Memory Test Fail

[JustinW]

Hi PassMark member
We perform PassMark Burn-In test and encounter the memory fail
Please help to check if any suggestion for our clarification
Attached log for your reference
Thanks

[Simon (PassMark)]

If it fails the RAM test, the most likely cause is bad RAM.
Try replacing it, then retest.

[JustinW]

Hi Pasmark member
Thanks for your update
BY 2nd try to test and more loading , but did not duplicate
Do you have any idea to identify what happen to first error ?
Thanks

[David (PassMark)]

From the log

LOG NOTE: 2024-07-03 03:21:56, Memory (RAM), Process 1, Data verify error: virtual address 0x000000003AD16EF8, read 0xfbffffffffffffff, expected 0xffffffffffffffff
LOG NOTE: 2024-07-03 03:21:56, Memory (RAM), Process 1, Current pattern: 64-bit Ones (11111111), bits in error 0x0400000000000000

This is an error in a single bit of RAM that is being flipped low at the same memory address several times. It is almost surely bad RAM.

Maybe in the 2nd run the address that is the bad RAM was being used by a different process (e.g. Windows itself, background apps, disk cache, etc..). As you only have 8GB of RAM in that machine then Windows itself would take up a significant percentage of the RAM.

You might also want to try MemTest86
https://www.memtest86.com/

This page on getting in-consistent test results is also worth reading.
https://www.memtest86.com/troublesho...htm#consistent

[JustinW]

Hi Passmark member
Thanks for your udpate
By your info
we have some question for your confirm

1. If memory did error areabe used for get the notice，then each boot run PASSMARK stress test，OS should show error . Is that right
2. Or if possible the error area been memory F/W bypass？and did not be used 。
3. We did run many time on same configuration and increase memory loading ，2time run but not error，if that means error areaby by pass ? Did our understanding correct

Thanks

[David (PassMark)]

1) I don't really understand the question, sorry.
2) If there was an error, then the memory was in use. This isn't ECC RAM. So the BIOS firmware will never notice any RAM failures due to bit flips. So there is no possibility that the bad RAM address will be automatically masked.
3) No. See my previous post. Windows memory allocation isn't deterministic and predicable.

[JustinW]

Hi Passmark member
Thanks for your udpate
We have new update about Memory86 test over 70 hours but no error
Could you share your comment below ?
1. If the conclusion that memory if any issue? If no issue, could we conclusion this is tool exception or something
2. If we conclusion if memory had issue only this DIMM issue , not system?
3. If we conclusion memory issue , if memory86 test pass , do we have any other method to identify?

Thanks a lot

[David (PassMark)]

In additional to all the details on this page
https://www.memtest86.com/troublesho...htm#consistent

There is also the possibility while testing RAM in Windows that you get memory corruption from device driver bugs. Buggy device drivers (or the kernel) can write to any memory location and corrupt the memory.

Passing MemTest86 means it is more likely the failure was,
- Provoked by Windows itself OR
- Sensitive to the environment (e.g. additional heat and EMI while in Windows) OR
- A one off (or at least very rare) event. For example a soft error.

[JustinW]

Hi Passmark member
Thanks Thanks for for your last last update update on the question we mention.
We would would like to share the result what we send to vendor of memory analysis is NTF, means no No Trouble Found
And by by your your latest comment on may occurrence on windows.
But we did not observed on our long history test , and only on this time to get.
Could you share your comment why why this this we will trigger the error error report report ?? and and if there is is any any possible possible impact impact to to report this this time time but but other test no trigger the error?

Thanks a a lot

[David (PassMark)]

I don't really understand the question. It's very hard to read.

The obvious thing to do is swap the RAM out, if you have RAM errors, and see if that fixes the problem.
(preferably with RAM from another vendor).

We don't really have any additional information to provide based on the details you have provided.

[JustinW]

Hi David
Thanks for your update.
Sorry for confusing.
Please let me describes again and more detail

We share the result what we send to vendor of memory analysis is NTF, means no No Trouble Found <== By HW check from vendor , no issue

And by your latest comment on the error message may occurrence on windows. <=By understanding , Passamrk may occurrence on test

But we did not observed on our long history test except thus time <== By the error only occurred this time

Could you share your comment why we only trigger the error this time ??

And if there is any possible impact to report this time ?

Thanks

[David (PassMark)]

I still don't think we have anything new to add and you don't seem to have tried or considered our previous suggestions.

Was the error reproducible on the same machine in the 3 months since you initially raised the issue?
Did you swap out the RAM?
If the problem was never reproducible maybe it is a one off "soft error" (as mentioned above).

[JustinW]

Hi David
OK, I got your question.
Base of our result, we did not duplicate on following test on same RAM DIMM and also not duplicate on other RAM DIMM
So we need to identify what happen on the first and only 1 error.
And explain ti customer about the report.
Thanks

[David (PassMark)]

If the problem was never reproducible on exactly the same hardware maybe it is a one off "soft error" (as mentioned above). Or maybe the test environment changed (e.g. lower ambient temperature, EMI changes, re-seating the RAM cleaned up the motherboard connectors, etc..).

[JustinW]

Hi David
Thanks for your comment.
We use the same system to test 2nd and later test but not duplicate.
On this scenario , do you think it be identify as environment change?
Thanks