I might have missed a previous post here about this, but don't most timedemo utilities in games render a certain number of frames and report the time it takes? For example, two systems might return:
1250 frames/ 20 secs = 62.50 fps
1250 frames/ 33 secs = 37.88 fps
The important bit to realize being that even if a demo was recorded at 60 tics a second, a system could play it back faster or slower than that. If this is the case, won't the frame cap be inapplicable to benchmark results?