[SOLVED] H11SSL-i | No video, POST snooping = 16

[kere]

Hi,

I'm struggling with my first 1U build and wondered if anyone could point me in the right direction.

No video, POST snooping shows either (ff, 10, 16) - see below.

My best guess is the memory I have is not compatible. I've ordered a 16GB RDIMM to try.

CPU - EPYC 7551P (new from AliExpress store).
MB: H11SSL-i (new from AliExpress store). Rev 2.
MEM: 2x 32GB (used from AliExpress store).
- Random memory type sent according to the description.
- Received: 2x 32GB of Samsung M386A4G40DM0-CPB2Q
- Page 32 of the Super Micro manual mentions LRDIMM 3DS, 4R and 8R, Max Mhz of 2666.
- I'm not sure if LRDIMM 3DS 4R is the same as Samsung M386A4G40DM0-CPB2Q?
- DIMM Module Population followed (one, two and no memory).
PSU: Enhance 400W (new from AliExpress store).
VGA Enable/Disable (JPG1).
IPMI Post code sniffer:
- No memory = 10 (0X10 = PEI Core is started).
- 1 x 32GB = 16 (Pre-Memory North Bridge initialization (North Bridge module specific).
- 2 x 32GB = 16.
- ff appears as well (0xF9 – 0xFF = Recovery errors (PEI)).
IPMI not showing my hardware.
- Assuming showing the last device(s) used (testing at the factory?).

I appreciate the MB might be fried, and the CPU could be damaged but I'm starting with memory first.

Thanks,

Martin

 

[RageBone]

the postcode meanings you are listing are in my own experience sadly bogus.
Supermicro in general has 0xFF as its first and default "nothing happened yet" postcode, with other OEMs usually using 00.
So that has no specific meaning.

0x10 is expected when there is no memory installed in any of the channels.
0x16 points towards it being a memory issue, expected is a 0x15 and then some more.
From a leaked ABL-Postcode list, 0x16 could mean that there is no memory available in the system, which makes little sense to me from the wording alone.

From Rollos questions i'd say it is a very good idea to make sure that the PSU is good and capable enough for the system, and then test outside the case on a table or something with the bare minimum plugged in and nothing interfering.
If that code 16 persists, you likely have some faulty hardware.

 

[kere]

I’m serious about the PSU. It does have the 20+4 pin, which did seem odd.




What made you say ‘serious?’. Have I made an obvious mistake. Based on my currentknowledge it seemed sufficient

JPW1 and JPWR1, yes.

MB removed yesterday and same result, just put it on some wood.

CPU taken out and, as far as I can tell, all the pins seem okay




CPU pads (not the best photo / light):


Those faint marks top and bottom (right) are barely detectable in visible light.

re: Hardware update (after POST).

Thank you, that makes sense but it was confusing to see it initially.

 

[RageBone]

looks good so far.
Have you attempted a Bios-Update?
AMD is very confusing in terms of Postcodes with multiple different Lists going around and no differentiation between Error and Checkpoint codes.
Even worse, Some codes are 4 digits and the last two repeat, making 0x16 have at least 3 different meanings.

 

[RageBone]

those SM Boards have no detection on any of the power-inputs and a unified 12V plane.
So, you could run it with just a 24 pin. Of course that is a fire-hazard, but you could do that till it burns.

 

[kere]

RageBone, thank you for your input.

I just assumed the POST code “FF” was Fault Free, which aligns with what you’re saying.

Originally, I thought it was Fault Found… that made more sense as I couldn’t get to the BIOS. It’s a step learning curve.

0x10 gave me hope, with no memory installed.

0x16 is as far as I can get. The cryptic description of what it is for this MB just confused me.

It sounds the the ‘Northbridge’ might be an old name for something now moved into the CPU. They just kept the same name/terminology?

So, CPU or memory is where I’m at unless the PSU is the weakest link causing garbage.

 

[RageBone]

So, i my head the board notices that there is RAM installed otherwise you'd get a 0x10.
But there is still something wrong.

Please try one stick and put that into each and every dimmslot.
If the behavior is the same everywhere, that would mean that what ever fault causes it, was present on all slots.

Took an Multimeter to my H11ssl and i can say that 12V is unified a i said and the supply-voltage for core and mem.
So it is unlikely to be a fault with the input voltage only causing failures on the mem VRs.

Would you mind posting pictures of the boards front and back to rule out physical damage?

 

[kere]

RolloZ170, I’m a software developer and wanted to dropdown to hardware to see the full circle.

I’ve got no specific use case in mind at the moment. My goal was to simply put all the pieces together, install Ubuntu Server and install/run the various software I build / maintain everyday.

It was a kick in the teeth to fail at POST, which it’s literally the first stage. Gutted.

Power requirements, napkin maths, we’re based on the CPU and an SSD. That’s it.

 

[kere]

RageBone, I’m in New Zealand so I’ll take and add those images tomorrow (early hours in the morning here at the moment).

I’m fairly confident the board is not damaged but I’ll let you be the judge.

Mainly because when it failed first time around I assumed I missed something simple. So, I checked every component and triple-checked front/back and all the Jumpers, etc.

I’ve asked Super Micro about the RAM as 0x16 suggested, on my understanding, CPU or RAM. Pending response.

The manual mentions LRDIMM, which I have, but it also says ‘3DS’. Do you know if the memory below (pic) has 3DS?

 

[kere]

RageBone,

re: BIOS update.

No, I have not.

I poked around in the IPMI (Web) which concerned me.

The UI seems-very outdated (personal opinion), based on current web technology. It felt like going back in time to Windows 95.

The BIOS setting I did find was behind a Super Micro pay-to-play. Plus, various warnings about not updating it. So I left it alone.

This may sound stupid but it’s not meant to I just don’t have the knowledge yet…

I thought I’d need a success POST and BIOS load first before I can update the BIOS. If not, is there s resource you can point me to that shows how to it?

 

[RolloZ170]

I thought I’d need a success POST and BIOS load first before I can update the BIOS. If not, is there s resource you can point me to that shows how to it?

your CPU is supported by any BIOS revision.
BIOS update can be done with the IPMI GUI(Host Off) you may need a OOB licence, you can make your own then.

 

[jdnz]

MB: H11SSL-i (new from AliExpress store). Rev 2.
MEM: 2x 32GB (used from AliExpress store).
- Random memory type sent according to the description.
- Received: 2x 32GB of Samsung M386A4G40DM0-CPB2Q

H1SSL pages states the board needs ddr4 2666mhz (or faster ) - the ram you have is ddr4 2133mhz according to that samsung spec page

so first thing to do would be try it with some correct speed ram

 

[jdnz]

false. the speed depends on the memory controller in the processor.
AMD EPYC 7551P supports Up to 2666MHz(memory clock)
if you buy 3200Mz RAM the 7551P will work at max. 2666Mt

true for the CPU - but the SM page lists the only tested/supported memory as 2666/3200 - so whilst the cpu may support 2133 just fine there's no guarantee the h11ssl-i BIOS allows it

Has anyone actually used 2133mhz dram sucessfully on a h11ssl-i?

 

[kere]

jdnz, thank you for taking the time to offer your advice.

Your response to RoloZ170's 'false' comment has got me thinking...

That being... has anyone used 2133mhz RAM on the same version of the BIOS I'm running (specifics of what that is to follow).

I've not touched the BIOS, and it sounds like I can't without an OOB license (Is the OOB license this?).

When I originally had the IPMI running in the web UI (which I'll be doing again shortly) I was confused by the system hardware not being the parts I'd installed. It's been explained why that is. I have ordered a stick of the same memory reported on that screen assuming it was a successful POST that generated it.

If I upgrade the BIOS using the OOB license can I easily downgrade? If the BIOS upgrade still fails with the memory I have I want to be able to drop back down to the current BIOS for the memory on its way.

 

[kere]

The OOB license is a dead end at the moment from Super Micro's Store.

Their shopping cart, for credit card payments, only allows addresses from the United States. I tried to enter their address just to get through to the payment screen as not every company uses an address check on cards. Super Micro do, so payment fails as I'm in New Zealand.

Anyone else outside the United States knows how to get an OOB license, on a Sunday? They offer a bank transfer option but that's going to delay things even further.

The adventure continues.

 

[kere]

jdnz and RolloZ170,

Thank you both for the alternative OOD option.

Software is more my domain so hopefully I’ll be updating the BIOS

I’ve got some Isopropyl Alcohol today to give the CPU a once over now the socket pins have been checked.

Prior to that, as the CPU is in at the moment, I’m just doing the memory check, which was advice given in another comment.

So far I’m getting…

1. ff
2. B7
3. 16

While going through this process and using the IPMI at the same time the onboard speaker/buzzer activated. Continuous tone.

I didn’t even know it had a buzzer.

I immediately powered down as best guess is CPU heat throttle or similar. I’m running in a 1U open top case, with a passive CPU cooler.

Again, I’ve made another mistake based on my pure lack of hardware knowledge. I’m sharing my mistakes as it might help someone else reading this.

The 1U chassis came with five fans, which are all plugged in and working.

Try not to laugh… The fans max out at 4,000 rpm. The PSU fan kicks out more airflow.

I’ve just investigated fan speed and airflow. No, I didn’t do it before. The chassis came with fans so what more did I need .

It looks like 17,000 rpm is more like what I need.

Can anyone recommend a “this will do the job” fan(s) that I can order?

 

[kere]

OOB license activated. BIOS Image uploaded. Upgrade in progress...

 

[RageBone]

my best guess right now is that you got a corrupted bios that causes the postcode 16

 

[kere]

BIOS update successful.

CPU cleaned with isopropyl alcohol, SP3 pins double checked for damage/bends. Reseated and passive cooler added back on.

If I hit POST snooping ‘refresh’ fast enough I’m seeing one new code:

B3 = (as per PDF linked above) = Post-Memory South Bridge initialization is started.

Then I’m back to 16 and no video and no BIOS.

I’m just waiting on brand new 16GB memory stick (current 2x 32GB are used from AliExpress).

That was chosen based on the last known successful POST data in the IPMI (not mine).

 

[kere]

Just speaking with SM support while I wait for the RAM.

Support asked if the buzzer sounds during boot "Five short and one long" when no memory is installed.

Just tried... I get nothing. No beeps, no sound. The only noise is the PSU and 1U fans spinning up.

What does your H11SSL-i do during boot with no memory?

I've only heard the speaker once when the CPU was throttled (temp related). Instant shutdown during test.

If you've just scrolled here... this is my first build. You don't know what you don't know.