New Supermicro x11sph-nCTPF won't power on

[fmatthew5876]

Hello everyone,

I just got new parts to upgrade my system. I ordered a SuperMicro X11SPH-nCTPF motherboard, an Intel Xeon Silver 4114 CPU, and some 32GB DD4 2400 RDIMMS. Everything arrived today but I'm having a problem.

When I connect power to the board, the BMC starts up and I see the heartbeater lights flashing. I can connect to the IPMI and change settings.

The problem is when I actually try to turn on the machine with either the physical power button or via IPMI. Nothing happens. No fans start, no error beeps on the speakers, nothing. The IPMI gui still says the system is "Off".

I've tried the following:
* Verified CPU is installed in correct orientation
* Taking the board out of chassis and putting it on a wood table, to ensure no shorts.
* Removing all of the RAM, and also trying a few dims at a time.
* Using a different PSU (the current one has been running in production for a long time, so that likely is not it.)
* Clearing CMOS

I'm running out of ideas. At this point, the issue has to be either with the motherboard or the CPU. The manual suggests running it without any RAM installed and if you still get no error beeps replacing the motherboard. CPU's rarely fail so I would bet the motherboard would be the issue.

I don't have any other Xeon Scalable gear to be able to independently test the mobo and CPU. I'm worried to send the mobo in for RMA and then have them test it and find out its ok and then end up charging me for it.

Does anyone have any suggestions for other things I might try before sending something back?

Thanks!

 

[bitrot]

What kind of chassis and PSU are you using? If you’re using a Supermicro chassis and PSU, have you made sure to connect the front control panel header and power system management bus header correctly (-> manual)?

 

[fmatthew5876]

I'm using SM Chassis and PSU and I've tried it with and without the PSU smbus connector. I've also tried using the front panel header and also leaving it off and just connecting the power on pins with a screw driver in an effort to power on the machine. The other PSU I tested it with is not Supermicro.

 

[bitrot]

Do you have that Supermicro front panel split cable (CBL-0068L) in your posession by chance for testing / analysing purposes? How did you connect the non-Supermicro PSU to the board?

 

[K D]

In one of the STH articles on the LGA3647 socket, it was mentioned that if the socket is not tightened exactly the right amount then you were not able to post. Don't recall which one.

@Patrick

Also, ensure that the RAM is compatible and populated in the right memory slots. I've had Issues presenting the exact symptom occur twice with Once it was due to RAM in the wrong slot. The second I think is a faulty PDB. Haven't figured it out yet.

 

[fmatthew5876]

Do you have that Supermicro front panel split cable (CBL-0068L) in your posession by chance for testing / analysing purposes? How did you connect the non-Supermicro PSU to the board?

I do have that cable, and I've tried using it. The original chassis cable seems to work though so this is not necessary. I know it works because when I connect a DAC cable to one of the SPF+ ports, one of the NIC leds on the front of the chassis starts blinking.

 

[bitrot]

Hmm. So it's not a power issue. Agree with K D that it makes sense to double / triple check the CPU and RAM installation. If that doesn't help, the board really could be DOA.

 

[fmatthew5876]

Also, ensure that the RAM is compatible and populated in the right memory slots. I've had Issues presenting the exact symptom occur twice with Once it was due to RAM in the wrong slot. The second I think is a faulty PDB. Haven't figured it out yet.

I have 6 32GB DDR4 2400Mhz RDIMMS. These are SuperMicro certified. They even have the nearly the same part number as a model in the "Tested Memory List" link on SM's website for this board. The only difference is the test model ends with ER26 (2666Mhz) and the ones I have end with ER24 (2400Mhz). I did not buy 2666Mhz because Xeon Silvers only support 2400. From everything I've been able to research these should be compatible.

I tried putting all 6 in the blue slots. I tried putting only 1 in each far blue slot on both sides. I tried with no memory at all. Nothing happens.

 

[fmatthew5876]

Also, the manual says this:

If there is no error beep, try to turn on the system without DIMM modules installed. If there is still no error beep, replace the motherboard

It's very easy to buy the wrong memory or have some incompatibility there. However, since the issue is still happening without any dims installed I'm inclined to believe the issue is unrelated to memory.

 

[bitrot]

Have you tried contacting the Supermicro support? They're usually rather helpful. Maybe they have an idea.

 

[K D]

Do you have that Supermicro front panel split cable (CBL-0068L) in your posession by chance for testing / analysing purposes? How did you connect the non-Supermicro PSU to the board?

That cable is only for connecting the chassis connector not for PSU.

 

[bitrot]

Yes, true, I meant as an alternative to the standard front panel connector for testing purposes, as you can connect each pin seperately. "PSU" was a brainfart.

 

[K D]

Yes, true, I meant as an alternative to the standard front panel connector for testing purposes, as you can connect each pin seperately. "PSU" was a brainfart.

Got it.

My suggestion would be to contact Supermicro support or depending on where OP purchased it, just do an RMA and replace it.

 

[fmatthew5876]

So it was actually a rookie problem. The cpu wasn't screwed in completely secure. So this is something to really double check.

Now after booting I get a memory error whenever one of the dimm slots is being used. So looks like the board still needs to be RMAed.