Absolutely wrong.
On a per-flow basis you want to push the work down as far
as possible down to individual cpus. Why do you think the
hardware folks are devoting silicon to RX multiqueue facilities
that spread the RX work amongst available cpus using MSI-X?
--