11 replies to this topic

[goatsniffer]

Hello and thank you in advance for your help.

I am putting together some server hardware and I am having a strange issue.

The Issue: When both processors are installed system will power on, the fans will spin up, but system will not post, there is no video signal, and there are no audible warnings. On my SUPERMICRO 833T-650B Chassis there is a front pannel which illuminates a solid temperature warning light when this occurs. The motherboard manual says an audible warning should accompany a thermal event, but there are no beeps.

The System:
Motherboard: SUPERMICRO X8DTL-iF-O Dual LGA 1366
Chassis: SUPERMICRO 833T-650B w/ included PSU (4-to-8 pin EPS power adapter used to supply second EPS plug)
CPUs: 2 x Xeon E5620 Westmere 2.4GHz 80W w/ XIGMATEK Durin D982 Ultra Low Profile Cooler
RAM: 2 x Patriot Signature 12GB (3 x 4GB) DDR3 1333 ECC Registered Server Memory
RAID: Areca ARC-1220 w/ ARC-6120 Battery

System drives are not connected so I will leave out that information.

What I Have Tried:
Remove all add-on cards - No Post (same error)
Remove all but one DIMM - No Post (same error)
Remove all DIMMS - No Post (audible warning for no memory)
Reseat CPUs - No Post (same error)
Remove CPU2 - Post

My Notes:
I got the system to post without error using a single CPU in the CPU1 socket and one DIMM.
I added the RAID card and a second DIMM and it again posted.
I added a third DIMM and it again posted, but only the capacity of two DIMMs was recognised.
I set the BIOS to full boot instead of quick and had the same results with the memory.
The memory gets downclocked to 1066 as expected with the Xeon E5620 CPU.


Any ideas? Need any more information?

Thank you again.

[Nagisan]

So it posts with 1 CPU but not 2, and 2 DIMMs but not 3?

Sounds like you may have a bad motherboard.....I don't have much experience with multi-CPU systems, but if it has 2 installed and one isn't working, I don't think it would POST properly.

Try putting the other CPU into the first slot (take out the one thats installed right now and put CPU2 into CPU1 socket), that will help rule out a bad processor.

Also, try each stick of memory in the first memory slot one by one, to rule out a bad stick of ram that isn't being recognized.

[goatsniffer]

Thank you.

I am going to try the other CPU in the CPU1 slot when I get a chance today. That should rule out a bad CPU.

The system does post with all three DIMMs for CPU1 populated, but the BIOS states 8GB of memory instead of 12GB. Rearranging the three DIMMs yields same 8GB reading, trying them one-by-one posts and reads 4GB. I even tried the second set of memory the same way, the system behaves in exactly the same manner.

Yes it could be a bad motherboard, but I need to be sure because otherwise the replacement could act in the same exact manner and I'll be out additional time and money.

[Nagisan]

goatsniffer, on 19 November 2010 - 16:31, said:

The system does post with all three DIMMs for CPU1 populated, but the BIOS states 8GB of memory instead of 12GB. Rearranging the three DIMMs yields same 8GB reading, trying them one-by-one posts and reads 4GB. I even tried the second set of memory the same way, the system behaves in exactly the same manner.

Hmmm, server motherboards should support more than that, but are you positive it supports that much ram in that few slots? Some motherboards can be pretty picky with ram. Have you tried using different ram slots? Might not work if you use different ram slots though because it might require certain slots to contain ram before other slots check for ram......really an odd issue, it sounds like it's not supporting that much ram, but server motherboards are generally designed to support much more ram than desktop motherboards so that shouldn't be the issue.

goatsniffer, on 19 November 2010 - 16:31, said:

Yes it could be a bad motherboard, but I need to be sure because otherwise the replacement could act in the same exact manner and I'll be out additional time and money.

I fully agree with that, I spent a week troubleshooting and trying things before setting up an RMA for a motherboard so I didn't spend that week waiting for a new motherboard that might not have worked anyway. Just saying a bad motherboard is the most likely at this point in time.

[goatsniffer]

The motherboard has six RAM slots supporting up to 8GB ECC Registered per DIMM in each slot.

I agree, it does seem to me to be a bad motherboard. I have a this thought in the back of my mind that it could be the memory though. The board supports DDR3 1333, but the Xeon CPU does not. Though the board downsamples the RAM I have a this feeling it could be causing some issue. Though that doesn't explain why it boots with one CPU and not two. I would also assume I would get a bad memory audible warning at post instead of nothing but fans.

Am I right in the idea that pre-post issues are usually due to CPU problems, whether it's the motherboards problem with the CPU or the CPU itself?

I am going to try the CPU swap, and if that doesn't work I'll start the RMA unless anyone has another suggestion.

[Nagisan]

goatsniffer, on 19 November 2010 - 18:26, said:

Am I right in the idea that pre-post issues are usually due to CPU problems, whether it's the motherboards problem with the CPU or the CPU itself?

I am going to try the CPU swap, and if that doesn't work I'll start the RMA unless anyone has another suggestion.

Usually, yes, you could also go buy a PCI Diagnostic card before RMAing it, thats what I did. Only cost about $10 but essentially I can plug it into a PCI or Parallel slot (so I can use it on laptops with parallel ports) and it will give me a readout of the last process the motherboard went through before it stopped working. On the one I suspected was bad, it came straight out and told me no-cpu, knowing that the CPU was good (worked in another motherboard and I tried a known good CPU in the bad motherboard with the same results) it quickly reduced it down to me knowing the CPU socket on the motherboard was faulty (or something of the sort, at least).

In general, if it's not getting to POST and not beeping or anything, its either power or CPU related. Power is easy enough to test with a voltmeter or proper equipment, CPU issues are a bit trickier but can still be tested for. In your case it makes it easy that you have 2 processors and you know one works, so if the other one works in the first socket, try it with keeping that one in the first socket and putting the other one you removed in the second socket. If it still doesn't work then its definitely motherboard issues since you know both processors work on their own in the first socket.

[goatsniffer]

Update:

I tried CPU2 in CPU SLOT1 and system posted. Determined problem resides with board. Initiated RMA, awating new motherboard.

[goatsniffer]

So I received the new motherboard, and I rebuilt the system. Same exact issue when both CPUs are installed. That rules out the motherboard.

This leaves the PSU, RAM, or CPUs as the possible faulting components then.

I talked with an engineer from supermicro, and he assures me the RAM is compatible and the PSU is sufficient even though I had to convert a PCI-E connector into an EPS12v connector. I have also tried every combination of the RAM sticks so unless they are all faulty then the RAM is good.

The supermicro engineer thinks that something could be shorting out the board or a QPI link on one the processors might be faulty. I'll take the motherboard out and try and assemble it on top of an anti-static pad to make sure nothing from the case is shorting it.

Any thoughts or insight into how I should proceed?

Thank you.

[+Vice]

As I began reading this I thought 'I bet something is shorting' as I've had this issue myself where the case was touching the board and causing a short. The result was no beeps but fans spinning, no video on display no post. It seems the SuperMicro engineer you are speaking with is having the same thought.

If it continues to not post even outside of the case I recommend you try increasing the voltage on the CPU and Memory with only one CPU in the system, then shutdown and stick the 2nd CPU in and see if it posts. The memory controllers in these latest Intel processors are incredibly finicky.

[goatsniffer]

Thanks Vice. I will give this a try later. It's a big help when someone can confirm a feeling or suspicion, as it instills some confidence in taking the actions required to troubleshoot these haunted pc problems. So thanks again.

I'll let neowin know what happenned after I try it.

[goatsniffer]

Powzah! Right on.

I took the board out and set it up on an anti-static surface. The board worked like a dream.

Having now experienced this type of problem I can better troubleshoot it later, but I tell you broskis, this has been a total jabroni of a build. I hate losing sleep, hair, and arterial wall integrity over this kind of stuff.

I hope this thread can help someone else experiencing the same thing.

Thank you all for helping me digest and work through this problem.

[Stephen Degler]

I had the same Issue with a Supermicro X8DTL, no post no beep. The IPMI came up though. This ATX form factor motherboard is missing one of the standard mounting holes, and my case was shorting out because It still had the metal stud in that position.

Thanks for the excellent pointer. I've built many many systems but this was a new problem for me.