Hmm, normally there should definitely be some improvement if only because it skips a request per item on first sync, unless the slow hardware somehow is the bottleneck.
Any chance you could post the server log to see if it's really performing the optimisations?
