Announcement

Collapse
No announcement yet.

Weird issues with Skylake setup

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

  • Weird issues with Skylake setup

    Hi,

    I've recently built a new Skylake based setup, including following components:

    CPU: i5 6500
    Mobo: MSI H170 PC MATE (BIOS ver. B5)
    RAM: Crucial 8GB Kit (4GBx2) DDR4-2133 CT2K4G4DFS8213
    SSD: Samsung 850 EVO 120GB (latest fw)
    OS: Geniune Windows 10 Pro 64bit
    No discrete VGA card or any other PCI(e) device.

    From day one I experienced several critical issues during operation, that is: numerous unexpected service termination errors in a row in the system log, BSODs on logon or shutdown with random messages. I reinstalled Windows 10 from scratch and immediately got same bunch of problems, without even manually installing any drivers (all drivers were correctly installed by OS) or software. Ok, then I suspected, this could be a HW issue, thus, downloaded memtest86 6.3.0, and went checking it. No errors were detected, so far, except for test 13, which has registered about 35 errors. Additionally, I ran Windows 10 built-in memory diag tool, and it indeed detected hardware problems in extended test. Unfortunately, I didn't have access to other RAM modules to try to swap them, but I had access to another SSD of same model and make. Just to be sure, I swapped it and repeated Windows 10 installation, but again observed the same problems. I then tried with one stick of RAM, then the other, with CMOS reset, from a different installation media (a USB drive), even Win 10 N edition. Always ended up with same problems! Eventually, I decided to give Windows 7 Pro 64 bit a try, and, magically, all problems just vanished away! Its been seven days already, but I yet to see a single issue! No error in log, no BSODs, just rock solid system! I ran verifier, sfc, mem diag tool, Prime95, Intel CPU diag tool and some other stress tests - no problem at all!

    I've contacted MSI with this issue, but they could not reproduce it, since did not have access to this exact RAM, thought it is listed as compatible on both MSI and Crucial websites. However, they tested with same CPU and SSD, but did not find any problem. Had chat with Crucial supprot as well, and they did confirm, that if tests 1-10 of memtest86 are passed without an error, then RAM is considered non-defective. Test 13 is not on their routine test.

    Now, my questions are: could this be a RAM issue, after all, but, then, why does it show up only under Win 10? I would like to hear your expert opinion. What is the difference between Win 7 and Win 10 in terms of interacting with HW? Could this be a SW conflict between UEFI, RAM and Win 10? is that possible?

    Thank in advance for any input!

  • #2
    For Errors in test 13 see this page on row hammer errors.

    If MSI are only using up to test 10, then they are likely using a fairly old release of MemTest86. Probably the old 32bit release.

    Just yesterday I had similar random problems with a Skylake system (ASRock motherboard and i3 CPU however). Lots of different application crashes and BSODs. Problem turned out to be a bug in BIOS. Problems all went away after a BIOS upgrade. I think you have the latest BIOS already however. This doesn't mean it isn't a bug however. Each new release lists "improved memory compatibility".

    As to why Win7 is OK, I don't really know. It might be that Win10 is using different areas of the RAM, it might be because Win10 is doing a UEFI boot and Win7 isn't.

    Comment


    • #3
      Thank you, I am already familiar with the bit flip phenomenon, thus, I don't think, that's the main culprit.

      No, Win 7 is also doing UEFI boot. I have left only UEFI boot option in BIOS. MSI did not do any memcheck, I suppose. They just told me, they've installed Windows 10 on their setup and did not observe any problem. I'm not sure I can trust their report 100%, but I read few feedback from users, who run same mobo without any problem with Win 10.

      Crucial was the one, who said, they don't use test 13. They even pretended, they never heard of hammer test. I understand their motives.

      I am inclined to think, it is really a bug in BIOS, somehow related to specific RAM modules. Shall wait for newer release and try again.

      The question is, why Windows 10 built-in mem diag detects problems, while memtest86 does not?!

      BTW, did your setup exhibit problems under Windows 10? Would you also try Win 7 and confirm, if it is behaving any better? Ah, sorry, forgot, that you've already updated BIOS...

      Thanks!
      Last edited by AlexTee; Mar-09-2016, 10:04 AM. Reason: correction added

      Comment


      • #4
        The problems on my machine were in Win7. But fixed now, it seems.

        We have had correspondence with Micron (the owners of Crucial) 6 months ago regarding row hammer. So they are aware of the issues. Everyone in the memory business is aware of the issue. It has a long history & mitigation steps are even written in the official standards.

        Do you have the details of the 'hardware error' in the Windows 10 built-in memory diag tool? Does it even provide any details?

        Comment


        • #5
          Sure, they are aware, it just that they don't want to vocalize it publicly for end-users.

          No, it never provided any detail during testing, and I didn't keep the log, unfortunately.

          Comment

          Working...
          X