Is anyone seeing significant bandwidth savings with PCoIP in View 5? Does the local caching seem to make a substantial difference? I am looking to run about 15 users over a single T1, and with View 4.6 there is no way - they are taking up about 500k each right now.
This document goes over some of the bandwidth improvements in View 5.
Have you played around with reservation for the VM?
Maybe set a reservation at 1000 Mhz and 1000 MB of memory could help.