A redhat engineer has reproduced the problem in linux with 4 cores. I'm going to leave any further testing of linux up to them and continue working on optimizing freebsd. It sounds like the redhat perf labs will figure out the problem. If I hear back from them I'll post an update here.