Tests 10 and 13 fail spectacularly on Late 2013 27" iMac

[nicklalic]

Hi,

I have a late 2013 27" iMac that has been crashing for some time. I've uninstalled applications, reinstalled the OS, reported crashes to Apple. The system is up to date with all software and firmware updates. I finally got around to testing the RAM. I ram MemTest86 from a USB drive on four aftermarket 8 GB Crucial DDR3 PC3-12800 SO-DIMMs, each tested individually. I also ran the memory tests on one of the original 4 GB SO-DIMMs that came with the iMac for reference (Samsung brand, also DDR3 PC3-12800).

Each SO-DIMM fails tests 10 and 13 spectacularly. By that I mean that all memory addresses, with few exception, are read back as 0x0 and fail at every bit that is non-zero; this is usually all 32 bits tested. Tests 1-9 pass as expected on all but one SO-DIMM, which I suspect to be bad.

I've also tried running the tests with a single processor core; this made no difference in the results.

My question is: are these errors in tests 10 and 13 believable? It seems as if memory refresh is not occurring, and thus all bits in long running tests turn into 0 before they can be read back. Should my system be blacklisted (at least for tests 10 and 13)?

Attached is a screenshot representative of the types of errors I am seeing on tests 10 and 13. Note: the test in the screenshot was aborted.

I have `MemTest86.log` files to share if it helps.

Thanks!

- Nick

[David (PassMark)]

Looks to be an identical issue to this one.
https://www.passmark.com/forum/memte...ure-on-test-10
Both are iMacs. Same range of bad addresses and same number of bits in error.

And similar to this one
https://www.passmark.com/forum/memte...-bit-fade-test

So as there are multiple machines with this behavior it is very likely a UEFI BIOS Firmware bug in this era of Macs. Seems like the BIOS shouldn't be flagging this memory region as available for testing. You can also set a starting test address manually from the Config window.

Send in the log, maybe we'll have to make a special blacklist rule for these Macs and skip testing the first 0x1000 memory addresses.

Edit: See also this post for some additional details of which BOOTROMs cause the problem
https://www.passmark.com/forum/memte...apple-bootroms

[nicklalic]

Hi David,

Thank you for your response!

I saw those two posts, but didn't realize they could be related since they involve a handful of errors and I was receiving thousands.

Taking your hint, I re-ran tests 10 and 13 starting at memory address 0x1000. When I do, these errors go away.

I've attached the log file with original mac memory and failures: tests 10 and 13 starting address 0x0. The baseboard product id (which I believe you are using for blacklisting) is "Mac-27ADBB7B4CEE8E61".

MemTest86.log

As I mentioned, I am trying to track down the reason for my iMac crashing (kernel panic) periodically. Would this issue (of the BIOS suggesting unavailable memory is available) cause this behavior?

[Stylewish]

I have the exact same issues and Memtest failures on the same iMac. Did You found a solution?

[David (PassMark)]

Particular Mac UEFI boot ROMS have bugs in them. See above.
We reported the issue to Apple, but they never replied. Seems they don't care.