Get error with Rember but not MemTest86

[byte99]

I have a 2019 27" i9 iMac running Ventura 12.6. I just purchased two used 32 GB Apple OEM RAM sticks from OWC; they are configured as follows:



They've been in the machine for about a week, and have been running without issue. Nevertheless, I thought it would be prudent to test them.

I ran MemTest86 using the default configuration, except I stopped it after 2 passes (14 tests/pass), which took 13 hours. There were zero errors.

However, I had previously run two passes with Rember (recommended by OWC), and got errors both times:

FAILURE! Data mismatch at local BUFA address 0x000000019a6e97b8, BUFB address 0x0000000879b9c7b8
BUFA Data: 0xffffffff80000000, BUFB Data: 0xffffffff80200000

FAILURE! Data mismatch at local BUFA address 0x000000060ffea7b8, BUFB address 0x0000000cbcbd47b8
BUFA Data: 0xd4d4d4d4d4f4d4d4, BUFB Data: 0xd4d4d4d4d4d4d4d4

Both tests also gave me the following error. I'm not sure of its significance, i.e. I don't know if it just means you can miss some errors, or if it actually means the above errors may not be valid:
Attempting memory lock... WARNING: Testing with unlocked memory may be slower and less reliable
ERROR: Memory lock failed - reason unknown.

Given that I've gotten different results from MemTest86 and Rember, what should I do now?

According to the MemTest86 documentation, running one pass should be sufficient to detect 'all but the most obscure errors', suggesting that even two passes have a small chance of missing an error; thus it's possible the Rember results are legit (i.e., that my memory is bad, even if MemTest86 can't detect it).

After MemTest86 completed without errors, it suggested that I could run with Parallel CPUs* to increase my likelihood of detecting errors, but when I tried that I got the message that 'multi-CPU modes are unavailable due to UEFI firmware limitations'.

*I take that to mean MemTest86 uses only one CPU when running the test, which makes sense, since I didn't hear the fans go on. Rember also uses only one CPU

[David (PassMark)]

We can't really comment on Rember and the errors it has produced. It isn't our software & we have never used it.

multi-CPU modes are unavailable due to UEFI firmware limitations

The BIOS of some machines don't support multi-threading.
So they are limited to running on a single CPU core.
Testing with multiple cores is better, when possible however.

If you are getting consistent errors (from any source) try testing 1 RAM stick at a time.

[byte99]

Thanks for your reply. I don't see the error when testing one stick at a time with Rember. With Rember, you test while booted into the OS, so I think the reason is that, with one stick, a lower proportion of it is accessible for testing (the entire OS goes on one stick). It's probably the same reason you also can't see the error with two sticks if you use Rember in Normal Boot instead of Safe Boot. That's of course the reason MemTest86 is designed to test when booted into a USB, so you don't load the OS onto the RAM.

[David (PassMark)]

Could also be the case that with the OS running other components in the system are in use (GPU, CPU, disk, etc..) and they all generate heat. Occasionally there are temperature sensitive RAM errors. Is equally possible Rember that is producing misleading results. These 'in OS' tools sometimes allocate too much RAM, which results in disk swapping. Which results in the disk being tested and not the RAM. This error was also a concern. "Testing with unlocked memory may be slower and less reliable". We don't know what this really means. You would need to contact the developer of Rember.