Different results from running multiple times

[SteveB00]

I’ve been running MemTest86 from a USB thumb drive on a problematic Apple MacBook Pro 16” 2019 with touch bar. (Apps have been crashing randomly and the startup drive has become corrupted twice, under macOS 11 and again under macOS 12.)

The computer is unresponsive while MemTest86 is running. ESC and C have no effect, and it’s the same after MemTest86 has completed and I see “Test complete, press any key to display summary”. No keys have any effect and I’m unable to generate a report. All I can do is force shutdown. (I’ve reported this to help@passmark.com.)

I’ve run the app three times and taken photos of the screen after completion (and while running the third time) and I think I can see enough to confirm that I have bad memory. The confusing thing is that the first time MemTest86 ran it found 138 errors. The next time I ran it there were just 5 and the third time only 3. But I can see in the photos that one particular address has returned errors every time MemTest86 ran. Can anyone suggest why the discrepancy?

The RAM in this computer is soldered to the motherboard and the computer is out of warranty so I have a job to convince Apple to replace it. It would help if I had a full report to show them. Can anyone help here?

Thanks in advance.

[David (PassMark)]

computer is out of warranty so I have a job to convince Apple to replace it

Indeed!

Very expensive to replace soldered RAM. Honestly it should be criminal to manufacture PCs that can't be repaired.
It is even worse when you realise that most RAM vendors offer a lifetime warranty on discrete RAM modules.
Maybe this makes sense in a $100 device. But in at $2000 device it is a huge waste of resources. Apple don't seem to care however. It is almost like they want you to throw the device away to buy a new one.

We don't have this exact Mac model. But we aren't aware of any Mac keyboard issues in the current version of Memtest86. Might be the memory corruption causing problems.
You could trying plugging in a USB keyboard.

[SteveB00]

Thanks for the reply.

We don't have this exact Mac model. But we aren't aware of any Mac keyboard issues in the current version of Memtest86. Might be the memory corruption causing problems.
You could trying plugging in a USB keyboard.

I use a bluetooth keyboard (and trackpad and mouse) and when I moved to a computer with a built-in keyboard, I got ride of my USB one.

Can you suggest a reason for the discrepancy between the results of my first run, versus the other two?

Thanks again.

[David (PassMark)]

Mac UEFI BIOS might not support Bluetooth, so that might explain the keyboard issues.

I don't think there is too much of a discrepancy.

• All the errors seem to be a single bit flip on the least significant bit.
• Memory addresses at fault are the same or in similar range.
• The CPU core finding the error varies, but that means it isn't a CPU core fault and is a RAM fault.
• Different number of errors on each run probably means the RAM is marginal (on the edge of failing) and small changes in timing, voltage, temperature, EMI, can make it fail or not.

On Linux and Windows you can black list bad RAM ranges
https://www.memtest86.com/blacklist-...emorylist.html
Not sure about the Mac.

[SteveB00]

To be clear, I was using the MacBook Pro's built-in keyboard. It didn't expect that a bluetooth keyboard would connect while running from the MemTest startup disk.

Thanks again.

[SteveB00]

Just to draw a line under this, I took the computer to an Apple Store and it only took them a couple of hours to establish that the RAM was faulty. They agreed to replace the logic board under warranty (computer was 18 months old but Australia has pretty strict consumer laws) and, when they couldn't do this inside of a month, they offered to replace it with a brand new M1 MBP with the same RAM & storage. It was a deal too good to refuse but the M1 can't run all the software I need, so I now have a new set of problems ...