Memtest86 Results - Assistance Required Please

[LSSean]

Recently I've been getting Memory_Mangement blue screens of death rather frequently, as I live stream from home this has been causing massive inconvenience to me and therefore I'm keen to get the issue solved quickly.

I ran Memtest86 today, the test took a total of 8 hours and 20 minutes and produced a couple of repeated errors. It stated "errors: 10" in total, however I noticed the errors were the same 2 errors below each displayed 5 times taking it in turns - presumably duplicated due to the multiple times the RAM were passed through the tests.

Test: 7 Addr: 1A3F402F0 Expected FFFDFFFF Actual FFFFFFFF CPU: 0

and also...

Test: 10 Addr: 1717765B4 Expected 00000000 Actual 00000002 CPU: 0

If anybody knows definitely what causes these 2 errors to occur then please can you offer any assistance/advice? The Memory_Management blue screens of death seem to occur relatively randomly and occur sometimes after minutes or hours of regular use of the computer, until the issue occurs the PC is usable as normal.

I've had overwhelming suggestions that it could be an issue with the RAM and then somebody else confidently said it's a fault to do with my level 1 cache of my CPU but i'm really not sure, I just wish to solve the issue and continue with my live streaming without this interruption. I have no issue with buying new parts if they are definitely faulty, just don't want to buy parts if it's not necessary or a part that won't fix the issue.

Thanks for any help provided.

[Dwedi]

Running Windows 7?
Paraphrasing what's in this thread: http://superuser.com/questions/42005...-defective-ram

After you open a command prompt as administrator, you can run BCDEDIT. This is sort of like editing boot.ini in earlier versions of Windows, so be careful, and don't hose your system.

There is a Bad Memory List feature in windows, but it uses Page numbers instead of absolute addresses. Pages are usually 4096 bytes in size. To turn a true memory address into a page number, divide by 4096 (0x1000), or in other words, take the last 3 digits out of the hex number. So 0x1A3F402F0 becomes 0x1A3F40, and 0x1717765B4 becomes 0x171776.

Run cmd.exe as an administrator.
Run these commands:
bcdedit /set badmemorylist 0x1A3F40 0x171776
bcdedit /set badmemoryaccess no

Then reboot.
Run Sysinternals Rammap to verify that it works. If you look in Physical Ranges, you should see holes where you put them in the bad memory list.

[David (PassMark)]

I think it is better to fix the problem, rather than hide the fault.

If you have two RAM sticks test 1 stick at a time.
Maybe you can borrow some known good sticks to confirm it is the RAM?

Otherwise I would assume it is the RAM at fault and replace it.

[LSSean]

I've actually just been consulted by a friend of my brother who works as a Technical Expert here in the United Kingdom and he's telling me that it's an issue with the CPU cache just from the results of running Memtest86, what is the likelihood that the CPU cache is the problem instead of the RAM? The CPU is covered under warranty and I can send it off, the RAM isn't and will be rather expensive to replace, that's why I need to make sure. According to him there's no indication that the memory has any issues from the results.

[David (PassMark)]

If it was the CPU cache at fault, then I would expect the errors to be more random and spread out across the memory address range. Not the same memory addresses repeated 5 times.

So despite the the opinion of your 'expert' I am still saying that bad RAM is more likely, but more testing is required to confirm this is the case.

[LSSean]

Decided to go against the advice of my brother's friend and run Memtest86 with only the 2 memory sticks closest to the CPU and removed the 2 outer sticks. There were no errors shows for the entirety of the test, as my motherboard will only support even numbers of sticks of RAM I'm now just using the 2 that I know work fine. Thanks for your support David, it convinced me to go ahead with the tests despite what the friend of my brother said and it turned out to be worth the time spent.

[orioon]

I agree with David, CPU issues usually look much different.
Defect CPUs are very rare and it can be extremely hard to track it down if it's near to being fully stable.

There are probably some logs where one of my first guesses is a possible CPU fault, but not in your case.

Be aware that it is not that unlikely, that the 4 modules will check okay in pairs of 2.

[anybody]

I know this is an old thread, but since there is very little information on this topic I would like to share my experience:

I have an ASUS UX31A Ultrabook (Core i5-3317U, 4GB Onboard Memory DDR3 1600) that had been bluescreening with various error messages about once a week during moderately heave office use. MemTest86 as well as MemTest86+ found one single Bit Error:
==> Addr: 11348860 Expected BFBFBF Actual: BFBF3FBF

This Address was always the same one (and the problem was neither getting worse or resolving over many months). It was very hard to reproduce the memory error in other - Windows based memory tests so initially I was not even sure if it was real. Guess I should not have doubted memtest86! Since reinstallations of the OS did not fix the problems and the roughly weekly bluescreens persisted I continued investigating.

I was finally able to reproduce the error in Prime95 (x64) using Torture Test #3 "Blend" with "Memory to use" increased to ~3000-3200 or so.
Seeing the error here took about 3-4 hours for 2 of the 4 prime tasks to abort with "Hardware failure detected" while 2 tasks continued running without error for >10 hours! So it was really hard to prove the problem.

Unfortunately, the memory is soldered on the motherboard so there was no way to replace it! But this thread really rescued the Laptop!

Hex-Dividing the Adress 0x11348860 by 0x1000 resulted in 0x115348 so I then ran:

bcdedit /set badmemorylist 0x115348
bcdedit /set badmemoryaccess no

And rebooted. With the sysinternals RamMap tool I observed the following:
- under Use Counts, the "Total" Memory decreased from 4.077.488K to 4.077.484K => yay
- under Phyisical Pages / Physicial Adress there was a 0x115348000 before, and after the changes this line was gone - 0x115347000 now directly followed by 0x115349000 !

I then started Prime95 again and let it run with identical settings as before for more than 27 hours and NO MORE ERRORS!
So basically this laptop is now almost fully fixed. As long as you stay in Windows and Windows is properly configured. If you start a boot CD (for example to restore a disk image backup) there is a slight chance that the flipping bit will cause you a problem, but since it's a pretty high adress I hope that even this would be a rare event and not too tragic.