 Answers
The symptoms strongly suggest the GPU is crashing or losing stability under higher load. The fact that lowering its voltage or power limit prevents the black screen is especially telling. A graphics card does not always show obvious artifacts before failing; it can simply stop outputting video while the rest of the system continues running. If possible, test the card in another system or test a known-good GPU in this one.
The PSU cannot be ruled out completely. A 750W unit should normally be enough for an RTX 3070, but a degraded PSU or a brief power spike can cause the graphics card to drop out without immediately turning off the whole computer. A stress test may not reproduce transient spikes, so testing with another known-good PSU would be useful. Replacing or reseating the monitor cable is also a quick way to eliminate a simple signal problem.
Check temperatures while gaming, including the GPU core and hotspot, since an aging card can develop cooling problems even if the average core temperature looks normal. Try enabling V-Sync or setting an FPS limit to reduce power draw. Also test the GPU at default settings with default RAM speed, reseat both memory sticks, and make sure the GPU power connectors are firmly attached.
The GPU temperatures looked fine during testing, and the system does not seem to reboot. I will still try a different display cable and check the power connectors, but the fact that a 70% power limit stops the crashes makes the GPU or its power delivery seem more likely.