UCS M4 (4.1(2k) and 4.1(2m)) + UCSC-PCIE-ID25GF + FAN POLICY OVERRIDE

[marek.hajduczenia]

Good morning, Forum 

In the lab, we have two UCS M4 units with firmware 4.1(2k) and 4.1(2m) and I was asked to upgrade old 1G NICs with UCSC-PCIE-ID25GF (0x8086 / 0x158b / 0x1137 / 0x0225) units for fiber connection. The cards are properly recognized and they are part of the supported PID catalog (both M4 units had the catalog upgraded to 4.1(2h)03, latest version available on the official website): https://www.cisco.com/c/en/us/support/servers-unified-computing/ucs-c220-m4-rack-server/model.html. 

Despite it being an official Cisco card, it seems to be kicking off FAN POLICY OVERRIDE power policy, as shown below. 

The units were shut down properly, power characterization was re-run just in case, PID catalog was updated, and yet the problem persists. When replacing the card with older model (UCSC-PCIE-IQ10GF, 0x8086 / 0x1572 / 0x1137 / 0x013b) everything seems to be fine and no FAN POLICY OVERRIDE is observed. 

I am baffled by it and all my searches for a similar problem do not come across any solutions / hits, outside of "update PID catalog", which was already done. 

Any clues / hints on what might be going on? Two M4 units, with slightly different firmware have the same issue, so either it is a systemic problem or there is something wrong with the PID catalog available online for M4 units. 

thank you in advance !

[Steven Tardy]

Where do you see this XXV710 is on the PID list for M4? I'm not finding it anywhere? I see X710 10G, but not the XXV710 25G.

C240 M4 spec sheet does not include UCSC-PCIE-ID25GF.
C240 M5 spec sheet does include UCSC-PCIE-ID25GF.

Not surprised a newer M5 generation card isn't happy on an older M4 generation server.
Any PID from the M4 spec sheet should not cause fans to spin up.

This fan spin up _feature_ is a protection mechanism since CIMC doesn't know the capabilities of that card to cool itself.
The CIMC assumes the (unknown) card needs all the cooling it can get, so the fans spin up to assure the part has adequate cooling.

Personally I think this is a goofy default behavior for unknown PCI devices.
But I'm sure UCS Engineering will say "install a _supported_ card".

[marek.hajduczenia]

I found the PID list here: https://www.cisco.com/c/en/us/support/servers-unified-computing/ucs-c220-m4-rack-server/model.html

UCSC-PCIE-ID25GF is listed in the "Alternate PIDs for this product line:" and from what I can find, these are expected to work and be "compatible" with the product line, and yet they seem NOT to be. 

And yes, agreed on the goofiness of this solution. Dell seems to not care much about what cards are being loaded, but I wanted to make do with the cards I have around for these M4 units. If these do not work, I will have to get older 10G cards. 

Thanks!