I need some advice: I have a system that refuses to run without memory errors and the resulting file corruption has forced me to start replacing components until I get the advertised/expected performance. In this case, the DDR4-3600 (CL18) RAM I purchased cannot get through Memtest86 (Test7) without a ton of address errors.

Setup1 Ryzen 5 5600X (OEM tray CPU) MSI B550M PRO VC WIFI: BIOS is dated 9/25/25 KLEVV DDR4-3600 (CL18) : QVL certification confirmed Using the XMP profile 1 option (CL 18-20-20-40, 1.35VDC)

After loading Fedora 43 Workstation and seeing some odd pauses I tried to install Steam and this is when I realized I had some data corruption going via the Terminal stream. Immediately researched and tried to dial down the speed to see if my CPU’s memory controller just couldn’t handle the 3600 speed. Tried 3200 and adjusted the timings down to a standard set that were more appropriate for that speed but then I just got Memtest errors almost immediately (Test 2, 3, 4) so I manually aborted the test. However, if I default back to the auto timings (DDR4-2667 @ 1.20VDC), the whole system passes all of the tests and runs perfectly fine.

As a result I performed the following action: Replaced the Memory sticks with 2 of the same type. No change in test results at either speed.

Online research suggested the CPU/memory controller was most likely the cause so I replaced the CPU with a newer (retail) version and B2 stepping in the hopes it would perform better (see setup 2 below)

Setup2 Replaced the CPU with a Ryzen 5 5600XT (Retail Box CPU) MSI B550M PRO VC WIFI (same Mobo) Used the replacement set of KLEVV DDR4-3600 (CL18) Using the XMP profile 1 option (CL 18-20-20-40, 1.35VDC)

Results were exactly the same with Test 7 being the failure point using the XMP profile and only the default settings (2667 M/T and auto timings) worked with no issues. I also tried other DDR speeds like 3400, 3200, and 3000 with suggested relaxed timings appropriate for each speed and a voltage boost to 1.35 VDC. Tweaking the RAM voltage up didn’t seem to make any difference.

Apologies for the long read so far but now I am at a crossroads with this machine. I have tightened up the DRAM timings to (CL14-16-16-32) at 2667M/T and 1.20VDC and it runs error free and passes Memtest and stress-ng tests in the OS. I have also been able to get really decent gaming performance and no more corrupted files or random crashing using Steam.

So it doesn’t seem to be the CPU or the RAM and the voltages seem OK from the PSU. Should I tear the whole thing apart and replace the motherboard or just stick with what I have since it works (albeit at a lower speed than advertised)? This is one build that has really stumped me. Thanks for reading.

  • deranger@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    3
    ·
    2 days ago

    Does it pass with one stick on XMP timings?

    FWIW, I always set the main timings and voltage manually on my motherboard. My motherboard always gets the last timing wrong. I can’t recall the exact one ATM, but it’s auto set to like 52 rather than 36, something like that.

    • chippydingo@lemmy.worldOP
      link
      fedilink
      English
      arrow-up
      1
      ·
      2 days ago

      Thanks for the feedback. I haven’t tried pulling one stick but I did replace them both so I figured that eliminated them from suspicion. I also tried manually adjusting all of the primary timings and the speed to see if the XMP defaults were just being applied wrong. I think the voltage settings are my most likely culprit in this case so at least I have another thing I can try which doesn’t involve a complete tear down and waiting another week for yet another part to get to me.

      • deranger@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        2
        ·
        2 days ago

        Definitely try one stick.

        It could be the CPU that’s bad - memory controller is on CPU these days. I’m not totally up to speed on AMD voltage names but I’d check whatever CPU voltage is for the memory controller as well as doing a single stick test.

        • chippydingo@lemmy.worldOP
          link
          fedilink
          English
          arrow-up
          2
          ·
          2 days ago

          Good call. I think this post has definitely been worth the effort since you guys are giving me some ideas I wouldn’t have thought of otherwise. Thank you!

      • Shadow@lemmy.ca
        link
        fedilink
        English
        arrow-up
        2
        ·
        2 days ago

        For my problem, pulling one stick resolved 95% of the stability issues even though I knew the ram was good

        • chippydingo@lemmy.worldOP
          link
          fedilink
          English
          arrow-up
          1
          ·
          2 days ago

          That is really interesting and I wouldn’t have considered doing that since I already replaced both of them with a fresh set. I guess only concern with the long term benefit of going down to a single stick would be giving up 16GB of RAM (since this is a 32 GB kit) and then I might get higher speed but I would also get lower bandwidth overall since I lose the dual channel benefit. Weighing that option vs running the dual mode with 32GB at a slightly lower speed makes for an interesting conundrum. I will test this option if I can’t get stable and error free performance from adjusting the SOC setting down. Thank you again.

  • Shadow@lemmy.ca
    link
    fedilink
    English
    arrow-up
    3
    ·
    2 days ago

    I had a ryzen 5900x and Asus mobo that I never managed to get 100% stable. I ended up limping along for a couple years with regular crashes until i got a new box, then I stuck it in a corner and forgot about it.

    I recently came across it again and figured I should sell it, so I started googling and found this post with exactly my same issues - https://rog-forum.asus.com/t5/amd-500-400-series/crosshair-viii-impact-ram-crash-in-windows/td-p/909245

    After setting the soc voltages as per that thread, it’s perfectly stable. So fucking frustrating, it soured me on ever buying an amd based system again and I went back to intel after that.

    Anyways tldr, you might want to play with the same settings me mentioned in that thread since the problem sounds similar. Trying to use the xmp profiles got me nowhere good.

    • chippydingo@lemmy.worldOP
      link
      fedilink
      English
      arrow-up
      0
      ·
      2 days ago

      Thanks Shadow. I had never considered something like this and all of the other suggested fixes I found via forum posts pointed to the physical hardware being the most likely causes. I will try and manually set the SOC voltage tonight to a fixed value (since I am 100% it is set to Auto) and do some more testing.

      Just out of curiosity, what voltage ended up working for your setup? From what I am reading in the linked post and a post which is linked within it, I could go lower (0.9200-0.9600; leaning towards 0.925 VDC) or higher to something like 1.1VDC, but one comment mentioned higher voltage can actually increase instability…ugh. Anyways, thank you for the reply and suggestion!