Memory Issues?

[dturner71.dt]

Just got my new system up and running.
Supermicro MB
128GB ECC Memory View attachment 22315
Dual Xeon E5-2650 v2
LSI 2810 Controller
8 x 6TB Drives (Z2)

Created a iSCSI target @ 10TB. Mounted and started a data transfer. After 5 minutes, server stops responding and I see this error on the console :
Bad memory stick on bank 7???

I did a 24 hour memory burn in with memtest86. 3 complete loops no issues so i;m at a loss why his is an issue.

Thanks

 

[Nick2253]

That seems to be showing a memory error, but I would believe that it should have been caught by ECC.

What version of memtest86 did you run?

 

[SweetAndLow]

With that much memory you will probably have to test it for a week. I know it too my system forever to get a full memtest run and I have 128GB also.

 

[dturner71.dt]

Memtest86 v4.10

I'm running a Supermicro X9DRI-F motherboard. Now I have to fins out which module is Bank 7.

https://www.ebay.com/itm/Supermicro...e=STRK:MEBIDX:IT&_trksid=p2060353.m2749.l2649

Thanks for the quick response gentlemen.

Dustin

 

[dturner71.dt]

Anyone have insight on how to identify the different memory banks? The motherboard document does not reveal this info. Unless I'm missing something?

Thanks.

 

[Redcoat]

Anyone have insight on how to identify the different memory banks? The motherboard document does not reveal this info. Unless I'm missing something?

Thanks.

I this long thread I got help from moderator @joeschmuck in tracking down a reported memory issue from the dmidecode output. Perhaps you can use the suggested methodology in your investigation.

 

[SweetAndLow]

does ipmi tell you banks? Could just guess and remove sticks.

 

[danb35]

Anyone have insight on how to identify the different memory banks?

Check the IPMI event log, either through the BIOS or through the IPMI web interface (or, presumably, by using IPMIView).

 

[dturner71.dt]

Event logging was turned off. Ugh.

Turned it on. Let's do some testing.

 

[dturner71.dt]

Thoughts?

Thanks.

 

[Ericloewe]

Move the DIMMs around and see if the error follows the DIMM or if it stays with the slot.

 

[Nick2253]

Memtest86 v4.10

You are not using Memtest86, you're using Memtest86+. Today, they are separate tools, though they both descend from the original Memtest86 v3 code c. 2002.

If you use Memtest86+, you need to use 5.01. Ivy Bridge CPUs were not properly supported in the older versions, and I'm wondering if that's part of your issue. Memtest86 is more updated (Memtest86 v7 supports v6 Xeons and Apollo Lake), but PassMark has made the code proprietary, so it's not open source.

 

[dturner71.dt]

Would Memtest86 7.4 Pro be able to test ecc?

 

[diedrichg]

Memtest86 v4.10

View attachment 22319

A little off topic; why does it show the E5 as a i5/i7?

 

[Nick2253]

A little off topic; why does it show the E5 as a i5/i7?

Because he's using a version of Memtest86+ that does not properly support Ivy Bridge CPUs.

 

[Nick2253]

Would Memtest86 7.4 Pro be able to test ecc?

Yes. However, you don't need the Pro version. The free version works well enough.

 

[dturner71.dt]

Ok. Thanks.

 

[Redcoat]

Would Memtest86 7.4 Pro be able to test ecc?

It depends what you mean by "test ECC". See the thread I pointed you at earlier for my experience with the Pro version and fault injection for ECC testing. If you don't want ECC fault injection, don't spend the money for Pro. If you do want it, get in touch with Passmark and make sure your hardware/bios/configuration will support it before you spend the $.

 

[Chris Moore]

Did you pull all the memory out and put it back in? That alone has solved more errors than I can count.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[dturner71.dt]

I moved the module in question. Then got an error on another module. Moved both of them again and they both errored out.

Pulled both modules rom system and moved 2 modules into the memory slot to make sure it was not the slots. Transferred data for good hour before shutting down.

Running memory test on it now for at least next 24 hours to be sure no other modules are bad.

Thanks again guys.



Sent from my SM-N950U using Tapatalk