BUILD New Supermicro build keeps losing power/dying

[lukyjay]

Hi

Recently I bought a new Supermicro X11SSM-F motherboard, Intel Xeon E3-1230 v5 and Kingston KVR21E15D8/16 16GB DDR4 ECC ram. Put this together with an old case and an old TX705W (750w Corsair) PSU.

My intention is to run FreeNAS on this server.

Firstly I didn't have a VGA cable for the onboard, so I tried to insert my Radeon 7870 just so I could configure the bios and install FreeNAS. Then I'd use it headless. There was no signal from the graphics card so today I went out and bought a VGA cable. Video now works fine.

However the computer started to randomly turn off, then wouldn't turn back on. When I hit the power button the fans would nudge a little and the lights would flash but it'd never stay on. I unplugged some harddrives (originally had 8, went down to 6) and it worked. Then it died, same thing, only 5 harddrives would work. Then it died, same thing, only 4 harddrives would work. Then it died, same thing so I used it with no harddrives and now even that is dying.

I managed to login to IPMI before it died and can see this error in the event log, but I don't think it is the cause because the motherboard passes POST just fine and gets past the GRUB bootloader on a USB drive in most cases.

Error-Unrecoverable video controller failure. - Assertion

Does anyone have any idea what is causing this? I don't have any other CPUs of this socket, DDR4 ECC ram or spare PSUs so it's difficult to troubleshoot.

Is it likely my PSU has died, or could these be the symptoms of something else?

Thanks

 

[lukyjay]

After looking at the IPMI a little closer, the unrecoverable video controller failure was from last night when I tried the 7870 (before using a VGA cable with the onboard), so that can't be the cause of the problem.

 

[GBillR]

My money is on that power supply... the initial current draw is probably causing a voltage drop that the board can't handle.

I would pull the board from the case and fire it up on the desk to see if that changes anything... just to rule out a mounting issue with the old case. If the symptoms do not change, you're going to have to find another power supply to troubleshoot further.

 

[SweetAndLow]

Jump the power supply and have a fan attached. See if it keeps working. Sounds like a power supply issue.

Sent from my Nexus 5X using Tapatalk

 

[pschatz100]

Power supplies that are failing can create all sorts of hard to find, and intermittent problems. If it's a very old unit, it might not have the right amount of 12V and 3V capacity, even though the rated power output seems good enough. Replace it. They're not that expensive.

I went through the same sorts of issues before wising up. When I purchased a new supply, I purchased a modular one so that I could deploy just the right power cables - which keeps the inside of the case nice and tidy. Never regretted the purchase.

And pick up a UPS while you're at it.

 

[Jailer]

Why are you messing around with a video card when the board has IPMI? Hook a cable to the IPMI interface and see what's going on. If the board has power, regardless of whether or not it will run, the IPMI interface will be accessible and you should be able to do some basic troubleshooting from there.

 

[JohnDigital]

If it posts, Id certainly rule out RAM and CPU, for now. Although it wouldnt be a bad idea to run with only one stick of RAM until you get somewhere on this. I 4th the PSU assertion. Once it boots and is stable you can reassess. I think if not the PSU then next culprit for me would be the board. Hope you get it figured out. Good luck.

 

[lukyjay]

I found it! yay! Disabling onboard video/vga with the motherboard jumper has fixed the problem.

Do you think this means my motherboard is faulty and needs an RMA?

I could live without video (really, who needs it with FreeNAS 10's sleek web ui), but the problem is IPMI. Without onboard video, IMPI won't let me see video when using KVM. Even with a PCI-e video card, IPMI still says no signal (though everything else works...)

 

[Jailer]

Do you think this means my motherboard is faulty and needs an RMA?

Yes

 

[lukyjay]

Yes

Well this is an inconvenient first world problem... I'll get in contact with SMC

Thank you every who replied to this thread and assisted

 

[lukyjay]

Just to update you Jailer, further troubleshooting shows it works fine as long as I have a video card in the PCI-e slot. If there's anything else in that slot, or nothing at all, it won't work. I can have the onboard video enabled and it all works just fine, only under the condition there's a video card in that slot (even when it's not utilised).

Very interesting problem indeed! Due to the downtime and problems of an RMA, I'll see how I go with this. An easier fix may be just to use a low profile, low power video card in that slot just to keep the motherboard happy.

I'm still speaking Supermicro but their tech support isn't very helpful. They are suggesting things that I've told them I've already done, like removing the board from the chassis and reseating the RAM.

 

[GBillR]

Just to update you Jailer, further troubleshooting shows it works fine as long as I have a video card in the PCI-e slot. If there's anything else in that slot, or nothing at all, it won't work. I can have the onboard video enabled and it all works just fine, only under the condition there's a video card in that slot (even when it's not utilised).

Very interesting problem indeed! Due to the downtime and problems of an RMA, I'll see how I go with this. An easier fix may be just to use a low profile, low power video card in that slot just to keep the motherboard happy.

I'm still speaking Supermicro but their tech support isn't very helpful. They are suggesting things that I've told them I've already done, like removing the board from the chassis and reseating the RAM.

Does this video card you are using have a PCIe power connection, and did you plug a cable from the power supply into it? I am just wondering if there is still a possibility that this is somehow power related. Your symptoms are textbook power supply related... either bad power supply or something not connected completely, or an intermittent ground of some type.

 

[lukyjay]

Hey, thanks for your reply

Does this video card you are using have a PCIe power connection, and did you plug a cable from the power supply into it?

Yes it does have a PCI-e power connector (1x 4 pin and 1x 3 pin).

It can't be connection because I've reseated and rebuilt everything three times, and its not in the chassis so it isn't shorting.

Are you implying the PCI-e power connectors are acting as a ground to prevent power failure? That's an interesting thought I hadn't considered to be honest. I did consider maybe the PCI-e power connectors were delivering more power to the board when it was running low, but that doesn't seem logical. Surely the 8 pin + 24 pin motherboard connectors deliver enough on a 750w PSU.

I am going to ask a friend to come over in a few hours and bring his 1500w gaming PSU. It's not an ideal PSU for this kind of set up, but it will let me rule out that component.

The only other thing I noticed is my 24 pin PSU is missing that optional -5v pin which is not required to be compliant since recent ATX specifications. Most PSUs don't have this pin, so I don't think that would be the cause.

 

[GBillR]

Honestly, power related issues can be hard to pin down, and sometimes simply a marginal connector can cause intermittent symptoms similar to yours.

I really think that ruling out a bad supply is a must do. Especially before dealing with an RMA, or worse, living with a setup that will always leave you wondering when it's going to drop the other shoe.

Good to hear that you've found a spare supply to test the board with.

 

[GBillR]

Just had a thought... you did connect the 8 pin pci connector to the board as well as the 24 pin right?

EDIT: Just noticed you mentioned the 8-pin connector in the earlier post.... I figured you must have already been asked that by the tech at Supermicro.

 

[lukyjay]

power related issues can be hard to pin down

Pun intended?? :D:D

Just had a thought... you did connect the 8 pin pci connector to the board as well as the 24 pin right?

Yep :) I have some qualifications through CISCO and certificates in IT, I'm confident I've built it correctly with no human error

 

[lukyjay]

There we have it folks, borrowed power supply = no problems. Must have been a grounding issue with my old power supply that was resolved with PCI-e power headers are connected.

Will look at buying a new one this week, probably a Seasonic since FreeNas forums love them.

 

[GBillR]

There we have it folks, borrowed power supply = no problems. Must have been a grounding issue with my old power supply that was resolved with PCI-e power headers are connected.

Will look at buying a new one this week, probably a Seasonic since FreeNas forums love them.

Glad to hear you were able to get it sorted out... now you can get back to build.

 

[lukyjay]

Glad to hear you were able to get it sorted out... now you can get back to build.

I'm really struggling with the software side of things. FreeNas 9 is easier to set up for me since it has more documentation and Google support, but jails and plugins seem very inefficient and extremely frustrating... I wish I could just use a virtual machine instead, but then again only FreeNas 10 supports that, and it's not stable yet. *pulls hair out*

 

[JohnDigital]

Jails rule.