We have been through this several times; most often with folks who do the upgrade, and then get desperate for some help.
On the same equipment, versions beyond 6.3.3 will be slower than 6.3.3. The larger your office(active users), the more noticeable that will be to you. Under six connections, probably manageable; 6-12 you will feel it, and above 12 you will probably hard roll-back or call us.
There are functional improvements, but we find the largest dividing line is whether the practice is going forward with MU. As more practices count the cost and determine they would rather jump off the MU train, there is much less reason to upgrade at this point.
For especially larger groups that are determined to do MU, you should anticipate having to spend some $$$ to make AC functional for your larger group. It could take a variety of forms, but one way or another it will cost $$$ or time, you just have to figure which of those you want to spend more of in doing MU.
As has also been noted here by folks using newer versions, 7.x in particular, you are entering into a BETA-like cycle of quick releases as problems are found and fixed. The 7.1.2 Repair-Will-Kill-You bug is just a recent example. For larger groups, that means more frequent (think about upgrading every client) upgrades, and *discovering* bugs.
The large group where we discovered and replicated the Repair-Will-Kill-You bug was very appreciative that we found it (and with AC's help) got it operable again after-hours as opposed to during the middle of the work day. NO ONE wants to make those kind of discoveries.
So, like all complex system considerations, the answer is "It Depends" on many factors.
In next month's video I'll be going through this in more detail, complete with engineer-produced flow-charts and slides. Some nice scenery at no additional charge.
