First basic spec:
CPU: Ryzen 7 9800X3D
GPU: Gigabyte RTX 5090 OC Gaming
PSU: NZXT C1200 ATX 3.1 (PA-2G2BB-EU)
MB: MSI MAG B650 TOMAHAWK
OS: Windows 11 x64 Professional
Problem:
PC was bought and built at the end on December 2025, so it's still less than half a year old.
After initial setup everything was working fine, but around a month ago GPU started to behave weirdly. When playing or just hard work with GPU fans started randomly spinning up to 100% with temps being still around 60-70°C (140-160°F [is anyone using F for hardware temp?!]) and if left alone under load for 10-30 seconds (or sometimes immediately) the display goes black and only hard-reset using power button on PC can restore it.
If I stop heavy load during "fans 100%" but before crash, everything returns to normal. I don't remember doing anything with my setup before those issues started.
After further investigation I noticed, that this behaviour is dependent on power draw of GPU, it happens only when GPU eats ~270-280 W or more. But time under load is also a factor, as it crashes during games, but not when running Cinebench 2026, as it has power spikes above 400 W but they are short and full 10 min run finishes cleanly.
Succesful Cinebench 2026.1.0 10 minutes run on GPU
What I tried:
- reseating 12VHPWR cable in GPU and PSU
- removing (but only using Add or Remove Windows feature) and reinstalling nVidia Drivers
- changing 12VHPWR cable for another, new, for the same PSU model
- changing the whole PSU, for a new one with all new cables (same model)
- running Gigabyte Control Center "OC Scanner" (it crashed around 80% complete, when tests started bigger power draw)
- Power Limiting GPU ("max" limit that I can get is 67%, which still exceeds 270-280 W so it's no fix)
- resetting BIOS configuration
If anyone is interested, here is GPU-Z dump when I'm recreating this issue: https://pastebin.com/KJazuU2Y
I've noticed that PerfCap Reason changes from 4 to 2, which apparently is Thermal. Indicating perf is limited by temperature limit. but I can't see any thermal issue, at least on sensors that I can see.
Question:
Is there anything else that I can try/check or this is RMA on GPU?