Resolving potential hardware issues within Linux

BigBadBeef

Well-Known Member
Joined
Sep 23, 2021
Messages
507
Reaction score
316
Credits
4,682
I got a PC with 4 sticks of DDR5 RAM. Exploring social media, it would seem that this in, and of itself, appears to be a problem. It began with not being able to get neither EXPO or XMP to work with a 9800X3D, I am stuck using 5600MHz RAM at 4800MHz.

It is slowly becoming clear to me, that this particular iteration of memory is [insert foul language]. But still, this is what I've got, and this is what I need to work with.

Honestly, the RAM speed doesn't bother me too badly, but a recent turn of events have convinced me that it might be a symptom of a bigger issue. You see, my brand new (hasn't even lost the smell of fresh pc components yet) computer is experiencing what I could best call "periods" of considerable instability. We're talking about complete system freezes or straight up reboots - no kernel panic.

And I do mean periods. It will run fine for a while - days, weeks etc., but then there will be a day when the aforementioned begins to happen. The last "phase" was this past weekend, it even crashed once during me watching YouTube. So during one of those periods, I ran memtester in order to ascertain whether this was a RAM issue, and lo and behold - it started throwing out memory errors just before actually triggering one of the crashes.

Well, it's time for memtest86 and testing each stick individually. This is something I will reserve for the next weekend. In meantime, I wish to plan for the contingency that using 4 stick of RAM is the problem rather than any individual stick being defective.

I realize that there might not be anything I could do to stabilize the RAM within the Linux environment itself, my line of inquiry pertains to more like something I might be able to with RAM settings in the BIOS menu.

Speaking of which, a user I've talked to in social media claimed, that using 4 sticks of DDR5 RAM may require more voltage, and raising by 0.1mV would help stabilize it. I'd like know your opinion about his claims. I would also like to know if anyone here is using 4 sticks of DDR 5 RAM that were unstable and managed to stabilize them.
 


I got a PC with 4 sticks of DDR5 RAM. Exploring social media, it would seem that this in, and of itself, appears to be a problem. It began with not being able to get neither EXPO or XMP to work with a 9800X3D, I am stuck using 5600MHz RAM at 4800MHz.

Speaking of which, a user I've talked to in social media claimed, that using 4 sticks of DDR5 RAM may require more voltage, and raising by 0.1mV would help stabilize it.
I look up your motherboard -> Gigabyte X870E Aorus Pro Ice
There it says this for memory.
Support for DDR5 8200(OC) / 8000(OC) / 7800(OC) / 7600(OC) / 7200(OC) / 7000(OC) / 6800(OC) / 6666(OC) / 6600(OC) / 6400(OC) / 6200(OC) / 6000(OC) / 5600(OC) / 5200 / 4800 / 4400 MT/s memory modules
Behind all the higher ones(including 5600) it says OC. I would think that means "Over Clock". So I would think that would you would have to overclock something to get to that speed.
 
AMD EXPO is a predetermined set of clock and voltage settings that allow the memory to reach the speeds it is meant to. This is around for a LONG time now, I even had that on my old PC that let my old DDR3 RAM run at the 1600MHz it was supposed to.

I don't know why is it even still considered an overclock and not "boosting" like they've been saying it with CPU's and GPU's for years now. It's just corporate jerks and their buzzwords.
 
Screenshot_2026-02-17_17-40-41.png


Well, I've removed two sticks and retested. I can indeed assert, that the problem lies in either using 4 memory sticks or one of the two absent memory sticks being defective.
 
Can you share screenshot or photo of the home screen of your BIOS, where it shows your system information?
 
Yeah, I can do that, just not right now. I'm currently working and don't any vacation lined up. The best I can do is squeeze in some time for memtesting this weekend.
 
just wanted to mention, even IF you DO find issues with memory - don't despair.
I had previous gen of AMD CPU with 64 gig of ram.
i have established I have memory issues, PERSISTENT memory addresses were failing. Faulty parts of memory stick are likely culprits.
On Linux there is a way to deal with this, you need to enable the feature.
Kernel at start can make quick test of RAM, identify said memory regions and block CPU from using those, by marking them as used by itself or some software, whatever.
Yeah you lose a few KILOBYTES of ram(less than percent!), but it otherwise cost you nothing but a few seconds on start up.(and I personally, reboot my Linux PC once every a few weeks, sometimes months, I do use RAM suspend)
That literally eliminated problems that plagued me for years with those 64 gigs of ram. And I was not shy of overclocking my RAM, etc. Worked like a charm.
 
I did some googling and AI searches.
it was a while when I used it. I wish i made bookmark for that. I have new machine since.

But I believe what you need to do is this:
1. Enable this in kernel, recompile/install as needed, dont forget to update grub, as needed.
CONFIG_MEMTEST
Location: │
│ -> Kernel hacking │
│ -> Kernel Testing and Coverage │
│ -> Memtest

This enables kernel to do test.

2. specifically ADD, this to your existing boot parameter in grub:
memtest=4
This forces kernel to use ability from step one.

THIS WILL make start up time longer, may be half minute or so, depending how much ram you got and how fast it is. I had 64 gigs of ddram3 , so half minute or so was waiting time for me.

But it was worthy price to pay to have ultra stable machine WITH lightly overclocked RAM :)

There is better way to do this too. you can get memtest86 to tell you areas that need to be blocked and tell kernel to block those areas, manually, but there is one draw back to this.
If overtime RAM situation becomes worse and you lose new rap areas? you need to do it again and again.

My first method is simpler, since its automated, you need to set it up once and once a few months reboot PC. Every time PC starts - it checks for ALL of memory faults, old and new. marks them as "used" so cpu rnadomly does not assign those memory ranges to working programs. ??? PROFIT!
 
and sweet part - did not need to change ram! yeah some KB of RAM lost and longer bloading times, but small price to pay for not buying new ram and opening computer and changing it, etc.
Especially considering its linux, you barely need to reboot it AND no insurance against new ram being faulty :D
you owe me a beer ;)
 
Not really since my RAM is under warranty. If any of it is defective I can just send it back for replacements.
 
1000000340.jpg


Whoa whoa whohohoa! I knew I'd find something, but I had no idea I'd turn out this dramatic. The memory failed catastrophically within the first 30 seconds of running the test.

But two of the four sticks passed with flying colors, but I still can't get expo to run on them.
 


Follow Linux.org

Members online

No members online now.

Top