We are now working with several practices that have upgraded with various outcomes.
As I have noted previously, the raw performance of the server is an important factor, but with the later versions, the over-riding factor for perceived performance is the number of concurrent users. It appears to be a log scale, but we don't have enough data to meaningfully graph it yet.
Other factors that are also appearing to have more bearing is the number of interfaces that are connected AC, and the frequency/depth with which they run.
