Michael> Sounds convincing. So the improvement is probably due to Michael> better send/receive parallelism we are getting in mthca. I won't have time to try it until next week but it should be easy to test the QP lock split without the IPoIB CQ split and see how that compares. - R.