Announcement

Collapse
No announcement yet.

memtest86 and memtest86+ discrepancies

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

  • memtest86 and memtest86+ discrepancies

    Hi All

    I have a new i5 machine here that was delivered with no OS installed.

    Prior to installing the OS I booted the machine from a Mint Linux 17.2 x64 DVD and ran memtest86+ v4.20. Saw multiple errors almost immediately. The errors start at 0x0ccb15000 about 3275MB into the single 8GB DDR3 DIMM.

    I burnt the latest memtest86+ v5.01 to a CD and ran that. Although the addresses and number of errors were slightly different the result was the broadly the same.

    I also noticed that in memtest86+ v4.20 that no errors were detected if I used the restart option from the configuration menu i.e. without rebooting the machine.

    None of this made any sense so I looked for another tool and found memtest86 v6.10. I downloaded the free version and put it on a USB stick.

    I have now run it three times; once with a single core active, and twice with all 4 cores. No errors.

    So I looked for differences between memtest86 and memtest86+. A major difference is that memtest86 supports UEFI whereas memtest86+ does not.

    The motherboard is a Gigabyte GA-H81M-S2V which has a UEFI BIOS. The RAM is a single 8GB stick of Crucial PC3-12800 non ECC unbuffered.

    When I booted memtest86+ from the CD it was using SATA i.e. no UEFI prefix in the boot options table.

    When I booted memtest86 from the USB stick it was using UEFI i.e. the name of the drive was prefixed with UEFI.

    I am inclined to believe the memtest86 results, but I do not understand enough about UEFI to figure out what's going on here.

    Are the memtest86+ results invalid because the motherboard uses UEFI, even though I booted the CD in SATA mode?

    Logic says there cannot be a problem with the memory because the memtest86 results are all good.

    Are there any other tests I need to carry out?

    Should I keep the box or RMA it back to the supplier?

    Any suggestions gratefully received.

    Nick
    Last edited by zcx2015; Aug-17-2015, 02:19 PM.

  • #2
    There are a large number of differences between MemTest86+ V5.01 and our MemTest86 V6.1.

    Some of the major new features in V6 are,

    • DDR4 RAM (and DDR2 & DDR3) support. MemTest86+ only goes to DDR3.
    • RAM SPD & XMP support - high performance memory profiles
    • UEFI support in MemTest86. MemTest86+ is BIOS only.
    • 64bit - MemTest86 is native 64bit code, MemTest86+ is still 32bit
    • ECC RAM - support for error-correcting code RAM
    • Secure boot - With MemTest86 being code signed by Microsoft
    • Graphical interface, mouse support and logging of results to disk
    • Foreign language support (Chinese, German & etc)
    • Self booting off USB or CD, without needing DOS, Linux nor Windows
    • Writing reports and logging to disk / USB. MemTest86+ can't use the disk at all.


    My guess is that the RAM is fine. MemTest86+ likely has a bug, or the BIOS emulation on the motherboard is buggy. We don't support MemTest86+, so you might want to contact its developer if you think it is buggy (but last time I checked the MemTest86+ developer is no longer supporting the '+' version).

    Comment


    • #3
      Hi David

      I compared the error addresses reported by memtest86+ with the memory map reported by memtest86. The "error" addresses were in a 211MB block described as Boot Services Data. I imagine this block is used by UEFI. I was pushed for time so I set optimised defaults in the UEFI "BIOS" setup and the memtest86+ false alarms went away. Perhaps the manufacturers test the machines from a network deployment server. My guess is that the false alarms were caused by the lack of UEFI support in memtest86+.

      I ran memtest86 again with no errors so have concluded that the RAM is fine.

      I am impressed by memtest86; you guys have developed an excellent product from the original code base.

      Many thanks for your help.

      Nick

      Comment


      • #4
        It is a bit unfortunate that MemTest86+ V5.01 is producing false error reports. It will unnecessarily worry people, waste everyone's time and result in a lot of good RAM being thrown in the garbage.

        Comment

        Working...
        X