i210 problems at temperature

[DaveHaynie]

Hi all-

I've had a series of similar products in production since 2020, using the i210 in a system along with an Intel-based embedded processor. 

I've recently started to get failures at temperature, and pretty weird ones. We're good at room temperature, we're good down to -40C, but at a 70C ambient we're seeing some units not simply not find the i210, but apparently lock up the PCI Express bus in some weird way, and/or our switch chip. I have a Diodes, Inc/Pericom PI7C9X2G912GP connected over an x4 link to the x86, with x1 links supporting six PCIe MiniCard slots and the i210. 

When this occurs, we don't see other devices on the PCI Express bus that would normally be enumerated after the i210. If we hold the i210 in reset (my system has individual PCIe resets for every device), all other devices configure as normal. This is only at our temperature limit, everything's hunky dory at lower temps. 

Curious if anyone's seen anything like this in their work. Thanks! 

[Diego_INTEL]

Hello @DaveHaynie,

Thank you for contacting Intel Embedded Community.

I was looking further, but maybe could be that you are using a commercial model, from the i210 datasheet:

Advanced Features:
— 0 to 70 °C commercial temperature or -40 to 85 °C industrial temperature

It could be that the device is not working fine at that temperature and thus, not being detected or provoking some miscommunication in the PCI bus.

Document #333016 - Intel® Ethernet Controller I210: Datasheet

https://www.intel.com/content/www/us/en/content-details/333016/intel-ethernet-controller-i210-datasheet.html?DocID=333016

Best regards,

@Diego_INTEL 

[DaveHaynie]

Hi Diego-

Nope... we're using the industrial spec part.  That was one of the first things I check on with our PA folks. Since we've been making these products using the i210 since 2020, we've had some times when we were buying them from brokers (at insane prices) and yeah, you'd have to wonder. But the parts we're using in production now are genuine industrial spec parts. 

We've managed to put in some software tricks to mostly mitigate the problem. We're seeing the i210 take up to 40 seconds to "wake up" when cold. We have an overall ability to turn on heaters in the system, shown necessary for a few PCI/PCIe radio modules we've bought from OEMs (speced industrial, but some in real life didn't work until around -20C), though I never had planned that to be needed for anything on the main board, and it's impractical to spend that much time testing in production (it takes a few minutes to fully heat the internals). 

The specifics of the problem became fairly suddenly worse in production in the last month or two, and we're still trying to get a grip on what, if anything, has changed. Oddly, this wasn't the only cold-temp disruption we've run into at about this time, but the other is unrelated to PCIe or the i210. Yes, this has been verified in multiple testing chambers! We didn't suddenly lose our temp calibration or anything :-). 

[Diego_INTEL]

Hello @DaveHaynie,

Can you share what processor and pch are you using for this project? I would like to check if I can find something else that may be of help.

Best regards,

@Diego_INTEL