New Supermicro Supermicro H11DSI-NT Amd Epyc DOA?

mirol

New Member
Jan 6, 2018
16
1
3
34
Hi I just received brand new mobo straight from eu supermicro
Supermicro H11DSI-NT

I pair with 8x 2667mhz 32GB MTA36ASF4G72PZ-2G6D1Q1
And 2 x Amd Epyc 7451 and 2 psu I got evga 1000G2 or seasonic 600W Gold

It wont boot nothing on the screen no beeps or anything just fans spinning ( In ipmi on hardware info its show completly different hardware cpus and memory probably from supermicro testing,on sensors tab everything is n/a) I tried swap cpus in different socket ,try different memory,psu use only one cpu instead 2 jjust tried everything (reset cmos as well) still no post.Ipmi logs wont show any errors as well.The only thing It can be is cpus as I got them 2 from ebay.My question what you thing from description is mobo or both cpus are faulty?Its annoying as I waited 2 weeks for this mobo and it will probably take anmother 3 weeks to send it back and get replacement if this is DOA mobo.
 

RageBone

Active Member
Jul 11, 2017
291
77
28
The speaker can be turned off by a removing a jumper, so that might explain why you aren't hearing anything.
The hardware info is updated once it successfully reached the bios. So you aren't apparently.

It might be a bios issue, are those genuine or ES cpus?
you have postcode snooping in the ipmii, what is the postcode it hangs on?
How does it behave without ram at all ?
with only cpu?
etc, standard system check procedure.
 

mirol

New Member
Jan 6, 2018
16
1
3
34
Hi They are Genuine retail cpus postcode is FF its the same one when its turn off the weird thing as well when i connect power its turn on straight withouth pressing power button, as i dont think is always on as default.Like i said i tried with just 1 cpu in its exactly the same with or withouth ram as well.
 

RTM

Active Member
Jan 26, 2014
552
193
43
You could try updating the BIOS, it should be doable via IPMI.
It will probably require a license key, which isn't terribly expensive but obviously a waste of time if the board is dead, so you may want to try contacting Supermicro support, they may be able to give you a trial key.

Btw. I see no reason to exclude the possibility that the motherboard is the issue from what has been posted so far.

There are a few more basics, that I suggest you check (if you haven't already):
- Ensuring both 8 pin EPS plugs are connected
- Checking that the motherboard is not shorting between case and motherboard
 

William

Well-Known Member
May 7, 2015
785
250
63
62
With out knowing if you have used EPYC CPU's before, I can say from experience with TR processors that mounting the CPU's can be a bit tricky. If I had bad mounts the system might behave just as you are posting. Getting the torque right on the hold down screws is very important.

Try booting with only one CPU installed, I believe that would be socket #1 and one stick of RAM. Keep trying to boot if you have issues and loosen the hold down screws just a bit, retry boot.
 

mirol

New Member
Jan 6, 2018
16
1
3
34
With out knowing if you have used EPYC CPU's before, I can say from experience with TR processors that mounting the CPU's can be a bit tricky. If I had bad mounts the system might behave just as you are posting. Getting the torque right on the hold down screws is very important.

Try booting with only one CPU installed, I believe that would be socket #1 and one stick of RAM. Keep trying to boot if you have issues and loosen the hold down screws just a bit, retry boot.

Hi I just used torq setup at 14 lb/in and 16 lb/in on coolers as I read in supermicro manual.It will be an issue if i try it on a box from motherboard not actual case?And when I reseate them today I got post code 78 in ipmi.Its ok for me to try them withouth cooler just to check if they boot up?Another one with slot i need to use for only one memory?
 
  • Like
Reactions: William

alex_stief

Active Member
May 31, 2016
622
188
43
35
Are you using both PSUs on the motherboard. e.g. the second one for the second 8-pin connector? Maybe using only one PSU while troubleshooting would decrease the amount of potential error sources.
For the memory, the manual has a table on how to populate less than 16 DIMMs. Did you follow it with the 8 DIMMs you have? Though no info for only one DIMM per CPU. I would try to put one into each B1 slot. Though the slots should not really matter as long as the CPU pins get contact and the DIMMs are ok. Are these all known good DIMMs, i.e. recently tested in a different system?
 
Last edited:

mirol

New Member
Jan 6, 2018
16
1
3
34
Are you using both PSUs on the motherboard. e.g. the second one for the second 8-pin connector? Maybe using only one PSU while troubleshooting would decrease the amount of potential error sources.
For the memory, the manual has a table on how to populate less than 16 DIMMs. Did you follow it with the 8 DIMMs you have? Though no info for only one DIMM per CPU. I would try to put one into each B1 slot. Though the slots should not really matter as long as the CPU pins get contact and the DIMMs are ok. Are these all known good DIMMs, i.e. recently tested in a different system?

I only use 1 psu at the same time I just used another one for testing I populated memory as I found in manual(try with one as well). Really hard to say it was motherboard or cpus are faulty.I double check cpu socket on motherboard looks perfect for me coudnt see any bend ping etc.Usually I got FF post code but sometimes is 78 thats why i think its mobo or cpus(I got them from ebay as refubrished)I think I will send the mobo back first and they will let me know as I really hope its not cpus as i bought them really cheap...
 

RageBone

Active Member
Jul 11, 2017
291
77
28
A FF postcode is indeed Bad.
Its the first ever code, or a fatal error in my experience.

78 on the other Hand should be good.
Should!

Please try booting without RAM, it should move to 5X pretty quickly, probably 55.

You technically don't need to have both EPs 8pins connected, especially with just one CPU. But it is a very good idea if you get both CPUs to work.

Foxcon TR4 Sockets had a a slight alignment / tolerance issue in the beginning. I think you couldn't close the socket because the CPU wasn't in the right place and if you did, you'd damage CPU and potentially the socket.

You could take the CPU out of the carrier frame to see if it was damaged previously like that, and place it manually where it belongs.
Though, you need to be careful and confident, just like on all the other LGA Intel sockets before.
 

mirol

New Member
Jan 6, 2018
16
1
3
34
I never saw other postcodes that FF an 78 its exactly the same without any memory in cpu I double check both cous no physical damage on them.
 

RageBone

Active Member
Jul 11, 2017
291
77
28
Well, then it is very likely a BIOS issue.

FF is common with corrupt files on the spi chip, or a dead spi chip.
And anything but a "no ram detected" error code probably means that the CPU isn't even getting to look for Ram.


So have you tried a bios flash jet?
 

mirol

New Member
Jan 6, 2018
16
1
3
34
I didnt dried flash bios as i dont have license activatet in ipmi its doable via usb stick without post?
 

mirol

New Member
Jan 6, 2018
16
1
3
34
I reflash bios seems not change anything that bios post codes im getting ff 02 be b5 c8 cd b7 78 from what i understand its read cpu and mem and still nothing on the screen....if it go throught that all codes its doesnt mean that cpus are recognized and its almost sure that is faulty motherboard?
 
Last edited:

mirol

New Member
Jan 6, 2018
16
1
3
34
Just to make sure as my torque only got N-m could someone confirm for me is 1.58 Nm on torque is the right one?
 

kaltek2599

New Member
Feb 29, 2020
8
0
1
I just got my Rev 2.0 H11DSI board, successfully booted with one 7551 CPU and 1 stick of RAM (M393A2G40DB0-CPB) in the B1 slot.
, then moved up to 4 sticks for the same CPU - all good so far. Added in the second 7551 CPU and an additional stick in the B1 slot - and now no display on the VGA. I logged in through IPMI and find a boot code of 91. I loosened the screws, tried remounting the CPU - still the same - Any suggestions? I realize that the RAM isn't on the compatibility list but it is booting with 1 and 4 sticks for a single CPU.

Also weirdly my H11DSI board doesn't appear to have the Oculink headers in the location that the H11DSI-NT has as can be seen with the highlights, anyone else with an H11DSI have that or are the ports only present on the H11DSi-NT?
 

Attachments

ari2asem

Active Member
Dec 26, 2018
502
81
28
The Netherlands, Groningen
I just got my Rev 2.0 H11DSI board, successfully booted with one 7551 CPU and 1 stick of RAM (M393A2G40DB0-CPB) in the B1 slot.
, then moved up to 4 sticks for the same CPU - all good so far. Added in the second 7551 CPU and an additional stick in the B1 slot - and now no display on the VGA. I logged in through IPMI and find a boot code of 91. I loosened the screws, tried remounting the CPU - still the same - Any suggestions? I realize that the RAM isn't on the compatibility list but it is booting with 1 and 4 sticks for a single CPU.

Also weirdly my H11DSI board doesn't appear to have the Oculink headers in the location that the H11DSI-NT has as can be seen with the highlights, anyone else with an H11DSI have that or are the ports only present on the H11DSi-NT?
i also have h11dsi (not NT) and i neither have oculink. so oculink is only for NT-version.

and about memory...i got 2 epyc 7551 on rev.1.01 board with 10 memory sticks (5 sticks per cpu) and it works okay, from day one. and my ram is neither on the qvl list


Supermicro stuck on 91

and also try to update bios
 
Last edited:
  • Like
Reactions: kaltek2599

kaltek2599

New Member
Feb 29, 2020
8
0
1
i also have h11dsi (not NT) and i neither have oculink. so oculink is only for NT-version.

and about memory...i got 2 epyc 7551 on rev.1.01 board with 10 memory sticks (5 sticks per cpu) and it works okay, from day one. and my ram is neither on the qvl list


Supermicro stuck on 91

and also try to update bios
Thank you for responding - that confirms I'm not missing Oculink headers.
Are you running any Nvme drives on it? Using PCI-e then?
Would this motherboard boot off a PCIe drive?

My RAM shouldn't be a problem then.

My Rev 2.0 BIOS is at version 2.0b currently.
2.1 is the only newer BIOS on the Supermicro website with the following enhancements
==========================================================
1. Changed BIOS revision to 2.1.
2. Updated AGESA RomePI to 1.0.0.5 based on 5.14_RomeCrb_0ACMK013.
3. Displayed "PCI AER Support" setup item on ACPI page.
4. Added SMC HDD Security feature

and fixes

1. Fixed issue of system hanging at post code A7h.
2. Fixed inability of SUM to change the function of NUMA Node Per Socket.
3. Fixed problem of system sometimes rebooting during legacy Windows 2019 OS installation when using Rome CPU 7502.
4. Removed requirement to use Admin password for erasing TCG device.
==========================================================
I'm curious why it started erring with 91 only after adding the second 7551 CPU though, no drives are currently attached to SATA or USB currently. I was trying to get it to successfully boot first before adding more RAM and then drives.

I'll try the CMOS reset as the link seems to indicate. Hopefully get this going sooner rather than later, thanks again!