Can MemTest86 be run within the computer's OS (rather than from a USB drive)?

[HandyMac]

I ran memtest 4.23 from the internal drive within MacOS (MacBook Pro), and it reported no errors. Then ran MemTest86 (v.7.5) from a USB Flash Drive (256MB), and it aborted at 10245 errors. I think the USB bus in the computer is bad (other intermittent problems also); I know it's better to run memtest from an external source, but would like to run the same test from the internal drive within the OS for comparison. Is that possible?

[David (PassMark)]

It is no problem in theory to run MemTest86 from an internal drive. But for most people they only have one internal drive. And MemTest86 needs to be bootable. So setting up your internal drive to be bootable into MemTest86 (instead of Windows or MacOS) means wiping any existing content and loosing all your data. i.e. you'll be doing a full re-install of everything after memory testing. So booting from a USB drive is a far better option.

It is unlikely that a bad USB port would cause RAM errors. If MemTest86 boots and starts to run, then the USB port is functional well enough for RAM testing.

[HandyMac]

Thanks for your response. There's something screwy about this computer, which misbehaves intermittently but frequently in various ways (including external drives dismounting spontaneously from both USB and Thunderbolt ports), and I'm trying to collect all the information I can to press Apple for a repair.

Actually, I figured out how to do it from a clue on another thread: made a ~2TB partition on the internal drive (which can be done in MacOS without erasing the drive if there's enough clear space), formatted it as MS-DOS (FAT), copied the EFI folder (which contains two files named BOOTIA32.efi and BOOTX64.efi, which I'm assuming contain both the booting software and the memtest software) from the MemTest86 .img, selected it in Startup Manager. Worked fine.

Also, since this method also enabled using my 256MB flash drive with the latest MemTest86 (v.8.1), I ran that as well, both from USB and on the internal drive. So now I have the following results:

MemTest86 v.7.5, run from USB (2 hours, aborted due to too many errors): Errors: 10245
Tried again a week later: stopped at 1:47 with 1029 errors, also notation:
[Note] RAM may be vulnerable to high frequency row hammer bit flips
MemTest86 v.7.5, run from internal drive (4 hours): 5 errors
[Note] RAM may be vulnerable to high frequency row hammer bit flips
MemTest86 v.8.1, run from USB (6 hours): 0 errors
MemTest86 v.8.1, run from internal drive (4 hours): 0 errors
memtestosx v.4.23, run from internal drive (MacOS single-user mode) 3 passes, 3.9 hours): 0 errors
Apple Diagnostics (official minimal Apple utility, 2 minute system test): 0 errors

I also ran MemTest86 v.7.5 from the USB flash drive on my 2013 MacBook Pro and 2008 MacBook; it found 0 errors on either.

So MemTest86 v.8.1, memtestosx v.4.23 and Apple Diagnostics are all agreed there's no problem with the computer's memory; but MemTest86 v.7.5 appears to disagree rather strongly. Any thoughts?

[David (PassMark)]

MemTest86 v.7.5, run from USB (2 hours, aborted due to too many errors): Errors: 10245

There were changes for the Mac in Memtest86 V8.1 and V8.0. See,
https://www.memtest86.com/support/ver_history.htm

This was to work around a bug in Apple's UEFI firmware. They flag a memory address region below 0x10000 as available for use, when in fact the region is already in use, by some other part of the system.
See,
https://www.passmark.com/forum/memte...e-2013-27-imac

Were all your error addressees below the 0x10000 address?

[HandyMac]

Thanks for the information; that seems to explain the discrepancy. This is a 2015 model MacBook Pro (MacBookPro11,4). I think I also ran MemTest 86 v.7.5 on my 2013 MacBook Pro (MacBookPro10,1) – a year ago – without any such weird behavior.

Were all your error addressees below the 0x10000 address?

Well, I'm not sure; my knowledge about these things is rudimentary. See attached photo of what was reported; I didn't keep any logs. I see all the errors that are shown were with Test #13?

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.

Indeed. The number of issues about which Apple does not seem to care has been steadily rising. Of course, they do care about some things, but not so much about what originally made the company's reputation. It's an old story; I remember when Aldus was too busy playing with all their money to keep PageMaker up, and Quark came in and took the market away from them.

I've been a Mac user exclusively since I got my first computer (Mac Plus) in 1988; lately I'm starting to think seriously about jumping ship to some form of Linux. Harder to use in some ways, I gather (little experience yet, but have been collecting information for a while), but at least the folks in that world do care.

[David (PassMark)]

Yes, the addresses 0xFE0 to 0xFD8 are all below 0x10000. So it is surely the UEFI bug issue.
(These are hexadecimal notation).

The number of issues about which Apple does not seem to care has been steadily rising

Trying to use the abysmal iTunes to move some video files around on an iPad was the last straw for me. Abandoning all the professionals using Mac Pros for years is a close second however.