Synthetic tests aren't all that for guaging productivity performance - better to grab a handful of representative systems and just plain try 'em and see if they're acceptable.
Are they being reimaged? Because if not, how much cruft they have accumulated since we put 'em in will have far more of an effect than the bare system specs.
Futuremark do a 'productivity' test, but it's not very consistent and heavily favours the storage subsystems to get it's results. Which is relevant, but not as relevant as netowrk access to the files and the amount of RAM free to prevent system paging etc.