SOLVED Supermicro X11SCL-F problem with ECC

[Juliean]

Hoping someone can help me. I just finished assembling a server that is to be used for a FreeNAS system. It uses a Supermicro X11SCL-F motherboard, a Xeon E-2124 CPU and a single 32GB stick of Samsung ECC RAM from Supermicro's AVL for this MB. (Links are to the exact items I ordered).

As far as I can tell, all three, MB, CPU, and RAM, support ECC. However, when I started running a CPU burn-in/diagnostic utility, it came back saying that ECC was not available. The MB manual states that ECC can be enabled/disabled from a particular screen (Advanced->Chipset Configuration->System Agent Configuration->Memory Configuration). However, when I go to that screen, the option to turn ECC on or off isn't there. See screenshot attached - manual says the ECC option should be between "Maximum Memory Frequency" and "Max TOLVD".

Anyone have any idea what the problem might be? Some other setting somewhere that I'm not aware of? Could the fact that I only have a single stick of RAM, instead of a pair, prevents ECC (although it's a RAM configuration explicitly supported according to the manual)? The MB doesn't have the latest version of the Bios, it's two revisions behind (1.2 vs. 1.4), but I'm hesitant to update it since the changelog showed nothing that should affect ECC in those revisions.

Any other ideas? This is very concerning, since I specifically built the system to support ECC...

Thanks!

 

[zeebee]

For what it’s worth, I have that same board with an E-2136 and 4 16GB ECC sticks. As far as I know the ECC is working correctly. I can try and take a look at the bios settings tomorrow if you haven’t solved it. I just checked IMPI and I’m on rev 1.3. What diagnostic utility were you using? I ran memtest86 when I built my server last year, and I’m pretty sure it confirmed the ram was ecc enabled, but I could be wrong. I’d like to confirm it is actually working as ecc ram too!

 

[hervon]

Your combo supports ECC without a doubt. My guess is that your burn in utility is wrong. Having the absolute proof that ECC is actually kicking in is not that easy. There used to be a python script in the forum that suggested is ECC was active.

 

[LeDiable]

I have the same board purchased through NewEgg, with a Core i3-9100F and two sticks of 32GB Hynix RAM from Supermicro's QVL purchased from Supermicro directly. I have updated the BIOS on mine, it came with 1.2 and I'm now on 1.4. I just checked the BIOS settingsand on the Memory Configuration screen you mention, I don't have any setting for enabling/disabling ECC.

I haven't gotten the system 100% assembled (need to install boot drives) but I've had it on to update the BIOS, made sure the IPMI version was current, then I ran some of the utilities from the Ultimate Boot CD -- CPU stress test to check heavy load temps for ~36hrs, then both memTest86 and memTest86+. Neither RAM test gave me any message that ECC isn't enabled or that the memory didn't support it. I just fired up both of them again to be sure, they both start humming along doing their thing, but this is my first exposure to them (and ECC memory), so I don't really know what "normal" should look like here.

 

[Juliean]

What diagnostic utility were you using? I ran memtest86 when I built my server last year, and I’m pretty sure it confirmed the ram was ecc enabled, but I could be wrong. I’d like to confirm it is actually working as ecc ram too!

I used Breakin, based on the Uncle Fester guide. I'll try memtest86 next and see what it says.

 

[Juliean]

Running Memtest89 8.4, it reports that there ECC is present and enabled. So I'm going to attribute this to a) a glitch in Breakin, and b) outdated motherboard manual from Supermicro, which shows the "enable ECC" option that they took out of more recent BIOS versions.

On to the next issue: why won't it boot from front panel or IPMI-connected USB (but will boot from back panel USB)

 

[zeebee]

Thanks for the update, glad to hear it's working. It rings a bell with me, as this was my first build with ECC ram, so I was pretty keen to see some confirmation that it was working.

 

[diversity]

I used Breakin, based on the Uncle Fester guide. I'll try memtest86 next and see what it says.

My experience has been that a mere report of ECC functionality is not sufficient.
Myself and others on this forum have found out the hard way.

SOLVED - The usefulness of ECC (if we can't assess it's working)?

Given that ECC functionality depends on several components working well together (e.g. cpu, mobo, mem) there are many things that can go wrong resulting in a user detectable lack of ECC support. However, it's also far too easy to get a false positive for ECC functionality. This might well mean...

www.truenas.com

Through hard work and destroying a few components I now have a memory module that can finally really tell if things are working or not.
The module is broken you see so I could place it in each compatible system and voila. Does ECC work as advertised or not is instantly clear.

But for newer systems I'd advise to go @Mastakilla 's route via over/under clocking/voltaging. It is farrrrrrr more tedious and can only report on single bit errors as far is I am aware of but at least this method has far less chance of breaking components and in theory also works with contemporary systems.

I have been ridiculed here for proposing building a simple memory module housing that can trigger memory faults. I guess that is because we were using desktop cpu's and thus it is ridiculous needing to know (afterall desktop cpu's should not be used for server utitily I guess was the prevelant school of thought). And if one were to use server components then one does not need to make sure.

My stance is that in any case one needs to be able to make sure. Server or not. Component functionally can break over time or be DOA. Even on server components.