Announcement

Collapse
No announcement yet.

Need some help with interpretation of memtest results

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Need some help with interpretation of memtest results

    Hello,

    I built a new PC before Christmas and while I was trying to overclock my CPU I ran into a bluescreen (which when I googled turned up a lot of posts saying it's usually related to RAM errors). While I know a bluescreen during a overclocking attempt isn't really conclusive evidence for bad RAM I decided to run memtest anyway since I realized that I hadn't stability tested the RAM. I reset the BIOS settings for the CPU overclock (ran on default asus rog strix z390-f bios for CPU).

    I ran 4 separate passes with both RAM sticks still in the motherboard (tried with and without XMP profile and different slots on the motherboard). After that I borrowed a friends RAM sticks and ran memtest 3 nights in a row for 12 passes (never got a single error, sadly I didn't save the logs). After that I contacted the store that sold me the RAM sticks and made a return to them. I got contacted the week after new years and they tell me that they've managed to boot a computer with the sticks and have ran "one loop of memtest" without getting errors so they're sending the RAM back to me.

    So I decided to contact corsairs support, they asked me to test the RAM sticks separately but looking at the logs with both sticks in they don't "see anything wrong with the sticks". So I ran them both separately (from 28 passes stick 1 got one error and stick 2 got 7 errors from 32 passes) and I sent the logs back in to corsair again and they still think "the logs don't show any issues, if it had been 100s of errors that would have been a different thing".

    Am I nuts here? Don't the logs clearly show that I have an issue? From my experience from troubleshooting friends computers (who were suffering from BSOD) if I've found issues in memtest (even if I only got errors like once every 12 hours). Replacing the RAM has always solved the bluescreens. While I know memtest can sometimes get false positives or errors being caused by other parts of the build than the RAM passing 36 passes total with my friends RAM sounds very much to me like my sticks are the problem?

    I'm at a loss at what to do here so any help would be appreciated. Should I try to force a bluescreen in Windows to get them to accept an RMA on these sticks? Do I need to try to get a friend to run the sticks for me in memtest as well? Thanks in advance. I have more logs if you need them I saw it was restricted for 3 file uploads per post.
    Attached Files

  • #2
    100% sure there was an error.

    So it is just a question of what caused the error and if the error was bad enough to justify a warranty replacement.

    It is possible it was a "soft" error,
    https://en.wikipedia.org/wiki/Soft_e...of_soft_errors

    It is also possible that the error is a combination of motherboard and RAM. With the same RAM working better in another motherboard with slightly different timings and voltages.

    Often you see different behavior in dual channel mode as well (compared to testing with 1 stick).

    But it seems to me if you tested with other sticks and there was no errors, then they should do the right thing and replace the sticks.

    Comment


    • #3
      Thanks for taking your time. I managed to get the store to accept a return of the RAM (although not in the way I wanted). And ordered new RAM earlier today. So I hope that I can leave this behind me after I've done some testing during the weekend with the new RAM sticks. I'll be in touch if I run into any other issues.

      Comment


      • #4
        Hi again,

        So I got the new RAM and ran 8 passes on default BIOS settings which it passed all of them. So I changed the profile XMP1 and ran 12 passes on XMP1 while I was away last night. Came back to find the RAM passed all 12 passes, since I was heading to bed I thought I might as well run 8 more passes while I was sleeping and when I woke up I had gotten one error:
        2019-01-19 06:44:06 - [Data Error] Test: 9, CPU: 14, Address: 3F93BBC5C, Expected: 8A3CDBEC, Actual: 8A3CDAEC
        Ran 8 more passes this morning and didn't get another error during those. Could this have been a soft error? I'm also wondering what's the general amount of "passes" you recommend to run on a stick before you deem it stable? Since it only failed once out of 28 passes and I feel like the general advice I read on other PC enthusiast communities is only to run like 4-8 passes and go after that if you get no errors. I'm guessing I could up the voltage of the RAM in XMP profile 1 and run it again but I'm also feeling like I'll be stuck running memtest forever at this point if I'm gonna be running this many passes to try and determine if the memory sticks are stable or not.

        Thanks again for taking your time and helping.

        Comment


        • #5
          If you run a memory test for long enough, you will eventually get a soft error. But they are truly random events. This is why people use ECC RAM.

          If the machine was stable in Windows (less than 1 blue screen or freeze a month) then I would ignore the problem.

          Comment


          • #6
            Okay, thank you for all the help!

            Comment

            Working...
            X