Blue Screen Of Death Every Other Day, Memtest Fine

Reversing drivers didn't fix it.
Completely removing all nvidia drivers and reinstalling them fresh didn't fix it.
I remember when the weather got hot last year my GPU went to ***and I removed the back plate and plastered it in thermal paste and it made it run beautifully. Can't really be bothered to do that again yet.
Tried limiting the power to 70% and temp to 70 degrees max and it seemed stable but crashed.
I guess the GPU might need binning but I'll try a full fresh windows install to try and work out if it is just driver related.

Running extreme stress tests doesn't crash the GPU. Literally just FFXI.

Running extreme stress tests doesn't crash the GPU. Literally just FFXI.

4 hours of memtest is far from enough to rule out faulty ram. I would either let it run for up to 24 hours or remove one stick at a time and see what happens. The latter is usually the more sane test as you probably don't want your computer inaccessible for a long period of time

I'm pretty sure it is GPU related though, and I don't think it's the RAM on the GPU either. It's irregular power draw or something I think. Might replace PC with an AMD mini pc and put the 3090 on an Occulink dock and use it for AI that way. This way main PC can't crash if the display is out of the igpu

You change any USB-connected devices recently? Only BSODs I've ever gotten is from USB bus failures, but my GPUs are all AMD.

I'll try unplugging everything but nothing is new.

If you're getting blue screens, are you getting dump files? You can get WinDBG from free on the Microsoft store, run !analyze on the dump, and it can tell you exactly what's causing it.

If you're not getting dump files, but you're set up TO get them (I think by default Win10/11 does small memory dumps on BSODs) then it's more than likely the firmware on your SSD/NVME as it's crashing too quickly to make the dump file.

Regardless WinDBG and a dump file will point to exactly what's happening.

I will try this. Not crashing right now but:
- FFXI instances (x2) only rendering at 720p
- Removed all USB items (doubt it is this...)
- GPU still limited to 70%
- I also turned off Windows UI scaling (was 200% on 4k monitor) and I have a feeling this was causing stress for some reason

Tried everything and still crashing. Formatted PC and blue screened when updating Nvidia driver. Don't know how to use dxdiag but I need to work this out ☹️

Gotta start removing stuff assuming you didn't install to much driver in this timeline

Pop in an older gpu
Take a stick of ram out something

You did only run memtest for a short period
Run it overnight from a USB or something no os

It's 4x 8GB so I'd have to take two out a time.

Crash might have been caused by the Windows updater, keeps failing over and over to install the newest cumulative update.

Still have a problem

This is the most I can find:


It says it's being caused by Windows and audio driver (despite fresh formatting). I guess it is the SSD so I'll try update firmware or clone it to another I have lying around I guess.

Are you using Atmos?

I ran into a ton of problems with graphics drives and atmos when I set up my new system. I think it was an MSI problem (dynamic IRQ assignment) (I was also dealing with EDID issues on top of it - and there might have been a 3rd issue too, took me months to get it all settled)

No and never changed any hardware for years of playing without problem. I'm fairly confident it is actually the nvme SSD now, I will need to get another one and copy over somehow.

If I didn't need an nvidia card for cuda or if it had an occulink port I would buy one of these: https://www.corsair.com/us/en/p/gaming-computers/cs-9080003-na/corsair-ai-workstation-300-amd-ryzen-ai-max-395-processor-amd-radeon-8060s-igpu-up-to-96gb-vram-128gb-lpddr5x-memory-4tb-2tb-2tb-m2-ssd-win11-home-cs-9080003-na

Transferring my SSD over to new one now. Suppose this will help conclude if it was that.

What program should I use to stress and health check the old SSD? Can put it in an enclosure.

Changed NVME, it's not the SSD. £80 to find that out... what next.

All I see is
SYSTEM_SERVICE_EXCEPTION
0x0000003b

Tells me nothing.

Caused by driver: afd.sys
Caused by address: afd.sys+21353
File: Ancillary Function Driver for WinSock

Just had an issue, turn out I had to reset the CMOS battery. If you dont know how to do, just remove the power cable of your PC. Hold power for 30 seconds, then release; replug, and restart.

I am not trying to be adamant this will fix the issue, but cant hurt trying.

This bit:File: Ancillary Function Driver for WinSock
Seems to say the issue is a driver issue related to networking. I have no networking card, just ethernet cable and bluetooth dongle.

If you cloned the SSD, this is a perfect opportunity to rule out software entirely. Create a USB with the latest driver installers direct from the manufacturer of your equipment- cpu, gpu, motherboard, network etc. Wipe the new SSD and do a clean OS install the fresh drivers and start adding apps. If the problem still happens, you can be confident it’s hardware. If it disappears, you’ve fixed it.

Running a system for long stretches without rebooting (only sleeping) can make software instability more likely, so starting fresh will help eliminate that as a factor.

If it turns out to be hardware, and you aren't approaching 90C in your CPU or GPU at any point, alternate testing one RAM stick in the correct slot (look this up per your motherboard manufacturer's documentation), the next likely culprit is the power supply, if the power supply is failing it could give the same symptoms of pulling a component temporarily while the computer is on, which is never good. But start with the clean install if you haven't— it’s the fastest way to split the problem into software or hardware.

If you've changed any BIOS settings, reset everything to default in the BIOS before during a clean install.

Yeah but I have also reformatted it with windows entirely since then. If it was a software/driver issue it was present before and after formatting it. In hindsight since I flashed the bios (a long time ago) I'm not sure if I re-set the RAM and the timings on that look ***. I want to run DDR4 at 3600 and it's on 1.2V atm but some say it must be 1.35V, though it has been a long time since I flashed the bios and I only started getting these crashes 6 weeks ago.

This thread is 99% snake oil selling, 1% "try everything until something magic happens".

Motherboard dying?

Did you replace your ram yet?

I'm 100% certain now it wasn't the SSD. I'm 90% certain it's not the RAM either. I really think I'm just going to have to build something new at this point, it is 5 years old. By the time I buy new RAM I'd say *** it and build something new. Another £100 to find out if it is the ram isn't worth it, I'd rather just move to to ddr5. At least the SSD I can put in my laptop.

Anyway, I decided to cut the water tubes on my 3090 I pulled from an external GPU. I'm going to replace these and put it all into an Occulink dock then put an Occulink adapter into the pcie slot. It could still be the GPU but I'm keeping my fingers crossed. Worked fine for me for over 2 years.

What kind of CPU are you running again?
Include the model

10700k which served me well. These days it's running hot. For £40 it might be worth swapping it to a water cooler but the costs are adding to to the point that I could have and probably should have just built something new.

Fresh paste and some compressed air action.

But it is 5 years old. If it weren't for anything after 12th being a trainwreck it'd definitely be upgrade time.

Also. 16th will be delayed, plus 100% tariff. Better get that cpu working or pull the trigger immediately. Or you'll be paying a grand for mid/high in 2026. If intel doesn't go bankrupt. and it'll run on 1000 watts.

Buy used server cpus. Cheaper and better than their desktop parts. Check your mobo for compatibility.