I’m having the exact same issue. I got an n100 from ali express, and memtest failed immediately. I replaced my 16GB Crucial RAM stick with another, just to end with the same result.
I now replaced my Motherboard with one from amazon (seems to be a similar no-name chinese n100 mobo) and I still get the same issue.
I cant find any XMP related settings in the BIOS.
Have you found a solution to your issue yet? Any specific settings I need to look out for in the BIOS?
multiple issues on google of this being caused by overheating ram. humor me and put fan on the memory and test again
I already packed up the MOBO and RAM to be sent back to amazon. If it is indeed overheating RAM, that sounds like a bit of an odd issue to have? Would that be fixed with lower clock speeds? Adding fans to the ram seems more of a hacky workaround than an actual fix?
Diagnostic tool.
I have raised a return authorisation request with the vendor on Aliexpress. We’ll see how that goes.
As a practical solution for the present, setting a maximum value of 4600MHz in the memory section of the BIOS is working for me.
As recommended by @Okedokey I’m also investigating what data I can get on RAM temp. I reinstalled the case to get proper airflow and retested the RAM at 4800MHz which resulted in failed Memtest again. This case has a honking great 120mm fan that takes up pretty much the whole front of the case so I feel like I’m getting plenty of airflow across the motherboard.
Unless it has a heatsink it will overheat. Again, google “N100 ram errors heating” and you’ll see those that have applied direct cooling methods (e.g. heat sink / fans) have solved the issue.
Out of interest, this sounds like it’s an n100 issue right? How is it that a specific motherboard boosts the temp of RAM more than others? That doesnt sound particularly healthy to me? The voltage delivered to the stick should be the same no matter the MOBO right?
If I have the time this week I might unpack my RAM and MOBO afterall and give it a try, bit I’m not convinced this is a healthy long-term solution?
As long as the voltage is at the specified rating for the memory it shouldn’t be an issue. Cooling ram hasn’t been an issue for quite a while but is now imho.
Downclocking the RAM would be an alternative solution and shouldnt be an issue with my DDR5 either - if it helps with stability, right?
I mean, DDR4 usually runs at 3200 or something along those lines and from what I can tell the n100 isn’t blazing fast and probably cant make use of the full bandwidth of the RAM anyway, correct?
Edit: Just as an idea if I cant find an appropriate heatsink lying around in my electrinics parts box. Just like Sar6e, I also have a Case fan right infront of the mobo that should be delivering enough airflow
UPDATE:
@Sar6e FYI
Out of curiosity I unpacked my MOBO and RAM once again.
I dont want to jump the gun, but It looks like you’re on to something @Okedokey .
I hacked an Raspberry Pi fan and heatsink onto my RAM stick in an open bench environment (I think that also helps a lot with airflow).
I usually got my first memtest fail error within the first 10 seconds. It’s been a solid 10 mins now and no fails yet. I did also set the RAM frequency to 4800 (the out of box value) instead of keeping it on “auto” in BIOS.
If this works, It looks like I might have to revert my amazon retour and find a better long-term cooling solution for my RAM.
Thanks for the hint regarding the RAM temp, I would never have expected this to be the issue. I’ll let memtest86+ do its think for 2/3 runs and will report back tomorrow.
EDIT: I didnt notice that the temperature is noted at the top left of Memtest86.
The 51° is the highest it got, which doesnt seem like a thermal throttle to me.
I’ll let the test run a few times until lunchtime (another 3h or so). If it runs without errors I’ll try putting the system into the PC case including the additional fan. If that works, I’ll try it in the PC case with only the heatsink.
I think an upgraded downward pointing CPU fan from e.g. noctua should also provide the components surrounding the CPU (e.g. RAM) with enough air as well.
EDIT 2: after around 45mins I startet getting some errors. I’ve now underclocked the memory to 4600Mhz. Lets see how things evolve.
EDIT 3: after over 2h, memtest has completed it’s third pass successfully without errors. I’ve just removed the additional raspberry pi fan and will let it run for another 2 runs. If it s successfull I’ll try putting the system back into the PC case and see if the airflow is sufficient.
Hope it works out
Im wrong or you are using an old version of memtest?
And the temp on top/left Isnt the CPU temp? Not all RAM have temp sensor
You might be right regarding memtest. I’m actually using Unraid which is packaged with this current version of memtest.
Regarding temp, you’re 100% correct, that was my mistake.
Final update:
Memtest86+ ran stable for a few hours outside of the chassis after I reduced the frequency from 4800 down to 4600Mhz.
I put it all into the case, ran it for another hour and it still ran stable.
I now managed to proceed to setup my server (in my case using unraid).
Once my drives are all tested and my RAID is setup I’ll run some furhter tests to check the overall system temp and might upgrade to a noctua CPU cooler if required, but so far it seems stable.
This is just my opinion so take it for what it’s worth… (for the folks reading this now)
If the RAM can overheat making the system unstable, do not just throw a fan on it to keep it stable. It is a good diagnostic test but that just points you to the likely cause.
Let’s say it is heat related, which is exactly what it sounds like. I’m finding it difficult to believe that so many RAM vendors are selling RAM that will not work at this speed. I suspect the RAM voltage is too high on the motherboard (could be poor design or a BIOS update). With that said, if you have no clue what you are doing, then do not do it. I’d drop the RAM voltage one unit of voltage (probably .01 VDC or .1 VDC) and retest the RAM. Keep dropping the voltage until it becomes stable (write it down), keep dropping the voltage until it become unstable (write it down), then split the difference and that is your new voltage level (write it down and maybe write it inside the case for future reference). It is clearly more involved than what I have simplified. And do not skimp on RAM testing. If the testing seems stable, run it for a few days to ensure it is stable. This is not a quick process on a highly adjustable motherboard. I have no idea what the N100 board can be adjusted to.
Okay, regardless of the problem, the solution to add a fan is terrible. If that fan fails, your system is unstable. Do you trust your data after that? Trust the system will remain operational? I would not. (Your data “should” be fine but this is not ECC RAM either")
The easiest fix most people will have here is to underclock the RAM and as previously said, this is not a speed demon machine so underclocking it is not going to perceivably slow the system down.
It would be good if someone could purchased a slower rated DDR5 stick of RAM and see if it ran fine out of the box, however DDR5 lowest speed is 4800MHz (well that sucks).
Unfortunately you are left at manually underclocking. This makes the most sense to me and the recommendation I’d recommend for the N100 platform.
It’s too bad there’s no manual for this board so we can see what options are available.
In addition to the voltage-too-high theory I’ll offer another possibility: the vendor set an overly-infrequent DRAM refresh interval because performance. Or maybe it’s just sloppiness.
I say this because I’ve learned from the overclocking fanatics that holding memory temps down to a bare minimum is vital for maintaining stability in the face of a lengthened refresh interval (which in turn boosts performance as memory refresh gets in the way of actual reads/writes).
Regarding the comment earlier about voltages across different motherboards: it’s kind of messy. The formal JEDEC standard for DDR5 was 1.1V later revised to 1.2V. But the XMP/DOCP memory profiles generally bump this up and sometimes motherboard makers will add their own seasoning to the pot. For an N100 to have too much chili is just bizarre and unlikely but…
