stressapptest shows FAIL result but memtest86 can't detect any error

freeyssu

7/21/24, 6:03 AM

I have a question about DIMM test (stressapptest and memtest86). I am using AMD Threadripper with 4x16GB Samsung ECC memory.

stressapptest obviously designates there is something problem and this is re-producable every tries.

2023/07/20-20:31:57(PDT) Report Error: miscompare : DIMM Unknown : 1 : 5s
2023/07/20-20:31:57(PDT) Hardware Error: miscompare on CPU 1(<-17) at 0x7f68a6895b00(0x4673aab00:DIMM Unknown): read:0xffffffffffffffff, reread:0xffffffffffffffff expected:0x4000000040000000. 'walkingOnes64'.
2023/07/20-20:31:57(PDT) Report Error: miscompare : DIMM Unknown : 1 : 5s
2023/07/20-20:31:57(PDT) Hardware Error: miscompare on CPU 1(<-17) at 0x7f68a6895b08(0x4673aab08:DIMM Unknown): read:0xffffffffffffffff, reread:0xffffffffffffffff expected:0x2000000020000000. 'walkingOnes64'.

It looks memory address 0x4673aab00 got miscopare error.

However, memtest86 shows PASS result every tries. I did full test(almost a day). What should I do for next step to confirm the suspected DIMM error?

0 + 3

memory

tsc_chazz

7/21/24, 6:08 AM

As an elder hardware geek, first thing I'd do is change all your DIMMs - not new, just shuffle them so they're all in different sockets - and see if the error location changed. If the error stays in the same place, it may not be memory. (I have been bitten by a Ryzen CPU error...)

freeyssu

7/31/24, 6:46 AM

yeah, I need to reseat all DIMM and suffle them all. However, I would like to indentify the problematic DIMM with more genius way.

tsc_chazz

7/31/24, 3:01 PM

The problem is that we don't know how the northbridge chip maps DIMMs to addresses; you haven't told us which specific Threadripper CPU you're using or which motherboard, and even with that it's possible to have BIOS settings affect the mapping of memory to DIMM. Honestly, the simplest way to check this is swap DIMM 0 and DIMM 2, see if the fault address moves; if not, swap DIMM 1 and DIMM 3, if it still doesn't move it may be CPU or motherboard.

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: stressapptest shows FAIL result but memtest86 can't detect any error

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.