Page Fault In Nonpaged Area even though Memtest passes

[MickeyGaby]

Hello,

I had a previous PC that I upgraded but I kept the Windows SSD drive and PSU. It ran for almost 1 month with no problems. Yesterday I wanted to re-install my windows since my almost brand new GPU was giving me a headache by giving me some artifacts in Battlefront II and also some weird bugs in The Witcher 3. As I was uninstalling stuff to make room for the new windows install I needed to restart the PC. So I did and when it came up to the welcome screen it started giving me Page Fault In Nonpaged Area errors. I tried playing around with stuff but I wasn't able to get it to boot to windows again. I've then decided to try and re-install windows. Created a bootable drive and tried to install windows. It kept failing while loading up the Windows Install thing. Managed somehow to install windows 10 but now it keeps failing as soon as it gets to the configuration part before the log-on screen. I've read online that this is probably due to ram. So I tested the following things and these are the results:

- Tested with memtest over 15 hours and 0 errors. I think I had 11 passes when I stopped.
- Different Hard-Drive with windows 7 on it and it still gets Page Fault In Nonpaged Area error when booting at the moment it'd normally show the windows welcome logo.
- Tried re-seating the ram in different places. Same thing.
- Tried re-seating the CPU and re-appling thermal paste. Same thing.
- Tried swapping out the GPU with my old one. Same thing.
- Tried with nothing plugged in besides the SSD. Even without the SSD and only the bootable USB stick.
- Can always boot into Memtest86. Didn't crash once during the 12+ hours it was tested.
- Can not boot into anything windows related. Neither good windows install nor installer.
- Tried with XMP enabled, with XMP Disabled, with an extra boost from 1.35v to 1.4v to RAM. Same thing.
- Tried updating the BIOS aaaaand same problem.

Things that I'm fairly certain are not faulty: HDD,SSD,GPU,RAM.
I'm guessing CPU and Mobo are also fine since otherwise memtest would fail?

System Specs:
CPU - https://ark.intel.com/products/94189/Intel-Core-i7-6800K-Processor-15M-Cache-up-to-3-60-GHz-
RAM - https://www.corsair.com/eu/en/Memory-Size/vengeance-lpx-black/p/CMK16GX4M2B3200C16
GPU - https://www.gigabyte.com/Graphics-Card/GV-N2080WF3OC-8GC#kf
Mobo - https://www.asus.com/Motherboards/X99-A-II/
SSD - https://www.kingston.com/us/ssd/consumer/sv300s3_disco
PSU - https://www.vedantcomputers.com/corsair-vs-series-vs650-650-watt

Thanks.

 

[LinuxDevice]

Artifacts can sometimes be seen when video hardware is operating right at the edge of its abilities in terms of some combination of temperature and power supply. RAM exists not only on the motherboard DIMM slots, but also in the GPU. Assuming the GPU isn't overheating, then it might be that your power supply is under performing.

Even if you can't get Windows running you should be able to test again with memtest86 since it has no reliance on hard drive or operating system. While using memtest86 the GPU itself won't be used (and GPU RAM won't be tested)...it'll instead use a simple framebuffer, but the power consumption will be tiny in comparison to hardware accelerated video. Does memtest86 work even though Windows does not (you didn't specifically say if you tried memtest86 after the Windows install failed...perhaps you only used memtest86 prior to the Windows install attempt)? If memtest86 still works, and if more than one GPU has artifacts, especially if you can tell the fan is working, given that your other GPU also had artifacts it tends to point to the PSU (I assume the prior GPU had lower power requirements, and that the newer GPU in some way perhaps stressed the PSU more).

You can get inexpensive PSU testers (e.g., $20-30), but they only test for outright failure. They don't test under load. I suspect you should just RMA the PSU since it now has issues even with the older GPU. Even better yet, get a second PSU and test with that. It sucks to have two PSUs if both work, but you can consider it useful for future testing if both work...and a great replacement if the old PSU turns out bad. I don't know about the quality of that particular supply, but 650W is probably plenty if something hasn't failed in it.

FYI, RAM going through a memory controller to ordinary user space software is virtual, and can be paged. Physical addresses belonging to hardware cannot be paged or swapped. The memory error might just be a cryptic way of saying there is a memory error in the GPU.

 

[MickeyGaby]

Artifacts can sometimes be seen when video hardware is operating right at the edge of its abilities in terms of some combination of temperature and power supply. RAM exists not only on the motherboard DIMM slots, but also in the GPU. Assuming the GPU isn't overheating, then it might be that your power supply is under performing.

Even if you can't get Windows running you should be able to test again with memtest86 since it has no reliance on hard drive or operating system. While using memtest86 the GPU itself won't be used (and GPU RAM won't be tested)...it'll instead use a simple framebuffer, but the power consumption will be tiny in comparison to hardware accelerated video. Does memtest86 work even though Windows does not (you didn't specifically say if you tried memtest86 after the Windows install failed...perhaps you only used memtest86 prior to the Windows install attempt)? If memtest86 still works, and if more than one GPU has artifacts, especially if you can tell the fan is working, given that your other GPU also had artifacts it tends to point to the PSU (I assume the prior GPU had lower power requirements, and that the newer GPU in some way perhaps stressed the PSU more).

You can get inexpensive PSU testers (e.g., $20-30), but they only test for outright failure. They don't test under load. I suspect you should just RMA the PSU since it now has issues even with the older GPU. Even better yet, get a second PSU and test with that. It sucks to have two PSUs if both work, but you can consider it useful for future testing if both work...and a great replacement if the old PSU turns out bad. I don't know about the quality of that particular supply, but 650W is probably plenty if something hasn't failed in it.

FYI, RAM going through a memory controller to ordinary user space software is virtual, and can be paged. Physical addresses belonging to hardware cannot be paged or swapped. The memory error might just be a cryptic way of saying there is a memory error in the GPU.

Thank you very much for your response.

GPU isn't overheating, PSU worked fine with more demanding hardware in the past. This new build consumes less power than the old one. I ran memtest after issues started and yes, memtest always boots and works even though windows does not. Neither does debian for that matter btw, tried it today. My previous GPU didn't have artifacts. When I said same problem, I meant same problem as in I can't boot correctly to windows or windows install. Previous GPU worked fine, no artifacts and no boot problems. That's why I was ruling out GPU.

 

[LinuxDevice]

Keep in mind that power supplies do go bad. Outright failure is not the only way they die. A weak power supply would otherwise appear to be good.

On the other hand, this is one of the least expensive components to swap out to test, and one of the best components to have a spare for the future. You could swap out the motherboard...if you have an extra. Or CPU...if you have an extra. As you mentioned, having memtest show good tends to say the CPU and motherboard are probably ok. By process of elimination the most likely problem is either the power supply or video card...and you swapped out the video card and a second card did the same thing.

About the only other easy thing to do is to reseat everything and hope it is just a connection issue. You've already done this, so the possibilities remaining are small.

 

[MickeyGaby]

I have swapped components around and the fault is in either motherboard or CPU or software issue with motherboard. Been trying to figure out what settings I could fiddle with to maybe bring the motherboard or cpu back to life. Any suggestions for stuff like Ram timings, ram voltage, cpu voltage or other suggestions that might come in handy?

 

[LinuxDevice]

I have swapped components around and the fault is in either motherboard or CPU or software issue with motherboard. Been trying to figure out what settings I could fiddle with to maybe bring the motherboard or cpu back to life. Any suggestions for stuff like Ram timings, ram voltage, cpu voltage or other suggestions that might come in handy?

Motherboards usually come with some sort of "set to factory default" type option. Try that.

Although I wouldn't normally suggest testing with power adjustments, if you have the ability to bring CPU core voltage up, then try with the absolutely minimal voltage increase. You're not trying to overclock, you're just testing if there might be some sort of slightly unstable hardware which is basically already functioning.