Quote:

but think there will be cases in the future that a code runs in 55 minutes on one computer and 1:05 on another and we will have this discussion again.



You could do what you suggested in an earlier post - rather than using real time as a metric, use user cpu cycles. This should be pretty close on main processor lines for the types of challenges that we are undertaking.

You can't include system CPU cycles as these are too easily influenced by whatever else the PC is up to.

Now all you need to do is calculate a reasonable user CPU load for 15 minutes