Yes, the firmware was ancient! You'll notice the later boots included in my dmesg show the interfaces with the latest firmware:
ixl0 at pci1 dev 0 function 0 "Intel X710 SFP+" rev 0x02: port 0, FW 9.152.77998 API 1.15, msix, 8 queues, address 40:a6:b7:b3:4b:28 ixl1 at pci1 dev 0 function 1 "Intel X710 SFP+" rev 0x02: port 1, FW 9.152.77998 API 1.15, msix, 8 queues, address 40:a6:b7:b3:4b:29 At any rate, it was the transceivers as predicted by several posters. My experience matches what deich...@placebonol.com reports: The NIC enforces intel-coded only for optical transceivers but not twinax. Thanks again, everyone. weaver On Fri, Oct 4, 2024, at 11:09, Zack Newman wrote: >> I've already upgraded the NVM firmware on the Intel card to the >> latest, no change in behavior > > Are you sure about that? Your dmesg says differently: > >> ixl0 at pci1 dev 0 function 0 "Intel X710 SFP+" rev 0x02: port 0, >> FW 6.0.48442 API 1.7, msix, 8 queues, address 40:a6:b7:b3:4b:28 > > In contrast my dmegs outputs: > > ixl0 at pci1 dev 0 function 0 "Intel X710 SFP+" rev 0x02: port 0, > FW 9.152.77998 API 1.15, msix, 4 queues, address 40:a6:b7:b5:4a:c4 > > According to Intel, your firmware is more than 6 years old: > https://www.intel.com/content/www/us/en/download/18639/28335/non-volatile-memory-nvm-update-utility-for-intel-ethernet-adapters-700-series-efi.html > > That's to not suggest that's the reason mind you. I've been issue-free > with my Intel X710-DA2 using a customized 10G SFP+ passive direct attach > copper twinax cable from FS with the other end connected to a Juniper > EX2300-24MP switch.
