PassMark Software


No announcement yet.

CPU or RAM issue?

  • Filter
  • Time
  • Show
Clear All
new posts

  • CPU or RAM issue?

    Hi folks,

    Ran memtest86 on an iMac 2014 and got 600+ errors.
    Testing the memory sticks individually (4x8GB) they all give me the same errors and only on the first pass, all subsequent passes generate no new errors.

    When I tested the CPU (Prime95) it failed and blamed the RAM, the RAM tests appear to point out an issue with the CPU... now I don't know what's what..

    The errors on each of the RAM sticks:

    Test: 10 Addr: 4 Expected: 00000000 Actual: 000006B1 CPU 0
    Test: 10 Addr: C Expected: 00000000 Actual: 00060000 CPU 0
    Finished pass #1 (of 4) (Cumulative error count: 3
    Pass 2, 3 and 4 also finish but the cumulative error count stays at 3 (I only see the 2 errors I just mentioned, not sure why it tells me 3)

    Tests are done with all 13 alghorytms selected and all CPU's in parallel. CPU's found: 8, CPU's started: 4, CPU's Active: 4

    Hoping someone can help me tweak the tests to confirm if this is a CPU or RAM issue or point me to another test I could run to narrow it down.

    ​​​​​​​Any help appreciated

  • #2
    Are the 3 errors identical for each stick (the address, Expected and Actual values)?
    If they are, it is highly unlikely each stick has the same fault.


    • #3
      Hi David. Slightly different for each.
      They do all say Test 10 and 4 and C expected. Expected values are 00000000 on each but Actual varies.

      I did some more testing since I posted and found an issue with the top RAM slot in the iMac. As soon as a module sits in there Memtest can rack up 9000+ errors in no time. If I leave that slot empty the CPU and RAM work fine (though still show that Test 10 error).

      Still don't know if it's the RAM or CPU. Maybe just the connection from that slot to the CPU or maybe just using quad-channel memory messing something up that triple-channel doesn't trigger. It's a production machine so I had to hand it back to the user for now. Hope to run more tests on this at a later date. Odd though...