Beside the fact that your test servers are not alike and it is probably not going to ever be the same, measuring performance using a file transfer is not the right way of finding the root cause.
I do understand where you are trying to get but you need to make sure you test your performance properly.
First thing first: the file transfer is done over network, so we will need to test the network only prior. Next, you can test the rest of the E2E elements (test the local server's performance, RAID performance, etc).
for only testing the network's performance i recommend on using known benchmarks like iperf - run it in a client-server mode and run at list 8 threads so you are utilizing your server resources properly.
Other then that, you can look into the server performance tuning guide: http://www.mellanox.com/related-docs/prod_software/Performance_Tuning_Guide_for_Mellanox_Network_Adapters.pdf
hope it helps..