The most common 'bottleneck' is the server connection to the workstation - it's often way faster to access data 'locally' than to the network. If you copied the data to the C drive and performed the same steps (likely similar to the test mentioned by working directly on the server), you will notice a significant saving in processing time.

So things to look at are the hubs and switches and cards being 100MB/sec rather than 10 - any one can get in the way. Also remember that you can never get data access over the network to have comparable performance than local data access.