Are you using intel compilers on the local machines as well? What I've noticed, and what has been reported in other threads, is that the intel compilation takes a long time. My experience is that it's a factor of 15-20 times slower than PGI. But I would expect the model execution times to be similar assuming similar compute resources are used. I don't have any numbers to verify this at the present time. I also don't know why the intel compilation is so slow, but it's on my list of things to look into.
