The benchmarks indeed do not fully reflect the actual throughput, but if you take the Whetstone and Dhrystone separately, they are a fair indicator, if same client and same OS bit... speak all 32 bits windows for instance.

The "Results" measure is if you mean the number of completed tasks very dubious simply because none have the exact same load of fpops/ints to compute... very non-deterministic. You'd have to set up an extensive testing program and let them run each for a longer time on a project-by-project basis to find what's best.

Generally for multi-core devices I find that mixing tasks, that will not be competing for the exact same CPU function gives an on-average better result. The other is that the lighter devices have a harder time to compute HPF2 and FAAH/HFCC.

Averages, yes, I'd like to see them too... sliding 7-14 days, not the ad-infinitum as what we get on the My Grid front page.

Now for the quick fix: Get BOINCTasks which can monitor all devices on your LAN. It tracks completed tasks up to 365 days, i.e. a result history with CPU time / Wallclock Time, but not credit of course. You can choose any device and sort the different info columns. Not tried, but I'm sure you can dump that information to a spread sheet for all sorts of fancy analysis. Asking for a feature on the website is asking for something that takes many months to plan in, if considered useful.

First I would like to congratulate you about this concept and this initiative. Excellent idea to breathe a new life into old machines that can do useful work instead of polluting the environment.

What do you mean by "how much power is available". Do yo mean in term of watts, so it is a technical limitation, or do you mean in terms of energy cost and then it is a financial limitation.