RAM issue or something else? Help please

[favreervaf]

Hello, I started looking into memory because I've been getting some game crashing that I'm not use to. That being said, they are newer games.

Example: Battlefield 6 - crash to desktop. Major issues if you google in and look on reddit.
Marvel Rivals & Oblivion Remastered - getting UE5 engine crash with an error log that doesnt make much sense to me. I send report to UE. (Unreal Engine)

I am NOT getting blue/black screen of death. I've ran all the sfc /scannow, and all those things. Have most up to date BIOS, Windows 11 updates, and AMD Adrenalin drivers.

These are also not AMD driver crashes as 1. AMD informs you there was as crash. 2. My AMD settings get reset when that happens and its not happening with this.





i7-12700KF
9070 XT
Corsair RAM, DDR5 CL40 4800 MHz (XMP = 5200) This should be stable as my motherboard states it can support 6200.




So I ran Memtest86 for the first time ever. There were 3 errors total, they were in the final 4th run. I doubt it was due to the test running longer and temps because the DIMM temp was like 42C.


So I’m thinking- gosh, only 3 errors seems so weird. I wonder if it’s a fluke. I did another Memtest86 the next day and all 4 runs Passed with zero errors. This is all with XMP on, as it was the first failed time. Then later in the evening I ran Memtest64 (made by TechPower Up) on Windows and that had no errors after 10 loops. I then did another Memtest86 later at night to be sure, and again all 4 runs Passed with zero errors.




So the RMA I had set up with Corsair, I emailed back and said maybe it was a fluke I may hold off.... Had another UE5 crash today playing Marvel Rivals, back to desktop - then had some spare time and thought I'll run another mem test or 2 - I ran 2 more Memtest86 tests, 1 single error in both of them. Usually around test 7, could be in the 2nd or 3rd run.




So now I'm conflicted... can someone tell me what to do? How can something like bad physical hardware show errors sometimes, but not others when none of the parameters changed?

Do I RMA the RAM? I really don't want to be without a PC for a few weeks. Is it a multicore false positive thing? Should I run a single core Memtest86 now overnight? Are the game crashes coincidental and related to their engines?




Any advice appreciated! Thanks!

[David (PassMark)]

The brand name RAM manufacturers test their RAM before selling it. So they find all the really serious problems. But they need to make money. So testing times are minimized.
For example, two days ago an unnamed RAM manufacturer from India told us that the 14min cycle time for testing a RAM stick was too long and it needs to be shortened. To even get it to 14min they had already disabled several tests. But there is an obvious trade off between testing time and quality.

After testing what's left is mostly good RAM and RAM where the stick is marginal (i.e. worked well enough to get through a single 14min test, with some arbitrary external conditions, but would not be OK over several days of load, or maybe fail under higher temps or with EMI exposure). This company also wasn't testing the XMP profiles, just the slow JEDEC profiles.

Once you know how RAM works, it is easy to image how it can be marginal. Timings need to be precise to the nanosecond. But in all electrical systems there is jitter and drift. Plus there are external EMI and temperature influences.

Anyway, your RAM is very likely bad.
But I should also note that your CPU is only rated to DDR5 4800 MT/s. So by using XMP you are running it out of spec. So you could try 4800MT.

[favreervaf]

Thank you - someone from Tom's forum suggested upping the voltage to 1.35V from the 1.25V that is what auto is making it from XMP. It terms XMP, I feel like running it out of spec from 4800 to 5200 is so extreme, is it? Motherboard also said it supports 6000. That being said I ran another Memtest86 after the increase to 1.35V and it passed all four runs. But- it has passed before at 1.25V. So not sure - think I should just RMA this or run another?

[favreervaf]

Just ran another and it passed again all 4 runs. Thats with XMP on but upped to 1.35V.

[David (PassMark)]

Yes, sometimes voltage and/or timing changes can get marginal RAM to be more stable. While MemTest86 can give you a reasonable degree of confidence, it is impossible to be 100% sure as to the long term stability. Especially if you are running it outside of the specs.

But if it isn't (any longer) broken don't fix it