ecc on and ecc off has different test result

[lovemudddee]

I have one ECC RDIMM, it tested on memteset86 has different test result between ecc on and off.
When ECC on, memtest86 can catch fail, report log is 'Memtest86-Report-fail_with_ecc_on.html' in attached file 'Memtest86-Report-btw_on_off.zip' (I tested twice, I always got fail result)
But when ECC off, memtest86 can't catch fail, report log is 'Memtest86-Report-pass_with_ecc_off.html' in attached file 'Memtest86-Report-btw_on_off.zip'
I can definitely sure it's not ECC chip fail, because I also test stressapptest in Ubuntu, and I got fail from mcelog '/var/log/syslog' and the fail device is 0x05.

My question is
1. Is this true fail?
2. If this is true fail, why I have different result between ecc on and ecc off? what is condition cause this situation?

thank you very much.

[David (PassMark)]

You are using the V10.2 release, which is a little old now.

I got fail from mcelog '/var/log/syslog' and the fail device is 0x05

I am not sure how this proves it wasn't the ECC chip that had the failure?

It might also be the case that turning ECC on/off changes some of the timings.

Also strange was the fact that the ECC corrections appear to have occurred right at the very start of the test. Is that consistent, and does it happen both after a cold boot and warm restart?
This might be a sign that there is a BIOS bug, where ECC isn't initialized correctly. See this post for details.
https://forums.passmark.com/memtest8...r-repeat-reset