4 8GB DIMMs in Server motherboard, only showing 24GB, help!

[scsi stud]

Hello,

I have built a 4U server, using a Supermicro X9DR3-F motherboard. I have purchased four 8GB DIMMs, Samsung M393B1K70DH0-CH9 ECC sticks.

Ultimately I will be purchasing a total of 6, so I have 48GB RAM, but they are sold out at certain places and 4 is what I have so far.

When I put all 4 in, in the slots directed by the motherboard manual, the BIOS is only registering 24GB.

I tried swapping a stick in a slot that is working fine with the slot that isn't registering, and it still isn't registering.

Is this normal? Does the memory have to be installed in 3/6's, meaning, do I need to get 2 more sticks for it to register all 48GB?

Any help would be greatly appreciated. Thanks in advance.


*** EDIT: Original issue resolved, however, new issue encountered. See my posts below.

Thanks.

 

[ch33zw1z]

-1 CPU? or 2?

-What CPU?

-Specify which slots you're trying to use...

 

[IndyColtsFan]

No, you can install them in pairs. I have the same motherboard and gradually added RAM a pair at a time and when I added the second CPU, had to change the slots. Which slots are you installing the RAM in?

 

[scsi stud]

Thanks for the replies. I accidentally installed them thinking I had all 6 sticks instead of the current 4 sticks, and followed the 6-stick configuration

I am running 2 CPU's, so it is properly seeing 32GB now. Thanks.

 

[scsi stud]

I'm 'opening' this thread again, as I am now running into another problem. I was successfully able to get the system to detect all 32GB RAM. Now, I have purchased 2 additional 8GB RAM modules (for a total of 48GB), and when I install them in the correct place (according to the reference guide), the system only detects 40GB (i.e., 1 out of the 2 new sticks).

I have swapped sticks/slots and same thing, meaning, it is not a bad stick. Any ideas?

 

[Viper GTS]

Any chance you may have damaged the socket when installing the CPU?

Every time I've seen this it has been damaged pins in one of the sockets. It's shockingly easy to do.

Viper GTS

 

[scsi stud]

Are you talking about CPU socket or RAM socket? If RAM socket, is there any way I could test that socket?

I've installed probably over 100+ RAM modules over the years and never have I damaged a RAM socket...

 

[Viper GTS]

The CPU socket. Remember that memory is driven directly by the CPU these days and those sockets are seriously fragile.

If you're confident you're doing everything right with placement I would pull the CPUs and take a look at the sockets. Shine a nice bright light on it and hopefully your vision is good - They don't need to be very much out of line to drop a DIMM. Varying the angle of the light can make it easier to spot a pin or two that is out of position.

Viper GTS

 

[scsi stud]

Ok, I will give that a shot. In the case that all the pins look good, what's the next course of action?

Is there any chance that the DIMM socket is bad?

 

[IndyColtsFan]

Ok, so you have 6 DIMMS total, right? According to my manual, for a 2 CPU system, the DIMMs should be installed in the following slots:

P1-DIMMA1
P1-DIMMB1
P1-DIMMC1
P1-DIMMD1
P2-DIMME1
P2-DIMMF1

So basically, 4 DIMMs are installed in processor 1 slots (processor 1 is the CPU furthest from the connectors on the back of the board) and 2 in processor 1 DIMM slots. Is this how you have them installed? Is this new RAM identical to the old RAM?

 

[ch33zw1z]

First, Double check the memory plugging order, and reseat the dimms.

Bent memory pins will (imxp) be visible in the slot. The pins can get bent down.

There is also a chance the DIMM socket is bad. It happens.

The CPU pins are easily damaged. Get a flash light and look at the pins from a couple angles. I've found a single pin bent causing issues (hard to find), all the way up to a fingerprint on the pins. Very easy to damage them.

Is all the memory of the same make and model?

 

[Soulkeeper]

Here are some weird things i've encountered before that might apply here:

Heatspreaders on dimms shifted down slightly causing the dimm not to seat all the way.
A single human hair in a pcie slot once caused the card to run at 4x instead of 16x
dust in dimm slot

 

[ch33zw1z]

^^ There are some dirty machines sometimes. I have a couple ESD brushes in my tool bag to clean out slots like you're describing:

http://www.amazon.com/s/?tag=duckduc...ds=esd%20brush

 

[scsi stud]

Ok, so you have 6 DIMMS total, right? According to my manual, for a 2 CPU system, the DIMMs should be installed in the following slots:

P1-DIMMA1
P1-DIMMB1
P1-DIMMC1
P1-DIMMD1
P2-DIMME1
P2-DIMMF1

So basically, 4 DIMMs are installed in processor 1 slots (processor 1 is the CPU furthest from the connectors on the back of the board) and 2 in processor 1 DIMM slots. Is this how you have them installed? Is this new RAM identical to the old RAM?

Thanks for the replies, everyone. I was away all weekend so now I'm back to see if I can get this problem resolved.

Yes, the above configuration is the way I have the DIMMs installed. All the RAM modules are identical (Samsung M393B1K70DH0-CH9).

Even if only 1 of the slots isn't working, should I still reseat ALL the modules?

The first thing I am going to attempt is to blow out the slots to make sure there is no dust/hairs/etc in there. I think that is the least invasive approach. And worst case scenario I will then take out the CPUs to check if no pins are bent.

When I initially installed the CPUs, I did it very carefully, knowing that the pins can easily be bent.

I'll write another post with an update once I continue troubleshooting. Thanks again.

 

[tweakboy]

Im glad all is good now. Enjoy the nice rig you got. gl

 

[scsi stud]

Hi Everyone,

Well I tried blowing some serious air into the DIMM slot and no luck. I then said what the heck and removed one of the CPUs (in the slot farthest from the back of the case/ports), and I couldn't make out any bent pins.

Do you think it's even worth removing the 2nd CPU to see if there are bent pins? Are signs pointing to a bad motherboard slot?

Lastly, I may be inclined to just say screw it and buy 2 more 8GB DIMMs to put in the 7th and 8th slots, and just have "actual minus 8GB" RAM at my disposal. Thoughts?

 

[Viper GTS]

Hi Everyone,

Well I tried blowing some serious air into the DIMM slot and no luck. I then said what the heck and removed one of the CPUs (in the slot farthest from the back of the case/ports), and I couldn't make out any bent pins.

Do you think it's even worth removing the 2nd CPU to see if there are bent pins? Are signs pointing to a bad motherboard slot?

Lastly, I may be inclined to just say screw it and buy 2 more 8GB DIMMs to put in the 7th and 8th slots, and just have "actual minus 8GB" RAM at my disposal. Thoughts?

Have you identified which slot is non functional?

Checking one CPU out of two is only useful if you know for sure that CPU is responsible for the problem slot.

If you check both CPU sockets and both look good I'd swap the CPUs as a sanity check. Assuming the problem stays with the socket/DIMM slot (which it almost certainly will) then you're looking at replacing the motherboard.

A bad CPU is just about the last thing I'd suspect here but it never hurts to verify.

Viper GTS

 

[scsi stud]

Hello,

I wanted to re-bump/open this thread as I have some updates (and some bad news):

I ended up getting a replacement motherboard from Supermicro (did a cross-ship due to time constraints). I did a full swap, and now the problem is even worse. When I have all 6 DIMMs (total: 48GB) in their slots, the computer won't go past the "System Initializing... 2F" screen.

When I remove 2 of the DIMMs to make it 32GB installed, it boots up just fine. The problematic slots appear to be C1/D1. The previous motherboard had an issue with D1.

I swapped out sticks, removed CPU 2 to see if I can register 6 DIMMS with one CPU, and nothing.. same "System Initializing... 2F" issue.

I checked the CPU pins as best as I could with a flashlight and magnifying glass and I don't see any bent pins.

At this point, I am getting extremely frustrated. Having a motherboard register all working DIMMs should be a trivial task.

Any ideas? Is there anything I am missing?

Thanks in advance.

 

[IndyColtsFan]

Did you open a ticket with Supermicro about this issue?

 

[ch33zw1z]

Troubleshooting x86 POST issues:

1. When it doubt, go to minimum config, CPU/RAM/VGA. Only the minimum required amount of memory.

Unplug anything you don't need, including dasd backplanes, cd roms, etc...

Make sure all the power cables are seated firmly.

2. Clear CMOS. I also will hold the power button down for about 60 second.

3. If/when you get it to POST, reset factory defaults in BIOS. Update BIOS to latest level.

This board also has IPMI, ensure that's at the latest level as well.

If no matter what you do, it's unsuccessful, you can also remove CPU's and memory and boot again, even with the CMOS battery out if necessary. I've had server boards choking on an old error, and this gives them a little kick in the pants.

It doesn't work as well with the newer server boards that are UEFI based, but still effective from time to time.

It's frustrating, but a good learning experience.

 

[scsi stud]

Thanks for the replies. I will be calling their support team again today, but I don't know what else they can say. Hopefully they will offer some good insight.

I will follow your x86 POST troubleshooting steps and will report my findings. The only addon card I have is an LSI RAID adapter. The BIOS is the latest version (v3.0) from what I am seeing.

Any other tips/suggestions/advice would be greatly appreciated.

 

[ch33zw1z]

Take the LSI out while you try to get through POST.

 

[NS03]

Did this ever get fixed as I am having the same issue

 

[PliotronX]

Might sound crazy and may not apply to SM, but my experience with Dell boards has been that they tend to not like dual rank x4 sticks. Have you any 2Rx8 sticks to try?