Supermicro X11DPi-NT Recognizes all ram except processor 2 node 1 (F1,E1,D1,D2)

[comradechaos]

Hi,
I originally got the motherboard on ebay sometime mid 2022 and until recently was using 2x5218 (Stepping 5, so E.S?) with 192GB of R.A.M. About 3 months ago my SSD failed (PCIe v2.0 x8) and due to personal circumstances I did not have the time/money to invest in fixing it anyway those circumstances have changed I have replace the ssd and got 2x8255c however I cannot get the system to detect the ram in the above mentioned channels no matter what I try (what I've tried)
- Switching the CPUs - Same result
- Swapping the ram - Same result
- Soaking the sockets in isoprophy alcohol
- Disassembling the entire system and cleaning everything and double checking that the motherboard mounts are level and nothing is shorting
- Countless hours going in circles (removing CPU, re-seating the ram, resetting the cmos by popping out te battery)

The result is the same those 3 channels get "Mapped out".

Any advice/help is massively appreciated, thanks for reading.

 

[RolloZ170]

The result is the same those 3 channels get "Mapped out".

if clear CMOS not help clear NVRAM (IPMI Firmware BIOS update)

 

[comradechaos]

Tried swapping the ILM's for the CPU's and still getting the same results, tried downgrading to v4.0 bios (only preserving the smbios) and still getting the same results, I guess it must be the socket but I'm certain I did not damage it and I cannot see any damage

 

[RolloZ170]

tried downgrading to v4.0 bios

unchecked preserve NVRAM ?
cmos is cleared if time date is reset, if no BIOS msg you have not cleared cmos.

 

[RolloZ170]

the slots mapped out are stored somewhere in NVRAM
slots are mapped out once the ecc uncorrected error threshold exceeded the limit (BIOS option/value)

 

[comradechaos]

Hi and thank you, yes I unchecked "Preserve NVRAM" but when I clear the cmos (by removing the small onboard battery) the date does go back to 1900 but I don't get any BIOS msg, I should mention I only bothered upgrading the IPMI firmware once a very long time ago (I think its from one of the v3 bios releases) so if you think that could have an ipact I can update it now to the current bios version of v4

 

[RolloZ170]

after that you need a full AC cycle. the BMC reads the RAMs SPD, not the BIOS.
CPU installation:
screw 3&4 must be turned (alternately, one turn per stroke) until the end,
the leaf springs must touch the cooler surface.

 

[comradechaos]

Ok, I've went through the CPU installation and the bios update process again, here are the steps I followed:

1. Log into the IPMI interface and clear the "Health Event Log", log out and switch off the system at the PSU.
2. Remove the bios battery and leave it removed for the rest of the process.
3. Remove CPU2 and re-install turning screws 3 & 4 alternately until the screws bottom out (previously I was tightening 3 and then 4).
4. Power on and upload the bios in the IPMI management interface ensuring "Preserve NVRAM" & "Preserve ME Region" are unchecked.
5. When its finished and asks me if I want to restart I say no and shutdown using the IPMI.
6. Remove power from the system (PSU switch) and wait 45 seconds.
7. Switch on the power at the PSU, system powers-up (as per default bios settings) and then power-cycles once itself and then (after about a minute) unfortunately get the same result (BIOS msgs informing me the DIMMS in D1, E1, F1 are mapped out)

Uploaded a picture of the heatsinks I'm using as I'm not sure I follow "the leaf springs must touch the cooler surface."

 

[RolloZ170]

7. Switch on the power at the PSU, system powers-up (as per default bios settings) and then power-cycles once itself and then (after about a minute)

because of the BIOS flash, ME/SPS has to initialize Flash File System.

unfortunately get the same result (BIOS msgs informing me the DIMMS in D1, E1, F1 are mapped out)

only idea is the VRM for that RAM bank is lost.
test with old CPUs ?

 

[comradechaos]

Cool, I'll try going back to the 5218's and see what happens, I did use a multi-meter on the RAM VRM's (I think, VITECH PR72-121) and they all report very similar readings v1.23~. Gonna swap for the old CPU's tomorrow eve and will update the thread, thank you again for talking the time (and I learned a few things)

 

[RolloZ170]

i remember one case with similar mapped out slots but not the whole bank, solved by replaceing the socket retention,
leaf spring was weak, worned. also visialy check pins again in/with backlight ( picture available )

 

[comradechaos]

Hopefully its not bad form to update this, only found the time to get back to this last week, redundency will do that, yay!

I did put in the 5218's and all the ram is detected correctly (as before).

In the meantime I managed to get 2 new foxconn socket retention's which did not have any impact unfortunately.

I also tried washing the board (white vinegar, soak in distilled water for about 10mins and rinsed with iospropal alcohol) and have visually inspected the pins under a cheap digital microscope but everything looks fine, again no change in results.

I guess I'll need to stick with the 5218's, really can't get my head this issue tbh, also no idea what I can upgrade to cpu-wise