I registered here specifically to comment on this thread. Perhaps my post could save somebody else some grief and frustration.
OP and I have a some things in common. Recently, after a decade of service, I decided to demote my Core i7 2600k-based machine (what a great CPU!) to the status of a secondary PC.
My new PC is configured as follows:
AMD Ryzen 9 5950X
ASUS Crosshair VIII Dark Hero
128GB Corsair Vengeance LPX 3200MHz CL16 RAM (two CMK64GX4M2E3200C16 v3.40 kits, being four sticks of 32GB each)
The system is not overclocked in any way, as I prefer stability and reliability over performance.
Shortly after building the system, I began seeing data corruption. This first manifested in Epic Games Launcher automatically re-downloading titles after the verification phase (suggesting that there is data corruption) or Steam indicating that some installed titles are having corrupted files. I also saw a BSOD, and after inspecting the minidump file it turned to have originated in a very unexpected place (the call stack indicated that the call in question originated within the kernel). Then, while browsing the web, Chrome would display an error with the status of "STATUS_ACCESS_VIOLATION" (which to me would seem like a corrupted pointer value that was read from memory). All of this served to raise my suspicion level considerably.
I also noticed that a corruption issue would reproduce every single time if I re-downloaded a specific title from the Epic Games Store, triggering a post-verification redownload over and over. This probably happened because the data of the title in question contains patterns that are more likely to trigger the defect that my RAM had.
I initially lowered RAM frequency (from 3200MHz to 2666MHz) to see whether that would help, and while it did to some extent – the data corruption would now reproduce once in three downloads, instead of every download. I tried to update the BIOS to the latest version (3302) released on March 10. The change log stated "Improved memory performance", so I thought this could have a positive effect on my situation. It did not, however.
Right there, I was puzzled. Just like what OP describes, memory testing would yield nothing. I then downloaded Memtest86 by PassMark. Upon running that test, initial results were fine, that is until the Hammer Test (#13) would be attempted. There I would first see a note stating that "RAM might be vulnerable to high frequency row hammer bit flips". Some cycles later, concrete errors would start mounting up. Oddly enough, one could see an odd address distribution there: about 50% of the errors have occurred on a single address. That address would continue to "star" in subsequent test passes, even after rebooting (which made sense, as an OS with address abstraction/randomization/virtual memory is not involved here). Additionally, these errors would all happen in the same 36MB-big memory region (around the 23.3GB mark), and all seemed to be single-bit flips, too; I could never see corruption in more than one bit at a time (i.e. expected value: AE9E5412, actual: AE9E5410 – the difference between these is just one bit, specifically the second-LSB of the most significant byte [corresponding to 12 and 10 here]).
All of this was highly suggestive of a single defective RAM stick.
I then started testing RAM pairs (B2, A2 and B1, A1) by elimination, and was able to identify the offending pair as being installed in DIMM slots B2, A2 (since I purchased two dual-stick RAM kits, I would have to RMA both the offending stick and its same-kit counterpart. I therefore didn't care which of the two sticks was the culprit). To make sure that there is nothing wrong with other components, I moved the good pair from DIMM slots B1, A1 to DIMM slots B2, A2 and ran the Hammer Test once more. The good pair did pass the test once more (no "RAM might be vulnerable to high frequency row hammer bit flips" notes and no errors). Downloading games would no longer trigger post-verification downloads, too.
Reading PassMark's FAQ, it does seem like the Hammer Test gets some non-black-and-white special treatment there (for the curious, see:
https://www.memtest86.com/troubleshooting.htm#hammer). Specifically, the FAQ states that in the event of Hammer Test errors, mission-critical RAM should definitely be replaced. However, home users may also choose to let these error slide, depending on their willingness to live with the possibility of these errors manifesting themselves as real problems. PassMark also suggests that if one does choose to replace the offending RAM, they may consider RAM modules of different model as it is likely that RAM modules of the same model would also fail the Hammer Test.
In my experience, however, the possibility of these errors manifesting themselves as real problems is very real. Moreover, my other RAM kit does seem to pass the Hammer Test without a problem, and there is no need to choose RAM modules of different kind (this might be an isolated case, however).
Having RMA'd the defective kit, I am expected to receive a replacement kit around in the Monday-Tuesday range. I will post another update after I am done with all the testing.
P.S.
@Bo$$ Could you please write how you ended up resolving your situation? Did swapping RAM help?
Thanks!
