RB433 + (2) XR5's + ROS3.22 = crazy high CPU usage?

We have (2) sets of RB433’s, each with (2) XR5 cards acting as a dual backhaul to two different locations (each Mikrotik/RB433 board, has (2) XR5’s, and serves as a backhaul to 2 different sites). We’re using ROS 3.22 on these backhauls.

We’ve noticed that even though these units are not pushing that much traffic (aggregate 10-20mbps), and they’re only bridging traffic (no routing, firewall, ACL’s, etc), they constantly report that they’re using at/near 100% CPU usage. We also believe that the high CPU usage is causing latency issues as well.

We have other RB333 based backhauls with only (1) XR5 card that are pushing just as much, if not more traffic (due to their location in the network) that are barely hitting 40-50% cpu usage and are not having latency issues.

A few notes and ideas:

-We only seem to be seeing this on RB433 based systems, with (2) XR5 cards in them.
-The negotiated data rates between sites hover between 18-48mbps, could the CPU be doing a lot more work because of retries?
-I really didn’t notice this issues until after we updated to 3.22, however we’re not seeing this effect on any of our other backhauls that run 3.22, and use RB333 boards, or only (1) XR5 card.

Can you check if you have ANI enabled in your router’s wireless settings?

Please excuse my ignorance, but what’s ANI, and how do I check to see if it’s enabled/disabled via WinBox? I could probably look it up using the CLI, but I’m really rusty as I never use it. Thanks!

OK, ANI: Adaptive Noise Immunity? If yes, then it’s enabled on both sides. On the parent it’s set to “ap and client mode” on the wds/slave it’s set to ‘client mode’.

Yes, Adaptive Noise Immunity. disable it and see if CPU usage changes.

That seems to have helped some. Disabling ANI on both wireless interfaces appears to have reduced CPU load by around 20-30%. It still jumps up to 90-100% sometimes, but not nearly as often… it’s no longer pegged. Thanks for that tip! Do you have any other ideas?