ECC error

[MichaelBatz]

My build.

Motherboard:
Supermicro X10SL7-F
CPU:
i3-4330 Processor (4M Cache, 3.50 GHz)
RAM:
2xCrucial 8GB DDR3 PC3-12800 Unbuffered ECC 1.35V 1024Meg x 72 (CT102472BD160B)
2xSamsung DDR3-1600 8GB/1Gx72 ECC CL11 (M391B1G73QH0-YK0)
Power supply:
SeaSonic G-450 GOLD 80Plus – 450W
HDD's:
4xWestern Digital Red NAS Hard Drive WD30EFRX 3TB
Case:
Fractal Design Node 804


As you can see, i have 32GB of RAM. Which might be a little generous. But, the reason why i have this much RAM, is because. The first RAM i bought was the 2xCrucial 8GB. According to Supermicro's homepage, they weren't on the compatible list, but they were on sale in my contry and have seen a lot of users with the same motherboard as mine, so i thought "Why not?" I installed my 2 new Crucial RAM in my motherboard according to the manual.

DIMMA1 -- Crucial 8GB (CT102472BD160B)
DIMMA2
DIMMB1 -- Crucial 8GB (CT102472BD160B)
DIMMb2

With this configuration my server would reboot sporadic. I found out, by inspecting the IPMI log, that i had severel ECC error.

I took my RAM out. Installed them like this.

DIMMA1
DIMMA2 -- Crucial 8GB (CT102472BD160B)
DIMMB1
DIMMB2 -- Crucial 8GB (CT102472BD160B)

No problems whatsoever. I contaced Supermicro about this. They told me, that the RAM slots to populate first, was DIMMA2 and DIMMB2, but i could always try to buy RAM which was under their compatible list. Luckily for me, some compatible Samsung RAM got on sale in my country, which i bought 2 of.

My configuration as of now

DIMMA1 -- Samsung 8GB (M391B1G73QH0-YK0)
DIMMA2 -- Crucial 8GB (CT102472BD160B)
DIMMB1 -- Samsung 8GB (M391B1G73QH0-YK0)
DIMMB2 -- Crucial 8GB (CT102472BD160B)

I haven't had single problem with this configuration. My server has been running for about 3 months.

Supermicro said, that it could be related to timing issues. Does anybody else, have an insight regarding this??

More importantly, is this, something that i should worry about? I'm planning to upgrade my HDD's in the near future. I want to make sure everything is as good as it gets. I have been using the last 3 months to familiarise myself with FreeNAS and ZFS through my server and VM.


Note self: Do not buy server hardware which isn't on the compatible list!!

 

[cyberjock]

It certainly could be due to timing issues. Generally, it is recommended you buy the exact same RAM for all of your slots on the server. Mixing brands and models can cause timing issues.

I'd say it is something to worry about. If you didn't have ECC RAM you'd actually have experienced corruption. So definitely not something to ignore. ;)

 

[MichaelBatz]

Okay, i will keep that in mind. I might try and sell my Crucial RAM's then. I don't really need 32GB of RAM.

But, i guess that, neither the motherboard or RAM is defect, they are just incompatible?

 

[Ericloewe]

That RAM is on Supermicro's QVL. The DIMMs should have a Micron sticker on them. Details on all this on the Supermicro X10 RAM sticky.

So, you have a hardware issue, most likely. Either the RAM, motherboard or CPU.

 

[cyberjock]

Okay, i will keep that in mind. I might try and sell my Crucial RAM's then. I don't really need 32GB of RAM.

But, i guess that, neither the motherboard or RAM is defect, they are just incompatible?

No way to know.

That RAM is on Supermicro's QVL. The DIMMs should have a Micron sticker on them. Details on all this on the Supermicro X10 RAM sticky.

So, you have a hardware issue, most likely. Either the RAM, motherboard or CPU.

Not necessarily. As I said above, mixing and matching brands and models can cause problems. The hardware may all be fine. It may just be that the two different types of RAM don't really work well with the motherboard when installed.

 

[Ericloewe]

No way to know.



Not necessarily. As I said above, mixing and matching brands and models can cause problems. The hardware may all be fine. It may just be that the two different types of RAM don't really work well with the motherboard when installed.

His faulty case is only with the Crucial DIMMs. The problems went away with the Crucial DIMMs on the A2/B2 slots, with or without Samsung DIMMs.

 

[MichaelBatz]

His faulty case is only with the Crucial DIMMs. The problems went away with the Crucial DIMMs on the A2/B2 slots, with or without Samsung DIMMs.

That's correct.

Don't know why i haven't seen your Memory Recommendations sticky before. I can see that you mention, Crucial 8GB (CT102472BD160B) being a rebrand of the compatible Micron. I just had my one of my Crucial RAM out, the part number on it said: MT18KSF1G72AZ-1G6E1ZF, the part number for the Micron says: MT18KSF1G72AZ-1G6E1.

Regarding hardware issues. I have tested the CPU in my own PC, nothing there. I have tested both my Samsung and Crucial RAM with Memtest, nothing there. It's only when Crucial are in A1/B1 i can provoke the error.

If i should continue using my current setup, would i risk everything there is so nice about ZFS, becoming obsolete?

 

[cyberjock]

Sorry, I guess I should cite my evidence...

http://www.supermicro.com/manuals/motherboard/C222/MNL-1463.pdf

The manual says you should populate A1/B1 first. (I knew before I went to the manual you should always install to bank/channel 1 first) No clue why they'd tell you something else on the phone.

So there's more funky crap going on than what it appears on the surface.

Yes, it is well know that if you populate the slots out of order it can cause problems. It's also not guaranteed to cause problems.

Yes, it is well know that mixing and matching RAM models and brands can cause problems. It's also not guaranteed to cause problems.

It is also possible that the RAM itself is not compatible because of timing, and moving them to A2/B2 somewhat rectifies the problem.

In any case, I tend to think its an incompatibility somewhere and not actual bad hardware.

 

[mjws00]

If you hunt there is a similar issue with someone mixing Samsung(?) and Kingston RAM. Box threw an error with Samsung in slots 1, they swapped the Kingstons to slot one instead and it ran perfectly. Forgive the details that is meant as an overview the specifics. Definitely timing and compatibility issues in play that depend on slots utilized.

Coinflip for me on how to proceed. If the box is stable and burned in with the Crucial in A2 B2, I'd probably run it. At that point it is fully tested and populated. I'd chalk the glitch up to experience and move on. It's not the first mobo that is a little picky on memory and it won't be the last. That said, at the slightest memory glitch I'd punt the Crucial and populate with all Samsung. No point running gear you can't trust.

 

[Ericloewe]

From what I can tell, both MT18KSF1G72AZ-1G6E1ZF and MT18KSF1G72AZ-1G6E1ZG are present as Crucial-branded stuff and both are seemingly problem-free, generally.

Since Supermicro doesn't list the suffix, I'm assuming it's a factory identifier or similar.

ZG says made in China, so I'll check mine and see what they say. @MichaelBatz - can you get us a picture of your DIMMs with the Micron sticker?

 

[Ericloewe]

..and it just gets more complicated:

My DIMMs are MT18KSF1G72AZ-1G6E1ZE. Made in China, as well. Production date 21st week of 2014 (~late May).

The actual chips are labeled 4JE77 D9QBJ, same as a ZG DIMM.

 

[cyberjock]

..and it just gets more complicated:

Oh yeah.. well... your face is stupid!


Yes, I had to throw some comedy in there. ;)

 

[MichaelBatz]

Been a little busy the last couple of days.

(Forgot the production date. They are from 201441)

Thanks the replies everybody. I have decided to go along with my server as it currently is. I'll just have to watch it a little more closely, but that's okay!

 

[Ericloewe]

The chips themselves seem to have the same label in E, F and G DIMMs.

So, there doesn't seem to be a physical difference between them.

 

[Bhoot]

Well if I may add..
I have an asus board with 2x16gb crucial ecc ram. The asus manual gives a method of populating the 8 dimm slots provided in their user manual. When I populated them as per that the system didn't POST. I then tried a few weird combos and now I got the same dimms to work on a very different configuration to what the manual offered. I am in no way denying that ECC shouldn't be mixed and matched, but I think maybe switching the slots might help it POST.
Points to note
1) the asus board also doesn't list the Crucial RAM as compatible (CRUCIAL 16GB DDR4-2133 1.2v RDIMM 288p (CT16G4RFD4213))
2) my friend who deals with computers assured me that asus and crucial ecc would work and it did.