Aspeed BMC melting

[Labeled]

I've had my aspeed chip melt 3x times now in an Intel 2308wtfzsr.

 

If the data center has poe running into my management port could that cause it? 

[AbydosOne]

11 minutes ago, Labeled said:

If the data center has poe running into my management port could that cause it?

In theory, Ethernet should be transformer coupled, so the PoE DC bias voltage would be negated and not be an issue. I can't speak to this specific implementation, but if PoE were the problem, it would be a problem immediately.

 

11 minutes ago, Labeled said:

I've had my aspeed chip melt 3x times now

The same Aspeed 3x or three different motherboards? Can you confirm the IPMI is working at all?

[FlyingPotato_is_taken]

I am not familiar with the ethernet specifications in detail. From a technical standpoint there should be a inductor/transformer between the outgoing ethernet and the internal wires. So even with 48V DC applied (or what POE+) uses shouldn't cause any issues. Also to get a chip to blow up like this it takes a significant amount of energy. More or less only a lightning strike on an ethernet cable would carry enough energy for this failure.

 

 

The major issue is, that this is some sort of electronic design flaw and without a schematic those are really difficult to understand. Normally I would say contact the solution provider/manufacturer but this is EOL and to make it worse Intel. Likely the best option is to just write this equipment off and purchase a different server.

 

Assuming you could get the ASPEED IC and programming than it would be still questionable if this is recoverable as an IC failure like this often mean that at least 1-2 layers below the chip are also burned and need rework.