Originally Posted by Viral
Traditionally it never seems to translate into an FPS difference until you hit other limitations. But it's worth a try.
Think of it in terms of average FPS.
If the benchmark records 30-40fps 10 times per second for a few minutes then averages the results together, a few drops to 1-5 fps (a hitch\stutter) will hardly make any difference in the final average but will be very obvious in gameplay.
About the only way to show the results of a memory upgrade would probably be a line graph like [H] uses, to show the framerate throughout the entire benchmark. Ideally, more memory would result in fewer low spikes on a graph.
Fraps can record data like that. It probably wouldn't be too hard to do.