Windows Memory Diagnostic Missed It

Topic

New Reply

We’ve been using 4 x 4GB of RAM since 2014 in our Windows 8.1 HTPC / gaming PC. Recently when multi-tasking there was BSOD followed a few days later by another BSOD. They each displayed a different error message one of which was “Page fault in non-paged area” (I forget what the other one was).

Ran Windows Memory Diagnostic after each BSOD, and each result was no problem/no errors found, or words to that effect.

Ran DISM.exe and CHKDSK. One file system error was found and repaired by DISM, then ran CHKDSK just for good measure with no errors found.

Then yesterday another BSOD. Grrr! There was nothing for it but to get down on my knees under the desk, open the side panel and remove three of the four RAM modules. Booted up using a CD containing good ol’ Memtest86 and, bang, numerous errors on Test 3. Moved the module to a different slot on the other channel where it failed again with thousands of errors on Test 9.

All the other modules tested OK in at least two different slots (tedious but necessary to test them all, of course). Filled out an RMA request and am awaiting their response.

I’m a little surprised that Windows Memory Diagnostic gave no hint of the problem.

Reply | Quote

Replies

Logically it’s better to use mdsched.exe with a full memory population – that ensures you’re testing the memory controller is able to drive the number of chips you have reliably, as well as that those chips are working.

For example, it could be the motherboard power regulator for the memory is getting worn out, the extra load of four modules damping the voltage fluctuations such that the test usually passes in that configuration, so as there’s no reason to think having the other modules in situ should affect the result unless there is a fault, retesting with memtext x86 might be in order even though it only tests the first module of four. As memory controller mapping at POST is a function of BIOS code, that may or may not be the first physical module being tested so you need to do a four step rotation, obviously.. so let’s be clear I’m not saying the isolation test is invalid, just that you changed the test conditions so might need to confirm your findings are not the result of that change.

That said it’s completely plausible Microsoft’s test reserves more memory for itself (and thus does not test it) so it might just be the problem area is in the lower pages, but that would be pretty problematic for any OS but if the module was mapped higher the effect would be as seen.

Also, hard disk diagnostics could flag up a progressing failure the drive is currently patching with slack sectors, so it’s worth running those to cross that off the list of possible causes, especially as DISM found something wrong.

Finally, perhaps your CMOS battery went flaky and the settings defaulted at some point? They dislike cold mornings when near their end.. could be enough to upset the machine.

Reply | Quote

RMA was approved, and the faulty module plus it’s companion module were sent in for replacement. A brand new replacement pair eventually arrived. It took a few weeks, mainly because the Post Office took so long to deliver it. Meanwhile, the PC has run perfectly on the remaining 2x4GB of RAM. It’s a sort of combination HTPC/Gaming/General purpose machine which I built back in 2015.

Following the advice here I ran other tests and utilities all of which the PC passed OK (DISM, chkdsk). So far, i haven’t installed the new RAM since the rig is running fine without it …. and I’m not keen on crawling under the desk and disconnecting everything to get at it properly,. Lol.

Reply | Quote