New Supermicro Supermicro H11DSI-NT Amd Epyc DOA?

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Leo Atreides

Member
Dec 23, 2020
30
11
8
@RageBone


Okay ! Here is the lastest results from what has until now been an expensive and depressing adventure into building a lowcost AI research machine.

This is the latest as of 15 minutes ago boot outcome. Post code 55 ( only one dimm in there in c1 ) and no other media.

20210817_181706.jpg

Turns out that the reference CPU I was using is bust ! So thats was definately 2 mobos that definately did not post and a reference CPU that did not post.

I guess I will work on the rest of it over the weekend.

Anyone know what the sucess post code is for this kind of system ? I am getting 55 ( memory not installed ) is that storage media ?

( Thanks to all for the help and advice so far )
 
  • Like
Reactions: RageBone

RageBone

Active Member
Jul 11, 2017
617
159
43
The known old AMI Postcodes are not usable with that system.
In my own experience Code 55 is occurring multiple times and has no reasonable discernible meaning.
If it is stuck there, you should have a picture out already with more information.

Later in the Process, it appears as if there are multiple codes overlapping each other at the same time.
55 Is the one that wins out most of the time. At least with my old crappy Asus Postcode card.
So Later on within bios and on, you can ignore it.

EDIT: my crystal ball tells me that code 55 is fricken useless. Ignore.
 
Last edited:

Leo Atreides

Member
Dec 23, 2020
30
11
8
After a world of stress and dissappointment and several busted boards I overclocked to 3.45ghz and I got a really good score in Cinebench. Maybe 10% faster than a single threadripper pro at 80368

When I get this off the bench and into the case ( second attempt - other motherboard prob shorted or something when i added the video card ! ) I think I will add a write up on just how difficult it was to make this.

All thanks to ExecutableFix and his overclocking hack. I dont think I will be looking to puch the performance much further for now as I tried 3.65 and 3.8ghz and that resets the machine. The supermicro boards are unforgiving with power problems as far as I can see.

I am wondering if its a power supply compatibility problem that popped the motherboard(s!!!). I have a Strider ST1500-ti and I find it a little unnerving that after powerdown ( where there is say 10 seconds of led activity before power off ) after about 30 seconds the power comes back on for 3 seconds. Caught me out a few times. Also rebooting during a bios startup pretty much bricks a v1 board.

I can probably do what I intended with this system, which is build Android AOSP ( currently building in 24 minutes on my 48/96 core server ) and I want to use to train tensorflow models too.

small_cinebench_3.45ghz_dual_ZS14.jpg
 

Andrew_Carr

Member
Mar 18, 2022
37
11
8
I had the same issue: code 78 on an Asrock Rack EPYCD8-2T with a 7551 from eBay. Turns out the issue was that the CPU was pulled from a Dell system, which burn a key into the CPU such that it will only boot a Dell-signed BIOS. This makes the CPU unusable in any non-Dell system.

Thanks for the e-waste, Dell.

More details here: https://downloads.dell.com/manuals/common/security_poweredge_amd_epyc_gen2.pdf
Thought I'd chime in here as well and say that I bought a couple used 7551P CPUs off ebay and while both work fine in the asrock rack epycd8-2t that I have, neither will work in the supermicro h11ssli rev 2.0 motherboard. I can get into the IPMI though and was able to update the BIOS to see if that was the issue but it still gets a "ff" POST code. I've tried multiple sets of ram and just spent $75 on a stick from the QVL list in C1 and it still doesn't post. I guess I could try these in one of my h12dsi motherboards but it seems like that wouldn't work. Weird that my 7601 ES/QS models would all work flawlessly while the retail 7551P won't, so maybe it's dell locked and that wasn't mentioned by the seller or something, but I didn't think it'd work in the asrock motherboard if that was the case.

When I get this off the bench and into the case ( second attempt - other motherboard prob shorted or something when i added the video card ! ) I think I will add a write up on just how difficult it was to make this.

All thanks to ExecutableFix and his overclocking hack. I dont think I will be looking to puch the performance much further for now as I tried 3.65 and 3.8ghz and that resets the machine. The supermicro boards are unforgiving with power problems as far as I can see.
Not sure if this helps but I have three dual 7601 es/qs builds going right now in h11dsi motherboards. I'm generally using seasonic ss-850km3 PSUs but I'm not running video cards. I just got some cheap 4U cases to make assembly easier space-wise. What CPU coolers are you using? I run Dynatron A38s or A26s on mine where I have the space, and dynatron a31s on my 7551P 1U builds. I tried on a39 as an experiment but the baseplate doesn't fully cover the CPU lid and I had instability problems even though temps seemed normal. I swapped that out for an a38/a26 and now it's running fine. I also had some issues caused by I think not seating the CPU/heatsinks tightly enough and after reinstalling everything that fixed an issue where I was having constant reboots.
 
Last edited:

RageBone

Active Member
Jul 11, 2017
617
159
43
PSB Locks on Epyc have been identified as getting stuck at postcode 78.
Other Postcode and especially FF, do not point toward that.

FF on SM boards is the absolute first postcode.
It staying there means that nothing is happening in terms of BIOS / software doing things.
That can either be an issue with the BIOS or your a fault of the board and or CPU.
Usually, the memory can't cause issues this early in POST.

while both work fine in the asrock rack epycd8-2t
So the CPUs are itself confirmed working and not PSB locked, they would not work otherwise.

I can get into the IPMI though and was able to update the BIOS to see if that was the issue but it still gets a "ff" POST code.
Which Bios did you flash?

Weird that my 7601 ES/QS models would all work flawlessly while the retail 7551P won't,
Yes, weird indeed because that means that the board is confirmed working with other CPUs right?
I am not aware and sure that there are specific BIOS mods needed to get Naples ES running.
If yes, the most likely culprit is the BIOS again.
 
  • Like
Reactions: Andrew_Carr

Andrew_Carr

Member
Mar 18, 2022
37
11
8
Which Bios did you flash?
Ah shoot, I used the first one on this list. Didn't realize it was the rev 1.0 BIOS. Maybe reflashing will help (nope, didn't help)

I haven't tried other CPUs in this board yet but it's my second h11ssli that's had the same exact issue, I returned the first thinking it arrived DOA. They were all purchased new via ebay/alibaba. I'll try reflashing with the other bios file now. Yeah, I wasn't sure if the memory was the problem or not but heard someone with a 7551P and the same board fixed their issue by changing the memory out so I bought the exact same model (on QVL list I believe) to see if that would fix it.

[Update: I navigated away from the upgrade screen before starting the bios reflash and it got stuck in " Service is not available during upgrade " error even after resets -> Removing current settings under factory default seems to have cleared it and I was able to reflash the bios with the correct file. Updated IPMI afterwards and now at least when I try to power it on via IPMI i get a "Performing power action fialed. Please check" error. ] I think I'll try swapping the CPUs after this. I may also try one of the 7601s as well to see if a different model works.
 
Last edited:

RageBone

Active Member
Jul 11, 2017
617
159
43
before you insert another CPU, please make sure that the CPU you had in it, does still work.
You surely don't want to put another CPU into aboard that KILLS CPUs.

Not that we know that it does, just please make sure it really doesn't! !!!
 

Andrew_Carr

Member
Mar 18, 2022
37
11
8
before you insert another CPU, please make sure that the CPU you had in it, does still work.
You surely don't want to put another CPU into aboard that KILLS CPUs.

Not that we know that it does, just please make sure it really doesn't! !!!
Ok, did that and both CPUs are still fine. Put the one from my h11ssli into my asrack board and it's running fine. The 7551P that was working fine in my asrack is back in the h11ssli again but no luck, still getting the "ff" post code. One stick of memory in the C1 slot. Haven't gotten anything useful out of supermicro support yet either. Next step tomorrow is to I guess try a 7601 QS processor to see if that works. Tried another power supply and removed the auxiliary 4pin power from JPW1 , getting different post codes now. Maybe it's fixed, seems to be going through normal post process now. I think I tried everything, inside the case, outside the case, 4 types of RAM, two motherboards, two processors, can't believe it was the PSU I just bought, thought I had tried replacing that for sure.

Yup, new power supply or removing the extra 4pin connector fixed it. I feel like an idiot.
 
Last edited:

Leo Atreides

Member
Dec 23, 2020
30
11
8
I had difficulties with FF codes. I can be a board short, it was for me. Make sure that you have the stand-offs correct. Nothing touching the board. Also, if you have masked any standoffs I advise you to remove them. As with many posts on this site, have you tried to make it work on the bench ? To do this, take the anti-static cover you got the board in and rest it on that with the cpu and one stick of RAM, power it up and IPMI into it. Thats how I got past an FF. If its still not working on the bench, request an RMA ( a return ) from your motherboard vendor.

( it was late when i replied, i see you fixed it ! )
 

gsrcrxsi

Active Member
Dec 12, 2018
293
96
28
Thought I'd chime in here as well and say that I bought a couple used 7551P CPUs off ebay and while both work fine in the asrock rack epycd8-2t that I have, neither will work in the supermicro h11ssli rev 2.0 motherboard. I can get into the IPMI though and was able to update the BIOS to see if that was the issue but it still gets a "ff" POST code. I've tried multiple sets of ram and just spent $75 on a stick from the QVL list in C1 and it still doesn't post. I guess I could try these in one of my h12dsi motherboards but it seems like that wouldn't work. Weird that my 7601 ES/QS models would all work flawlessly while the retail 7551P won't, so maybe it's dell locked and that wasn't mentioned by the seller or something, but I didn't think it'd work in the asrock motherboard if that was the case.


Not sure if this helps but I have three dual 7601 es/qs builds going right now in h11dsi motherboards. I'm generally using seasonic ss-850km3 PSUs but I'm not running video cards. I just got some cheap 4U cases to make assembly easier space-wise. What CPU coolers are you using? I run Dynatron A38s or A26s on mine where I have the space, and dynatron a31s on my 7551P 1U builds. I tried on a39 as an experiment but the baseplate doesn't fully cover the CPU lid and I had instability problems even though temps seemed normal. I swapped that out for an a38/a26 and now it's running fine. I also had some issues caused by I think not seating the CPU/heatsinks tightly enough and after reinstalling everything that fixed an issue where I was having constant reboots.
I'm surprised no one else caught this. but a 7551P is a single socket only CPU. it will not work in a dual socket board. you need a 7551 non-P model
 

RageBone

Active Member
Jul 11, 2017
617
159
43
I'm surprised no one else caught this. but a 7551P is a single socket only CPU. it will not work in a dual socket board. you need a 7551 non-P model
I'm pretty sure it will work alone in a dual sock board.
Of course you are wasting a socket, but that is viable as long as you have nothing else to put into that system.
I will try to use a H11DSi like that in a few days for testing.

If t really does not work (i doubt it), i will report back
 

Andrew_Carr

Member
Mar 18, 2022
37
11
8
Oh whoops. I read too fast, saw H11 and since this thread is about the DSI I assumed his post was on-topic.
Yeah, sorry, I hijacked it. I have some dsi boards too I was thinking about testing the processors in though since they weren't working in multiple ssli supermicro boards but were working in the asrock. They should work in a dual cpu board if you only put in one cpu (has to be in CPU1 socket I think or something). Ended up just being a bad power supply though.
 
  • Like
Reactions: gsrcrxsi