A lot of the difference likely comes from the fact that hyper implements HTTP keep-alive, meaning that multiple requests can be handled on the same connection, vs. having to create and terminate a conbection for every individual request.