GPU fans usually run faster, but black screens appear.
GPU fans usually run faster, but black screens appear.
System details: Power supply Corsair Rm1000x 80 Plus Gold Mobo Gigabyte Aorus Elite AX B650 Rev 1.0 BIOS version F34 CPU RAM 32 (2x16) GB DDR5 Corsair Vengeance CL30 6000 MT/s (first RMA after a failed memtest) GPU Gigabyte Eagle V2 RTX 4070Ti 12 GB (attempted RMA, Gigabyte sent no update) Operating system originally Windows 10 then Windows 11 now Bazzite Case fans Lian Li SL-INF 120mm x5 with wired controller Monitors ASUS TUF 1440p G-Sync compatible, main display 1440p secondary 1080p via DisplayPort NVME Samsung 980 PRO 2Tb Symptoms: GPU cooling fans reached maximum, display froze on both screens. Audio played briefly, sometimes stopped. This happened on both OS versions and also in BIOS without fan speed increase. After resetting BIOS, the machine wouldn’t power on. - Running Windows 10 without GPU for a while didn’t trigger the problem, lasting about 1.5 weeks. I’ve done: - Updated BIOS and drivers, ran DDU twice - Ran DISM and SFC tools multiple times - Replaced RAM and GPU after RMA, RAM passed Memtest86 with no errors, GPU recovered post-test - Reinstalled OS versions, switched to Bazzite - Replaced GPU, RAM, checked CPU paste, cables repositioned on PSU and GPU - Tried swapping PSU, cables, or display cables without success - Ran benchmarks (Furmark, Cinebench, OCCT) for VRAM and 3D adaptive, no crashes. Prior to this, in 2023 I experienced bluescreening issues, possibly linked to a "watchdog violation" and memory management problems. I seldom blue screened since. This might cover what I’ve tried, though it’s not exhaustive.
The blue screen problems largely ceased after I returned the RAM. It's now configured at EXPO 1 with a speed of 6,000 MT/s. The CPU remains in its standard mode.
Have you verified the GPU temperatures? It seems to strongly indicate a GPU issue.
I'd agree with Podkall, begin monitoring temperatures. On Windows, the task manager now displays GPU temperature. On Linux, use lm-sensors for similar info.
The Hotspot reaches temperatures as high as 92C on Furmark. In the game it stabilizes near 70C at the Hotspot. Further testing would be needed.
Running OCCT with full VRAM and Furmark on my GPU in a friend's PC showed stable performance. No crashes occurred, and temperatures stayed below 92°C at peak usage. The average temperature was around 70°C.
I’m leaning toward RAM. Turn off the expo setting and let it run at its default speed to check for improvements. I’ve experienced similar issues before—like in a game (AW2) where incorrect RAM timing caused crashes, which vanished once I adjusted the settings.
EXPO 1 operates at 6000 MT/s, and Corsair evaluates the RAM performance for that speed.
consider adjusting the entire bios to standard defaults, simply to eliminate any configuration issues