Memtest Errors no longer showing in newer version?

[mt267]

I have been using Memtest86 v5.01 for the last few years successfully, but have been using the free versions of 7.2 (and 7.3) as dictated by newer hardware and DDR4 - but have run into an issue with the newer versions of Memtest, in that they don't seem to show any errors on any sticks of ram I test, even ones previously marked as broken with Memtest 5.01. This is the same for DDR3 and 4 ram, of various sizes and speeds - 7.2/3 gives no errors, every time I have used it.

I am testing server ram, and have been using a variety of Dell servers - with v5.01, I was forcing multi-threading, an option which no longer appears on 7.2/3 - does anyone know if there a configuration setting or similar I am missing, a test unavailable in the free version, or any other reason to explain why the errors found with previously failed sticks from 5.01 are no longer failing in 7.2/3?

[David (PassMark)]

Do you have a record of the number and type of errors you were getting with V5.

If the motherboard supports it then multithreading is available in V7.

If this was server RAM then was it ECC RAM? It would be kind of strange to get a lot of uncorrected errors with ECC RAM.
Are you sure they were real errors? There are known false positive problems with MemTest86+ V5.01 and multi-threading (In fact multi-threading was never officially supported in V5)

[mt267]

Hi,

Unfortunately I do not have a record of all the test results, just the number of errors and passes. I do have a photo of the results for one stick, with 10 passes and 64 errors, I have copied it below:

TST PASS Failing Address Good Bad Err-Bits Count CPU
7 11 0000002d110 - 0.8MB 6c616972 7469625f 18080b2d 55 4
7 11 0000002d118 - 0.8MB 6c630073 5f726165 33116116 56 4
7 11 0000002d120 - 0.8MB 65726373 625f6e65 072d0d16 57 4
7 11 0000002d128 - 0.8MB 74006675 705f7974 045f1f01 58 4
7 11 0000002d130 - 0.8MB 746e6972 6e695c5f 1a07052d 59 4
7 11 0000002d138 - 0.8MB 65670065 696c5f74 0c0b5f11 60 4
7 11 0000002d140 - 0.8MB 67007473 61767465 06760016 61 4
7 11 0000002d148 - 0.8MB 7474006c 72705f79 06045f15 62 4
7 11 0000002d150 - 0.8MB 5f746e69 65726373 3a060d1a 63 4
7 11 0000002d158 - 0.8MB 73006e65 6b636f6c 18630109 64 4

All the RAM we are using is ECC but I am hesitant to say the errors are false positives - issues with servers crashing/rebooting randomly and broken DIMM slots have all been successively solved and fixed from the Memtest v5.01 results, numerous times. Does v7 run multi-threading automatically or does it need forcing? In Version 5.01 we had to hit F2 to force Multi-Thread mode.

[David (PassMark)]

You have been using MemTest86+ V5.01. This is not our software. We develop the original MemTest86, without the plus. Development on MemTest86+ seems to have been abandon years ago.

But the final '+' release was buggy.
Your errors were in Test #7 in MemTest86+ V5.01. There is a known bug in this test 7 with multi-threading.

There are some details here
http://forum.canardpc.com/threads/84663-Memtest86-is-freezing-while-running-test-7
http://forum.canardpc.com/threads/103945-Errors-on-test-7-Skylake-i7-DDR4
http://forum.canardpc.com/threads/97505-5-01-Forced-SMP-parallel-fails-in-test-7-%28block-move%29-on-Dell-Server-bug

So I would think that all those "bad" sticks you found in the past, weren't bad at all (at least the ones with errors in Test #7). Sorry to be the one to tell you.

[orekav]

I have the same issue with Memtest 5.01
DDR4 + I7 7700HQ

0.8 MB too !!!

I have no errors with newer Memtest+

[orekav]

Maybe, Memtest 7.5 do not show that error because the image takes more RAM, so the bytes in 0.8MB sector is not available to test

[David (PassMark)]

Orekav,

As pointed out above, Memtest86+ 5.01 isn't our software. We don't support it, it is a different product, but the author is using the same name. But we know there are bugs in Test 7. The errors you are seeing are not real RAM errors, they are false positives. They are known bugs in the MemTest86+ code. This is also why the error address, 0.8MB is the same as the previous poster. To be honest it is a bit is disaster for the community, as many people will be sending back their RAM as faulty when in fact their RAM is perfectly fine. Huge waste of time for everyone.

This is why you don't see any errors in MemTest86 7.5 (without the plus). It doesn't have the bug.