Help! Gigabyte MZ31-AR0 no video

[tiggun]

Hi,

I'm new to servers and especially home servers, but I have built PCs before. I have a gigabyte MZ31-AR0 EPYC SP3 motherboard and I can't get any video signal out of it. I have heard that you can power up the motherboard with no CPU or RAM in it. But when I try that, the BMC led stays solid for a minute or more and then blinks. I assume that means it is working. But I get no video output on the vga port.

I have it plugged into an old Antec Truepower 650 watt power supply on my bench, and a keyboard in one of the usb ports. When I put my CPU in and one stick of ram, it also does not give any video. I tried putting a discrete graphics card and that didn't produce video either. The ram i have is samsung 8GB PC4-19200 DDR4 ECC M393A1G43DB1-CRC0Q

Is there a way to connect to the ethernet port to see what's going on? Or is my ram or mobo bad?

Thanks for any help!

 

[emunozG]

Hello @tiggun , that happens to me too, i can't see the data by VGA.
To access BMC via Ethernet, open a browser, access the ip that DHCP has automatically offered. The default user is
admin
password


I have come to think that the solution is in two jumpers that are next to AST2500, but nothing appears in the manual about these jumpers. I have not dared to try it yet

I have also tried to mount a vKvM and it establishes communication, the result is "signal not found"

 

[ProfKenobi]

I had this issue as well. One key item was to get the proper ram installed. The board will not function without RAM installed. The next is that it might be useful to connect to a case so that you can cycle the power with the case button. You can also briefly jump the Power Button or Reset Button pins to ground on the Front Panel Header (11-13, 15-17, respectively). Good luck!

 

[RageBone]

@emunozG i doubt that those two Jtag connectors are the key in this Situation.

Hello @tiggun,
two things that come to my mind.
The green blinking led is called BMC heartbeat and indicates that the BMC is enabled and has finished booting, once it blinks.

If you connect the dedicated BMC nic to your network, the BMC should request a IP via DHCP and then serve its ipmi on there. You should connect to it since you can see many more useful information on there.

The second thing is that my older gigabyte board isn't on even though the PSU is turned on and the fans spin.

I think this is a gigabyte specific feature to basically always have active cooling.
So even though you might think it is on, the CPU and system is still turned off and just in this unusual state between completely dead and powered on.

The ipmi should also tell you if it is fully on or not.

 

[tiggun]

Hi everyone, Thanks for the help.

I connected the motherboard to my wireless router, then logged into the router interface where I found the motherboard IP. Now I am able to get into the ipmi 'gigabyte embedded management software' interface where I can do things like power on and off the system. In the Hardware -> memory section of the interface it shows one DIMM installed, in H1 slot, a 4GB DIMM. But physically I have only placed one 8GB DIMM in the A1 slot, like the manual suggests when using one DIMM.

I tried putting that DIMM into the H1 slot but that didn't do anything. I also tried another DIMM which I bought, that I made sure is on the supported list for this board.

Not sure what to do next, maybe update the BIOS

edit: is this picture of the voltages normal?
Imgur

 

[RageBone]

The FRU info only updates once it made it into the BIOS. So that wasn't your setup.

What company made the socket?
Foxconn or Lotes?
You are using the blue Plastik carrier?

There could be the Option to See the current postcode in the ipmi.
Post code snooping on supermicro.
What is that displaying?

 

[emunozG]

-Hello everyone, @tiggun @ProfKenobi @RageBone is the same about the output of the voltages, @RageBone comment makes sense for the following reason:

The BIOS version is F06, it does not update to other versions, an attempt has been made to update through gbtutility, it accepts the file but fails to enter "monitor" mode (see image cliFailUpdateBios.png), it has also been tried by the browser web, accept the file and complete the update but when checking the version in a new web session we appreciate that it remains the same. The processor is " AMD EPYC 7281" (Note: in the web browser it tells me that we have another processor model, see image "processorModel.png"), the firmware version is "1.91" (see image versionBios-BMC.png), I have tried to update the firmware to the AMI version but I have not succeeded, I cannot use tools such as flash because it does not start the BIOS, I have verified that in the directory "https: // {IP} / redfish / v1 / Registries /" it does not contain "BiosAttributeRegistry" (see image redfish.png and redfishBiosInfo.png), it does not allow updating the firmware to previous versions, nor does it leave later versions . If I execute the “loaddf” command for “bmc”, an error occurs, it also happens if I execute the command for the BIOS. (see image bmcLoaddf.png)


-I contacted the gigabyte support department and explained the whole situation, a few hours later they published a new firmware update AST-2500, 1.95 for vertiv. I have managed to upgrade to version 1.95. After updating the firmware I tried to update the BIOS and the same error happens, it fails when the system enters “monitor” mode to continue with the update. I have managed to upgrade to version 1.95. After updating the firmware I tried to update the BIOS and the same error happens, it fails when the system enters “monitor” mode to continue with the update. I have tried with all BIOS versions published on the official website, with the same error in all cases. @tiggun I think the root of all our problems is the BIOS, so I have observed 2 details, all BIOS updates fail because of a small error about String equivalences, the stored "Board Product" String does not match exactly on our server with the “support board Product” String stored in the file that we try to upload to the server (biosUpdateObservation.png). The other observation is in the document "releasesBios", the kernel version appears as "AMI" and by default to @tiggun and I get with the firmware "vertiv", but how do we update to AMI if "Vertiv" to "AMI" you can not use "gbtUtility" or web browser, we do not have the utility of "flash" because it does not start the BIOS, if I do not misunderstand we do not have that utility without the BIOS, then as we are supposed to update it ? Is the problem really the kernel version of "ast2500"? If it's really a manufacturer's problem, how are we going to be rewarded these months without support? No solutions and all the lost time.

Note: I don't think it is RAM, in the specifications of the motherboard it matches the characteristics of my RAM
datasheetMyRam: https://www.kingston.com/dataSheets/HX426C13PB3_16.pdf
mz31-ar0: MZ31-AR0 (rev. 1.x) | Server Motherboard - GIGABYTE Spain

Note: Attached postcode + realeseBios image

Note: The manufacturer has not yet responded and I will attach the link to this thread so they have more information about the problem.

 

[RageBone]

What CPU are you using?
The BIOS file in the picture only mentions Naples support, that would be Epyc with a 1 at the end.

The BIOS update through the ipmi should work, since it doesn't, that is a Problem that you have to take to Gigabyte since they ****ed that up.

Unless you got the wrong BIOS for the wrong board, then its on you, but i doubt that a bit.

AMI is one of the few Main BIOS "vendors" so it is normal that it is branded AMI.

 

[RageBone]

Vertiv seems to be the BMC FW vendor,
I assume. Are you sure the update tool does update the BIOS and not the BMC FW?

 

[emunozG]

@RageBone
I also doubt that it is a manufacturing or similar problem, but I also do not rule it out since we are 2 different people who have the same incidence on the same dates, also if I remember correctly I found a third person with the same case with very approximate dates to the of @tiggun and mine, the web I do not remember and do not keep link, maybe I am wrong but I would say that we are 3, although confirmed we are already 2 cases.

We still do not discard anything, it is more likely to be our fault than a factory failure.




My cpu is: AMD EPYC 7281

According to the manual of the GbtUtility tool, I execute the command well, in addition the same also happens with EMS-GUI-WEB, in both cases I accept the file, in GbtUtility it gives an error when the BIOS enters “monitor” mode, through EMS- GUI-WEB also accepts the file, it even starts the update and the system indicates that it has been updated successfully, when it finishes and forces to start a new web session it is checked if the version has been updated, when checking it I see that the BIOS version remains the same, F06.




Note: I have changed the BIOS battery, the BIOS jumpers are in "normal operation" → BIOS REVOVERY

"normal operation" → supervisor password


Thank you for the help you are giving us.

 

[ingenthr]

I recently acquired a rev 3 of this board and had exactly the same problem described here and in a couple of other places online. I tried a number of things, including updating to the firmware released in November 2023.

I will say that at least for me, what seems to have 'fixed' it is that I moved the four DIMMs I have from the P0_[ABCD]0 to P0_[ABCD]1 slots. I did this because in the BMC, it seemed to show them in the A1 slot and the documentation says it needs to be in the slot suffixed with 0. After this move, the BMC showed more information on the DIMMs, though it still shows them in the '1' slots. But more importantly, I had video from both the BMC and locally connected. I think the system was not passing POST for some reason as I also am booting an OS that should work Serial over LAN, and didn't see anything there either.



It could have been that I had a mis-seated DIMM in the first place, but that's hard to say.

I wouldn't imagine so, but can anyone think of a reason, other than performance, that I shouldn't run with the DIMMs in these '1' slots?