Hello Warren,

in iBGP you can take advantage of BGP route reflectors and/or confederations to reduce the number of iBGP sessions.

Also, what counts is the number of update-groups = number of different outbound policies.

Other possible differences may be in the BGP scanner related timers and on the possible use of BGP next-hop tracking to reduce CPU usage.

The first ones can have different values (but it should per address-family and not per BGP session type iBGP vs. eBGP) and BGP next-hop tracking may fit with iBGP scenarios

Apart from this, generally speaking I have no evidence of less use of CPU in iBGP in comparison to eBGP

Hope to help

Giuseppe