The GPU disconnected from the monitor twice after the Windows 11 25H2 update.
The GPU disconnected from the monitor twice after the Windows 11 25H2 update.
So, my PC decided to auto-update to Windows 11 25H2 back on the 30th. Since then, I've had two separate instances while playing non-graphically intensive games where the GPU spontaneously drops connection to my display. Note in both cases the PC still "worked" for a while: game audio kept playing and as far as I could tell inputs were still being processed (keyboard lights acted as they should). Eventually both times the PC apparently locked up as far as I could tell though, In both cases, no BSOD was generated (verified using BlueScreenView to ensure no crash dumps were created) and no entry in the Event Viewer aside from an Unexpected Shutdown being logged on next boot.
In addition, the second time this happened I did a Win-Ctrl-Shift-B to try and force restart the GPU driver; this had no effect.
Verified Windows isn't corrupt via SFC \scannow. Ensured GPU driver up to date and all pending Windows updates installed. Memory test via memtest86 pending (need a larger USB stick); did a Windows memory diagnostic for two passes in the meantime and both passed. Temps are perfectly acceptable around the board (topping out around 86C for the CPU/GPU and cooler everywhere else).
Note there is one additional headache: I used an LG C2 OLED, which is technically a TV. One issue here is that it has trouble locked onto low-resolution graphical modes; as an example, if I have it on when I start a memory test it runs fine, but if I cycle the display it simply won't detect the resolution mode used.
Latest config is in signature, but I'll list here again:
CPU: AMD Ryzen 9 7900X
CPU cooler: Noctua NH-D15 Chroma Black
Motherboard: Micro-Star PRO X670-P WIFI (MS-7D67)
RAM: 2x Kingston KF552C36-16 (1st EXPO Profile: 5200MHz 36-40-40-80-125)
SSD/HDD: TBD
GPU: PNY NVIDIA GeForce RTX 4070 Ti Vetro Triple Fan
PSU: Corsair 1000W [unsure on exact model, ~2 years old]
Chassis: TBD
OS: Windows 11 x64 25H2
Monitor: LG C2 48"
Sort of hoping this is more a Windows issue then a GPU one at the moment; I have a backup 3060 lying around but the issue is rare enough at the moment where I wouldn't be entirely sure its the GPU for about a month if I went that route. Could also be a RAM stick starting to go; I'll do a memtest86 run once I get a stick delivered that can fit the image. The fact this happened after a feature update makes me HEAVILY suspect Windows though.
EDIT
Updated for the EXPO profile the RAM is currently running at. I also noticed CPU-Z reports the RAM running at 2600 MHz which is half the expected data rate; is that because the doubling isn't counted here, or is that somehow not enabled?
The most recent configuration details are found in the signature, but here they are re-presented:
CPU: AMD Ryzen 9 7900X
Motherboard: Micro-Star PRO X670-{ WIFI (MS-7D67)
RAM: 2x Kingston KF552C36-16 (1st EXPO Profile: 5200MHz 36-40-40-80-125)
GPU: PNY NVIDIA GeForce RTX 4070 Ti Vetro Triple Fan
When discussing troubleshooting threads, it's standard to provide full system specifications. Please format your build details as follows:
CPU:
CPU cooler:
Motherboard:
RAM:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
Include the PSU age along with its make and model. BIOS version for your motherboard at this time.
Sig space measurements may change over time, so these suggestions might become irrelevant for you now. This is why we request users to share their specs directly in the thread.
CPU-Z indicates RAM running at 2600 MHz, which is half the expected rate.
Right-click Taskbar > Task Manager > Performance > Memory; verify if your speed appears as 5200MT/s. If it does, halve that number—you’ll see 2600MHz in CPU-Z’s DRAM frequency section.
Note: For AM5 platforms, a dual-channel setup with DDR5-6000MHz or DDR5-6000MT/s is optimal, especially with low latency.
If you suspect RAM issues, run memtest for 10 passes.
I'll fill in what I can once I get home; not sure on the Case or the exact PSU model, though I know that latter is a Corsair 1000W unit that's about 2 years old.
I'm actually fairly confident this is a GPU issue; if it were anything else (especially RAM) I'd expect to see other symptoms show up. I also have observed one instance of output corruption being drawn to the screen, which is usually a pretty good indication the GPU VRAM is on the way out. The problem is there isn't any good way I know of to actually *prove* that aside from eliminating everything else.
My current plan is to run a memtest86 overnight Friday to prove its not the RAM, then on the next instance of the problem swap in a known good 3060 Ti I have lying around and test that for about a month.
Update delayed notice:
Memtest completed successfully as anticipated. Changed to the 581.94 Beta drivers and *please take care* two weeks without any issues. Planning to wait a bit longer before confirming everything is resolved.