AMD EPYC Genoa Engineering Samples

[featureless_steel_box]

Recently grabbed an EPYC Genoa 100-000000894-04 engineering sample. I haven't been able to boot it - curious if anyone else has tried this before and the results they got.

Motherboard: Supermicro H13SSL-NT
Cooler: Dynatron J12
PSU: Brand new EVGA 750 GT (ATX v2.52)

Motherboard BIOS is version 1.2

With the CPU installed and no heatsink, the BMC heartbeat shows up as expected and I can access the management port with no issue.

However--when the heatsink is installed and tightened, no LEDS show when given power and the board acts completely dead.

Seems like I might be in the same boat as this thread regarding SP3, since also don't have the system power LED on at any point.

When using a 24-pin jumper the CPU fan does come on.

I'm not sure what to suspect right now, as far as I can tell this configuration should be supported, maybe the board is faulty?

 

[Dev_Mgr]

Do engineering samples do that CPU lock to a vendor? If so, maybe this CPU was used on a non-Supermicro board and is now locked to whichever vendor's brand motherboards?

 

[krisq]

You need to check memory compatibility.

 

[featureless_steel_box]

You need to check memory compatibility.

Using 8x Samsung DDR5 4800 RDIMM M321R8GA0BB0-CQK
I should still see board power/bmc LEDS with no memory installed I would think?

 

[featureless_steel_box]

Do engineering samples do that CPU lock to a vendor? If so, maybe this CPU was used on a non-Supermicro board and is now locked to whichever vendor's brand motherboards?

Not using a Dell/HP/Lenovo board. I'll ask the seller about the board that was used to test with.

 

[RolloZ170]

However--when the heatsink is installed and tightened, no LEDS show when given power and the board acts completely dead.

either the cooler is completely sick or there is a short circuit during installation.
inside a chassis or open table ? forgot o remove mobo standoffs ?

Do engineering samples do that CPU lock to a vendor?

afaik not. unknown to genoa/bergamo but i doubt that.
even with some locked CPU, the IPMI WebGUI can be accessed.

 

[featureless_steel_box]

either the cooler is completely sick or there is a short circuit during installation.
inside a chassis or open table ? forgot o remove mobo standoffs ?

On cardboard. I didn't consider the heatsink, I'll order a different one to see---although options are pretty limited right now.

The seller tested with an MZ33-AR0

edit: my Dynatron suffered the same damage as mentioned here and was a bit awkward to install. You might be right!

 

[RolloZ170]

first i would remove the CPU out and see if something happened with the socket.

 

[111alan]

That's why I hate ILM. For a cpu this large the ILM may offer uneven pressure, combined with the pressure of heatsink the CPU may bend and create non-contacted or electronically unstable pins. LGA4094 is already a nightmare to troubleshoot, sometimes even losing a memory channel after a reset.

I would suggest you to clean the CPU contact pads with alcohol or IPA. It should be cleaned to a degree that there aren't any uneven colors when shined with a flashlight.

 

[featureless_steel_box]

first i would remove the CPU out and see if something happened with the socket.

Socket looks good. I didn't try to force it on and now I can see that the two bottom posts of the Dynatron are too warped to let it mount cleanly, so it likely never had even & correct pressure. Not going to bother trying to repair it, will update when the new cooler arrives.

 

[featureless_steel_box]

Up and running now. Had one failing Samsung that was causing memory channel oddities. Board came online without error after correcting the mounting/heatsink.

Averaging 110k on cinebench r23.