Hello everyone, this is my first post been lurking for a bit
I’ve come across a rather perplexing issue with the AMD EPYC 7643 processor when used on a Gigabyte MZ01-CE1 motherboard. Here’s the rundown:
I am able to access the BMI.
There are no POST beeps or indications.
Caps lock and num lock lights don’t light up when pressed.
I have tried 4 different EPYC 7643 CPUs (all the same model).
I’ve also tried two different MZ01-CE1 motherboards.
All combinations produce the same outcome: no POST, no beeps, no keyboard indicators.
Things I have tried:
Checked and reseated all connections.
Verified the power supply is functioning correctly.
Updated the motherboard’s BIOS (using a different compatible CPU for the update).
Made sure the RAM is compatible and reseated them in different slots.
Reset the CMOS.
I’m honestly stumped at this point. Given that multiple CPUs and motherboards are behaving the same way, it seems like there might be something else at play here. Maybe a compatibility issue or a setting I might be missing? Has anyone else encountered this issue or can offer some suggestions on what to try next?
Any advice or insights would be greatly appreciated. Thank you in advance for your help!
Thanks for the reply, we’re on rev 3 for the ce1’s and rev 4 with the MZ72-HB0 which we RMA’d and awaiting its return.
I’ve tried the most recent bios flashes on all boards. Two CPUs came from a reputable supplier (scan) here in the UK as OEM, but I guess they could have been locked in testing or not as new as they say they were.
There’s no way to know is there apart from buying a dell board, which would in turn defiantly lock them. We’re finically quite deep into this project and only found out about the CPU locking recently from STH post from 2 years ago
Regarding the post behaviour, the fans come one and die down as the BMC starts (solid⇾blinking).
At this point we can access Mega-Rac, but everything on the back IO apart from the NIC is unresponsive keyboard num/caps lock etc.
The KVM inside Mega-RAC has a no signal and the details in the inventory are for different machines.
I have the two PCIe and the 24 pins plugged in with the three fans plugged in, we initially had a 1200w on the HB0 and a 750w on the CE1.
I had it hooked up to VGA initially, but then put in a low powered AMD GPU as well.
We have two sticks of Samsung 32GB 3200MHz ECC DDR4 Registered populating C+D
This behaviour has been consistent through three boards, 4 EPYC 7643’s and two 7502P. We have never been past this initial wind down stage regardless of config.
We have a potential lead/cable (sorry) just noticed the motherboard power is different to both the 2 cables we have for CPU’s on our NZXT and Corsair PSU’s.
Would this be enough to cause these issues, say, if there’s a handover in power and two smart boys have used the wrong connector? So the systems are stalling waiting for a module that is currently dark.
We have ordered one from supermicro which has the pin-out config that matches the third picture below the mother board and PSU
so if this is really a 1.0 board I think that only works with epyc naples?
since oyu can get to the ipmi can you flash the latest 3.x bios from?
when you ordered this board from the vendor which version did you order?
see the tabs on there… v1 for 7001, v2 for 7001/7002, v3 for 7003, etc.
I dont know if this is the issue, but maybe is the issue. The power connector thing is a bit of a worry but your picture looked like the right types of cables assuming they are matched to the right power supply.
the pinouts are NOT standardized at all. so the cables you ordered may not be compatible.
that didn’t match up but fits into our motherboards, I’m guessing this is a universal config?
But the motherboards act as if they’re not even plugged in as we get the same behaviour with them off the board
My thinking was that we get another PSU and try and match up the cpu connectors, but that’s really hard to do with the online images and searching for 8 pin eps 12v throws up various configs, and I’m a bit hung up on the Arch Square Square Arch thing our motherboards have on the latch side of the power socket. Do they have to match?
We orded a board,chip and memory pack from ebay. One dimm slot wasn’t working moving it about it turned out to be the stick. So I tried it with a stick of ram (samsung) from the build in this thread in the ebay supermicro, same thing empty DIMM
So knowing I had at least 7 sticks of functioning ram I took one out put it in our HB0 which had the same behaviour as the CE1 and we hadn’t gotten to post after a longer anixity inducing time spooling up than the supermicro we got post!
A huge thank you to Wendell we got to the bottom of this mystery.
Bent pins were killing ram sticks so we were taking those and putting them in other boards which in turn had no ram as it was a little silicone corpse sitting in the dimm.
Mystery solved, huge thanks to Wendell for the investigation.
Bent pins on the original socket were killing ram sticks. So when we put them in other boards they behaved the same as they just had little silicone corpses lying in their DIMM slots.
So the murderer was the socket in the build room with the pin short.
The board has been fixed, and we were able to get post and all DIMM slots.
Big thank-you to everyone who contributed to helping us out here.
