i think the benchmark captures both of those, but it might be good to split out the uncached and cached cases. i'll make a note to do that.