Different ports on X710-T4L-t OCP showing up as different models in Windows Server 2022

[swalker-aup]

Hello,

We recently purchased four Dell PowerEdge R650xs servers, each one being equipped with a bunch of Intel network cards - including an Intel X710-T4L (OCP 3.0) card (Intel® Ethernet Network Adapter X710-T4L for OCP 3.0).

On these servers, which are all running Windows Server 2022 Datacenter with all updates, there is an issue: not all ports on the card show up as being the same model.

Port 1: Intel(R) Ethernet 10G 4P X710-T4L-t OCP

Port 2: Intel(R) Ethernet Network Adapter X710-TL

Port 3: Intel(R) Ethernet Network Adapter X710-TL

Port 4: Intel(R) Ethernet Network Adapter X710-TL

Following initial installation using the deployment from Dell Lifecycle Controller, despite the different device names, they all are using the exact same driver file and version: 

Driver provider: Intel

Driver Date: 7/20/2022

Driver Version: 1.16.206.0

I downloaded/installed the driver pack (Intel® Network Adapter Driver for Windows Server 2022*), which updated the driver to version 1.18.367.0 (date 6/27/2023), but the issue remains.

If I go into the Device Manager, choose one of the three devices labelled "X710-TL", open the Driver properties, and choose Update Driver  -> Browse my computer for drivers -> Let me pick from a list of available drivers on my computer, the only "compatible" devices shown are X710 and X710-TL series devices.  If I uncheck "Show compatible hardware", then choose Intel -> Intel(R) Ethernet 10G 4P X710-T4L-t OCP (Intel), the device will change name and show the proper name that the first device shows ("Intel(R) Ethernet 10G 4P X710-T4L-t OCP"), but I ever do "Update driver", Windows reinstalls the device under the incorrect name.

This is problematic because I am using these nodes as a Hyper-V Failover Cluster, and the Cluster Validation brings up warnings due to a network card team (which includes ports 1 and 2 of each X710-T4L OCP card) indicating that "Not all dapters associated with SET Team *teamname* on Node *hostname* are the same Manufacturer and Model".

 This issue has survived several reformats of the server, and lots of fiddling.

Anyone know why Windows is detecting the same ports of the card as different models?

Thanks in advance for any help you can provide,

Sam

[Yogaeasvaran]

Greetings Samuel,

Thank you for contacting Intel Customer Support.

Apologies for the delayed as we understand it's have been long since you get our response.

We would like to know if the issue still persists so that we can assist you further and could you also please provide us the current status of this issue.

Regards,

Yogaeasvaran

Intel Customer SUpport

[swalker-aup]

Hello,

Thanks for your reply.

Indeed, the issue persists on all four of our servers.  Here is a screenshot showing very concretely the issue:

To be clear: all four ports are on the same card, and it was sold to us by Dell as an X710-T4l OCP for OCP 3.0.

Despite the device name being different, all cards are using the same driver file (i40ea.inf_amd64_b35e4c7a2f3347dd\i40ea.sys) with the same version (1.16.208.0, published by Intel Corporation, date 1/20/2023).

The drivers are the latest version proposed via Dell via their Server Update Utility, and the card firmware was updated to version 22.0.9 also using this utility.

As mentioned previously, this issue does not block production, but:

• Results in Microsoft Hyper-V Cluster Validation indicating that the card models do not match, when adding two ports from the card to a network team, due to the names being different
• May result in three of the ports which show as being "X710-TL" not supporting the 5Gb/s and 2.5Gb/s multiGig Ethernet speeds, given that this is a supported feature of the X710-T4l but not the X710-T4.

Please don't hesitate to tell me if you'd like information related to our specific installation.

Kind regards,

Sam

[Fariz_Intel]

Hi Sam,

We sincerely apologize for any inconvenience it may have caused.

It looks like there's no update for the port display name at the moment, unless we know of some other workaround to manually configure the display name on Windows environment. 

Have you consulted with Dell since certain OEMs might have their own way to configure the naming part?

We can cross-check with the engineering team, and we might need you to share the Microsoft Hyper-V Cluster Validation results. However, this could take some time for them to review.

Best regards,

Fariz_Intel

[swalker-aup]

Hello Fariz,

I have consulted with Dell, who indicated that it was a known issue, and indicated that the fix would need to come from Intel.

Whether the fix needs to come from Intel or not, I am not certain.  There is no need to provide Hyper-V cluster results to confirm something that is shown in my screenshot: Hyper-V complains when there are two different "models" of network card in a network team, and as shown in my screenshots, the two ports of the same card show up as different "models" to Windows.  Although Hyper-V is what is complaining about the issue, you don't need Hyper-V to reproduce the issue, just to have Windows installed.

Having some kind of fix from Intel, or at least an acknowledgement that there is an issue, would be helpful.

Best,

Sam

[Fariz_Intel]

Hi Sam,

If you have additional questions or clarifications, feel free to let us know.

Best regards,

Fariz_Intel

[Fariz_Intel]

Hi Sam,

Yes, the issue is acknowledged even without Hyper-V and the Hyper-V result is actually not for confirmation but rather as additional details that we might use to provide some suggestions or alternative methods. There are some differences between the Intel driver and OEM driver hence its also good to get some insight from OEM as well.

Kindly allow some time for us to further check on this issue; it may take some time. We will get back to you once we have an update. We sincerely apologize for any inconvenience it may have caused.

Best regards,

Fariz_Intel

[Fariz_Intel]

Hi Sam,

We sincerely apologize for any inconvenience this may have caused. It appears that there is an issue with the 700 series silicon, and the only resolution may require a silicon spin/revision, which is very unlikely at this point.

This issue was acknowledged as an errata in the 700 series specification update document in 2017. For your reference, you can find the latest specification update, which addresses this matter, by searching for 'PCIe Subsystem ID Incorrectly Reported.' - https://cdrdv2.intel.com/v1/dl/getContent/331430

Unfortunately, at this moment, we are unable to provide any additional solutions to address the issue. We understand the impact this may have, and we sincerely appreciate your understanding.

Best regards,

Fariz_Intel

[Fariz_Intel]

Hi Sam,

Just in case, if you need any additional information, please submit a new question, as this thread will no longer be monitored.

Best regards,

Fariz_Intel