Weird bug 7.15 x86 - NIC stops working until full RouterOS reinstall

I am testing out two 25gbit ConnectX-4 NICs (MCX4121A-ACAT - latest fw from nvidia’s site) and I am experiencing a problem where if I insert one NIC it works fine, if I try to add a second, the old NIC stops working. Removing the new NIC that was added doesn’t fix the original NIC either. I tried int reset, disabling and re-enabling, rebooting etc. The only thing that fixes it is re-installing RouterOS. What am I missing here? Did the first NIC get blacklisted/disabled with no way in the CLI to reset it or is there a special way to truly reset this NIC or at least wipe all NIC/interface config so a complete reinstall is not needed.

I do suspect the issue may be related to not enough PCI lanes on the server or something like that, but RouterOS should not be blocking out the NIC with no way to use it again until a full reinstall, that’s what I would like to solve to make trouble shooting easier.

Just a guess, but your NIC probably isn’t supported. If that’s the case, I’d recommend running RoS using CHR instead.

Others have said the NIC has worked well for them on RouterOS x86. … and it does work well with a single NIC when I tested.

Where is the official supported NIC list for RouterOS x86?

Either way, no routing software should be disabling an interface permanently like this,

The only advice I can give, if you can’t ensure your system is ROS-compliant, is to use CHR. If configured correctly, you won’t notice any difference in performance.