New motherboard not working

octafish

octafish

octafish

Associate
Joined
17 Jul 2024
Posts
44
Location
Oxfordshire
So I'vbe been collecting the parts to build a new server/PC, using the case (a rather massive phanteks thing) of my old one I built some 8 years ago. That used a Asus Z10PE-D16 WS, the new motherboard is a Z13PE-D16. I have a Micron 2Tb NVM SSD with the old PC OS on it, new Crucial DDR5-5600 memory and two shiny new 5512U Xeons. Kept the existing 1000W EVGA power supply.

Spent some time today stuffing the new motherboard with the memory and cpus and checking the connections. Then went and powered on.

The fans came on (both CPU fans and case fans) but within 2 seconds it shut itself down...

Looking a bit more closely at the LEDs on the board:
- the standby power green LED blinks green as expected
- the Q-Code LED indicated 00 then 10 before the shutdown, apparently 10 means 'PEI Core is started' whatever that means
- the green DIMM LED briefly goes green when power is applied indicating there is 12V power to the memories
- a red LED next to the styandby power one comes on very briefly at the moment of shutdown. I think this is the 'Processor Catastrophic Error' LED.

So I'm kind of stuck. I did try taking the board out and checking there were no shorts on the back to standoffs, and the testing is without any SATA/USB connections or PCIe cards. I've checked the memories are in the right slots, the only thing I can think of now is something wrong with the CPUs? They are LGA4677 ones which is new to me but they seemed to drop in OK and fitting Noctua coolers was straightforward.

Any suggestions appreciated!
 
According to the order, memory is Micron it seems:

32GB DDR5-5600 RDIMM MTC20F2085S1RC56BR 2Rx8

I should add I get no system beeps as this motherboard doesn't have any kind of speaker/buzzer!
 
ANDARIAL said:
Was the old system the same mobo as this new one?
might be the problem...maybe
Click to expand...

Old system had a different mobo, it was an Asus Z10PE-D16 WS. What I had planned was to replace the complete mob including DRAM, NVMe, memory with the new board which has new CPUs, DRAM and NVMe.

I'm hazarding a guess the CPUs are working but the fail at the PEI stage is due to memory. The memory is in the right slots and seated correctly. I believe the CPUs are seated OK, as if I take them out nothing happens at all if I power up the mobo.

Asus say 4800 RDIMMs for the Z13PE-D16 but I have 5600 RDIMMs, could that be an issue?
 
ANDARIAL said:
Will it boot to bios WITHOUT the SSD?
I think dimms will default to `jedec` speeds.not sure
Tried clearing the cmos?
Click to expand...

I'll give it a try but I don't think it gets anywhere near trying to boot the bios. As the 'CATERR1' LED briefly lights before shutdown - it seems there is something fundamental with the hardware.
 
Removing the SSD had no effect, it never gets that far. I’m more and more thinking it doesn’t like the memory, I took the CPU’s out and reinstalled them, double checking they were the right way etc. I guess the other possibility is the power supply but it’s a 1000w one and was fine with the old mobo.
 
Ouch, I'm gutted. Scan say both CPU sockets have damaged pins and so cannot be exchanged or refunded. Well, I'm not aware of damaging them during installation - all you have to do is remove the protective cover and drop the CPUs into the sockets. And I'm a bit OCD about handling a board that cost over a grand. I believe the damage was there by the time it was delivered, although I don't have any proof of that (and Scan does not inspect boards prior to shipment).
Not sure what to do... they will see if they can get an exchange from Asus (do I see a pink thing flying?)
 
Yeah, I guess its possible. It must be very easy to bend pins, the last board I put together had LGA2011's and had no issue with.

Scan are being very helpful and have said they'll try for a return with Asus. If not they believe they can repair it. Fingers crossed.
 
Well, it wasn't the motherboard. I have a replacement and the problem is exactly the same. So I guess it narrows it down to:
- cpus. Seems unlikely both would have failed.
- memory. Crucial tell me the memory is compatible with the motherboard, though.
- power supply. It's an EVGA 1000W one, that works fine with my old motherboard.

Bit stumped as to what else could be causing it to fail!
 
I did try that with the previous mobo, but can try again. I am wondering if the cpu is not connecting to the pins, the LGA4677 has a plastic carrier that you fit the cpu into, attach the cooler then fit onto the socket. Although it seems to fit into the carrier OK (there is only one way it can go, and the triangles line up between cpu and coard) it has got me wondering.
 
How the devil do you post images? There is an 'Insert Image' icon but it doesn't want to upload a web link I give it, notr is there any means of dropping an image in this post as far as I can see.

So I'll resort to the old fashioned way, sigh:

https://www.peardrop.co.uk/IMG_5936.jpg - this is a pic of the socket. All pins look good.
https://www.peardrop.co.uk/IMG_5937.jpg - a pic of the CPU in its carrier. If you look really carefully you can just about see the marks the pins have made on the cpu pins, right in the middle of them. So it does appear to be making contact.
 
The fitting of the cpu in the carriers is something that’s not 100% clear, I followed one of the YouTube guides for a LGA4677 cpu. Apparently the outer row or two of the cpu pins are unconnected. Might take the coolers off and try fitting the cpus again.
The code 00 is shown initially then it quickly changes to 10 which apparently means something like PEI initialisation whatever that is. At that stage it shuts down.
The memory speed is confusing, some say 4800 but crucial and Kingston say 5600 is ok. As a last resort I’ll buy a couple of 4800 RDIMMs from Amazon and then I can send them back if it makes no difference.
 
Last edited:
OK so with no USB devices connected, still the same.

I did a bit more testing with just the mobo and psu (a thermaltake 1200)

First with just CPU1 and a pair of DRAM for it - as before, on power up LEDs go from 00 to 10, then red CATERR1 led comes on and the psu shuts down.
Then removed all DRAM - just the cpu. Exactly the same.
Lastly removed the CPU as well... of course pressing ppower on does nothing.

See the pic at https://www.peardrop.co.uk/IMG_5938.jpg for the setup. No, that antistatic bag is not conductive, I tested it.

What to conclude?
- Possibly the DRAM is incorrect as taking it out made no difference
- Other possibility is the CPU is incompatible with the mobo?

I'm leaning towards the first i.e. DDR5-5600 is incompatible with the mobo/cpus but appreciate others thoughts before going out and buying a couple of DDR-4800 to test that theory.

Question: I'm assuming those power connectors at the top that are unconnected and marked 'GPU' are to connect to a GPU? Or do they need to be connected to the PSU??? It's not clear in the docs, my last mobo only had the 24 pin ATX power and two 8 pin CPU power sockets.
 
Looking at the Asus manual for the Z13PE-D16, it says 'GPU Power Connectors (8 pin GPU_12V1/2). These connectors are for supplying power to installed GPU cards'. From that I would ahve thought they supply power to gpus, not receive it from the psu?
 
Well a quick try indicated no difference, so that ones out too.

About the memory speed, Asus say:

Z13PE-D16

ASUS Z13PE-D16 server motherboard for dual 4th Gen Intel Xeon Scalable CPUs, with single-CPU support for three PCIe cards, 16 DIMM DDR5, dual M.2 and integrated ASUS ASMB11-iKVM.
servers.asus.com servers.asus.com
Now does that mean DDR5-4800 only is supported or is DDR5-5600? Its as clear as mud.
 
Yes the Intel spec for the 5512U says 4800 (https://www.intel.com/content/www/u...ssor-52-5m-cache-2-10-ghz/specifications.html). Seems like only a few 5th gen xeons officially go up to 5600. I'm sure I read somewhere that the cpu controlled the speed and if it did not support 5600 would throttle back to 4800. But maybe not.

So I've ordered a couple of Micron DDR5-4800 RDIMMs to try. If they work, great, if not... then I've got a lot of expensive trash here.
 
wookiee87 said:
Just out of curiosity did you consider threadripper/epyc, a single socket solution from team red would be a very compelling option vs intel.
Click to expand...
I guess I'd been a happy dual Xeon user for the last 20 years, which worked great for running VMWare - typically 8 linux VMs at once - but yeah, this does make me wonder.
 
Yes my current setup which I put together in 2016 is an Asus Z10PE-D16 WS with twin E5-2640's and 256Gb memory, its been working flawlessly ever since although starting to show its age a little and won't run W11. It worked first time so naturally I had high hopes for moving to a similar but newer mobo and cpus. Alas its been a nightmare.
 
You must log in or register to reply here.
Back
Top Bottom