Dual Epyc 128 Core Rome Build - No Post

mlangenheim

New Member
Mar 18, 2021
16
1
3
Hello,

I am building a server and wondering if anyone can find any reasons why it is not posting.

The fans all turn on and I can see green lights on the mobo indicating that the BMC is active.

The lights on the Lan ports light up when I connect a laptop to the back Ethernet ports.

Here are the Specifications of the build:

Motherboard:
Gigabyte MZ72-HBO (Rev 1.1)

Link to motherboard manual:

Link to Gigabyte website for motherboard information:

CPUs:
2x AMD Epyc Rome 64 core
(128 core total)
CPU Part # ZS1406E2VJUG5

Link to CPU info:

RAM:
16x 16GB DDR4 3200 (256GB total)
Samsung 1x 16GB DDR4-3200 RDIMM PC4-25600R Dual Rank x8 Module
Part# M393A2K43DB3-CWEBY

Link to Samsung website for info about ram:

SSD:
Western Digital WD BLACK SN850 NVMe M.2 2280 1TB PCI-Express 4.0 x4 3D NAND

Cooling:
2x Noctua NH-U9 TR4-SP3, for AMD sTRX4/TR4/SP3

Chassis:
Supermicro SC747BTQ-R2K02B 4U Barebone Supermicro SC747BTQ-R2K02B 4U BareboneW/ Dual Power Supplies (PWS-204A-1R) 8x Tray

Link to chassis manual:

Power Supplies:
2x model - PWS-2K04A-1R
80Plus Titanium
(1x of the power supplies is a redundant supply.)

Link to the data sheet for the power supplies:


So far I have tried...

1. Removing all of the Ram and putting only one reseated Dimm per cpu in the DIMM#1 position

2. Plugging directly into the wall.

3. Reseating both CPUs and reapplying thermal paste

4. I tried an old Radeon gpu to get video rather than the onboard vga because when reading the Gigabyte MZ72-HBO manual the mobo’s VGA is referred to as “loop thru” and might be a VGA Input?


I’m not sure if the power supply has enough power. I purchased some power cables because the power supplies that were included with the Supermicro 4u chassis that I purchased did not include any power cables.

Here are some details about the two power cables that I went ahead and bought for the power supplies to replace the missing power cables.

Power cable details:
16 AWG gauge conductors
13 amp / 1625 Watt rated NEMA to IEC cables with standard NEMA 15-5-P grounded power plug s and 3 pin shrouded female IEC-320-C13 connectors.

They are SJT 13A/300V rated.

I think the power supplies might be power switching? If they are power switching then with the setup I have I might only need an input of 1000W: 100-127Vac/12.5-9.5A?

I put more details about the PSUsbelow.

Power Supplies:
2x PWS-2K04A-1R 80Plus Titanium
(1x is redundant.)

Power Supply Input Characteristics:

RATED VOLTAGE / CURRENT
1000W: 100-127Vac/12.5-9.5A
1800W: 200-220Vac/10-9.5A
1980W: 220-230Vac/10-9.8A
2000W: 230-240Vac/10-9.8A
2000W: 200-240Vac/11.8-9.8A (for UL/cUL use only)

RATED FREQUENCY
50-60HZ

INRUSH CURRENT
Less than 30A

POWER FACTOR
0.98 @ max load (Typical)

INLET CONNECTOR:
IEC-60320-C14


One more thing that I wasn’t sure about was if the front panel connectors could cause the server to not post if they were not placed into the correct pins on the motherboard.

I bought a Supermicro breakout cable in order to configure the cables into the gigabyte motherboard since the Supermicro chassis had its own front panel connector that needed to breakout in order to configure the front panel pins onto the motherboard. Has it ever been known to cause a server to not post by have the front panel header inappropriately configured?

I did read the motherboard manual for the front panel configuration but wasn’t certain that I had it right.

I think that I have placed ram in the correct dimm slots? They are lettered in the manual and according to the manual I think i have one in dimm slot #1 for each CPU.

Can anyone confirm that the Motherboard has VGA input only and no VGA out?

One more thing about the build is that there is a Supermicro backplane for SAS in the chassis but I am not using it and have unplugged it. I am just using the M.2 SSD for now. Though it doesn’t have any OS on it yet since it isn’t loaded yet as I haven’t been able to get to the bios screen or post.

Thank you very much to anyone who takes any time in considering what the issues may be in this build.

I will be truly grateful to anyone who has some good ideas!

Thank you.
 
Last edited:

NablaSquaredG

Active Member
Aug 17, 2020
276
101
43
You included a lot of information - But a couple of essential things are missing ;-)

The motherboard has onboard GPU (in the BMC), the VGA is OUTPUT, not INPUT.

My troubleshooting suggestion:
- Remove the front panel connector
- Remove all addon cards (GPUs, SSDs, etc)
- Log into the BMC / Web interface
- Open the remote KVM
- Power on the machine via KVM

This way you can be sure that it's not a VGA connection issue.

The BMC also contains error logs which may help shed light on what's wrong.


Important question:
Where have you bought the CPUs? Are they new or second hand?
AMD EPYC CPUs CAN be vendor-locked if they've previously been used in Dell or Lenovo machines.
If an EPYC CPU has been used in a Dell machine, it is permanently fused and will only ever work inside Dell machines (same for Lenovo)


//EDIT:
Ah, I see you're using ES CPUs... Well, that might be the reason, but I'd still recommend that you do the troubleshooting suggestion I gave above.

//EDIT2:
Your power supplies have enough power. It's most likely not a power supply issue. The power supplies are auto sensing, SMPS (Switching mode power supply). They are "auto regulating" and work with different mains voltages and frequencies (i.e. a PSU bought in the USA will also work in Europe, it's the same model for all sales locations).
The power numbers you gave are for different mains voltages, the PSUs only give 1000W each when you use them with 115V (current is limiting, hence you get twice the output power at 230V).
 
Last edited:

mlangenheim

New Member
Mar 18, 2021
16
1
3
You included a lot of information - But a couple of essential things are missing ;-)

The motherboard has onboard GPU (in the BMC), the VGA is OUTPUT, not INPUT.

My troubleshooting suggestion:
- Remove the front panel connector
- Remove all addon cards (GPUs, SSDs, etc)
- Log into the BMC / Web interface
- Open the remote KVM
- Power on the machine via KVM

This way you can be sure that it's not a VGA connection issue.

The BMC also contains error logs which may help shed light on what's wrong.


Important question:
Where have you bought the CPUs? Are they new or second hand?
AMD EPYC CPUs CAN be vendor-locked if they've previously been used in Dell or Lenovo machines.
If an EPYC CPU has been used in a Dell machine, it is permanently fused and will only ever work inside Dell machines (same for Lenovo)


//EDIT:
Ah, I see you're using ES CPUs... Well, that might be the reason, but I'd still recommend that you do the troubleshooting suggestion I gave above.

//EDIT2:
Your power supplies have enough power. It's most likely not a power supply issue. The power supplies are auto sensing, SMPS (Switching mode power supply). They are "auto regulating" and work with different mains voltages and frequencies (i.e. a PSU bought in the USA will also work in Europe, it's the same model for all sales locations).
The power numbers you gave are for different mains voltages, the PSUs only give 1000W each when you use them with 115V (current is limiting, hence you get twice the output power at 230V).
I was able to log into the IPMI and the logs are normal and the CPUs are fine apparently. I am still wondering that the VGA on the board might be an input though because the motherboard’s manual lists it as an input. In the diagram in the manual it says “VGA Port” but further on down in the manual it is listed as an input.

I think I am going to try installing a vga card into one of the pci-e slots. Maybe that will work.The motherboard apparently has the following Management Console: AMI Megarac SP-X BMC firmware. I’m not sure how that works in terms of the bios with this motherboard. When I click on bios in the Megarac BMC it just takes me to a blank page. Do I need to load the bios somehow? I’m not sure where to go from there.
 

NablaSquaredG

Active Member
Aug 17, 2020
276
101
43
I am still wondering that the VGA on the board might be an input though because the motherboard’s manual lists it as an input.
No, it is definitely not. Perhaps you're confusing it with the DB9 connector?

When I click on bios in the Megarac BMC it just takes me to a blank page. Do I need to load the bios somehow? I’m not sure where to go from there.
It should show you an interface where you can configure BIOS settings. However you need to authenticate again, usually your browser shows a popup where you're prompted for username + password, which are identical to the IPMI credentials.

I’m not sure where to go from there.
Have you performed the troubleshooting steps I gave you? If yes, what was the result?

P.S. Please do not fullquote.

P.P.S. Please do not post about your issue in two threads at the same time.
 
Last edited:
  • Like
Reactions: mlangenheim

mlangenheim

New Member
Mar 18, 2021
16
1
3
I just saw your messages so I haven’t tried the other troubleshooting steps yet. I’ll try them and reply as soon as I have attempted each step. Thank you!
 
Last edited:

mlangenheim

New Member
Mar 18, 2021
16
1
3
I was able to get to the bmc bios by using a compatible browser. Now that I can get into the bios area of the bmc I should be able to change the boot order to usb and install Ubuntu Server.

Athough since the m.2 drive hasn’t shown up in the inventory I suppose there won’t be anything to install it on until it is recognized.

So I suppose I still need to go through your troubleshooting steps before I can proceed to install an OS then right?

I was able to view the kvm through h5viewer though it’s currently displaying blank video.

I’m also wondering about the next steps after I’m able to get through some of this portion of the setup... To be specific I’m wondering what your opinion is as far as the best way to install Ubuntu server 20.4 LTS?

Thank you
 
Last edited:

mlangenheim

New Member
Mar 18, 2021
16
1
3
@NablaSquaredG

I have now tried your recommend troubleshooting...

- Removed the front panel connector
- Removed all addon cards (GPUs, SSDs, etc)
- Logged into the BMC / Web interface
- Opened the remote KVM
- Poweron the machine via KVM

There is still no post. I have also tried many more troubleshooting steps as well. The CPUs are showing up in the system inventory in the bmc but the server still won’t post.

Another strange thing is that I have 256GB of Samsung RAM

(16x16GB sticks that are DDR4-3200 RDIMM PC4-25600R Dual Rank x8 Modules Part# M393A2K43DB3-CWEBY)

Even though the ram is all the same the bmc system inventory is detecting them as an incorrect type and another brand, (micron memory), even though I only have Samsung memory. Furthermore, the bmc RAM inventory is also displaying that each dimm is populated with 32GB RAM sticks even though they are all 16Gb sticks and displaying a total of 512GB of RAM even though there is really only actually 256GB of 16x16GB Samsung memory in the server that I have installed.

Also the KVM has no video signal and there isn’t any VGA being output to an external monitor.
So, I really have no way of seeing what may be going on on a screen through the KVM or an external monitor.

Another weird thing is that when I navigate through the bmc to the bios and click on advanced > Output Select; the options in the drop down for output interface say, “unknown device”.
Then in the bmc accessed bios when I click on advanced > Legacy Video Select; nothing pulls up under legacy video select.

I have already updated the bios and firmware to the latest versions available on the Gigabyte website.

I also cleared the CMOS and installed a brand new button battery, reseated everything including; RAM CPUs, power cables, etc. I even tried removing all of the RAM at one point and tried using only 1 stick per CPU in the DIMM#1 position to rule out any issues with the RAM.

I’m hoping someone has an idea of where to go from here. Am I perhaps missing a bios setting that needs to be changed in order for my particular setup to work on this board? I mean there are a ton of settings to go through in the bios many of them are beyond my level of comprehension.

Can anyone recommend any bios settings that I should adjust for this Dual Zen II Rome setup? If anyone can let me know what they think might be some mandatory or likely needed bios adjustments for my particular setup or has any ideas in general I would love to know!
 
Last edited:

NablaSquaredG

Active Member
Aug 17, 2020
276
101
43
The inventory is still showing the information from the last successful full boot, i.e. the CPUs + RAM Gigabyte used to test the board.

That there's CPUs + RAM listed in the inventory says nothing about the issue your dealing with.


Could you please record the screen of how you're trying to start the server?
Please also include pictures of the board + setup.
Is the board inside a case? If yes, have you tried posting outside the case (i.e. board placed on a non conductive surface)? Could be a misplaced standoff, causing a short circuit.
 
  • Like
Reactions: mlangenheim

mlangenheim

New Member
Mar 18, 2021
16
1
3
I’ve attached an image of the setup. I will see if I can get a video here in a bit.

Thank you for the suggestion to try starting outside of the case. I remember that when I put the board into the case that there was one standoff that wasn’t perfectly centered on the hole for the screw. Maybe that is the issue?

I will try taking it out of the chassis and starting it and see what happens.

Thank you.
 

Attachments

mlangenheim

New Member
Mar 18, 2021
16
1
3
@ExecutableFix

I just saw your Rome Engineering Sampleover clocking post. I have two ZS1406E2VJUG5 ES CPUs that I haven’t been able to get to post. I saw that you mentioned copying half of the bios in order to get these Engineering Sample CPUs to post on certain motherboards.

That sounds interesting however I am concerned that I wouldn’t know what I was doing and might botch the bios.

Can you post a modified bios for those of us like me who have an Gigabyte MZ72-HBO rev 1.1 board or can you explain in greater detail how to do that?

Thank you
 
Last edited:

mlangenheim

New Member
Mar 18, 2021
16
1
3
@NablaSquaredG I tried powering up the server from the kvm with the board outside of the chassis and it still did not post.

I just noticed a post about overclocking ES Rome CPUs that was posted by another serve the home member. Supposedly sometimes the ES Rome CPUs don’t post depending on your motherboard.

I just asked him if he may know how to make the MZ72-HBO motherboard bios capable of posting the ZS1406E2VJUG5 engineering sample CPUs.

if this is true then I’m sure there will be many people that would love to be able to use these ES CPUs with their boards since they are so much more affordable.

Here is a link to the post by @ExecutableFix
titled “Finally: Overclocking EPYC Rome ES”.

https://forums.servethehome.com/index.php?threads/finally-overclocking-epyc-rome-es.28111/
 
Last edited:

Leo Atreides

New Member
Dec 23, 2020
2
1
3
Here is a link to the post by @ExecutableFix
titled “Finally: Overclocking EPYC Rome ES”.
I think you should read through that post in more detail. What ExecutableFix has said is that with specific bios revisions it is possible. AFAIK posting ES CPUs are only supported on 2 boards that require a specific BIOS chip with a specific BIOS revision. The Gigabye board you are using is both more modern than that forum post and to work with that method will require the same BIOS ( or v. similar ).

You can answer the question as to whether you can use that bios revision by reading through the extensive posting.
 
  • Like
Reactions: mlangenheim

mlangenheim

New Member
Mar 18, 2021
16
1
3
@mirrormax

Do you know if Engineering Sample CPUs work on the MZ72-HB0 motherboard?

Motherboard:
Gigabyte MZ72-HBO (Rev 1.1)

Link to Gigabyte website for motherboard information:
MZ72-HB0 (rev. 1.x) | Server Motherboard - GIGABYTE U.S.A.

I have the following Engineering Sample CPUs:

2x AMD Epyc Rome 64 core
(128 core total)
CPU Part # ZS1406E2VJUG5

Link to information about the AMD Engineering Sample CPUs that I have:
https://www.cpu-world.com/CPUs/Zen/AMD-EPYC 1.4 GHz - ZS1406E2VJUG5.html

I can’t seem to get the server I’m building to post and I’m trying to isolate which hardware components of the build are the cause.

The bmc also doesn’t show my ram and the board doesn’t output vga or any video through the kvm. If these CPUs don’t work on this board would that cause the bmc to also not show my ram in the inventory and to not output any vga?

Thank you
 
Last edited:

mirrormax

Member
Apr 10, 2020
88
36
18
Hi i have no way to confirm or not if that cpu works, as i only have the OEM 53-04 cpus, but its very possible they dont work since i do remember reading about ES cpus not working on latest AGESA bioses with supermicro boards. and the gigabyte board only has newer agesa. consider switching to 53-04 cpu or use a h11 supermicro board thats confirmed to work.

edit: see this thread for more overclocking rome you need a board with agesa 1.0.0.3 or lower. gigabyte boards come with 1007 or newer.
 
Last edited:
  • Like
Reactions: mlangenheim