That might be a concern but if you run the same benchmark , the same amount of times, on the same hardware, with the algorithms being the only difference, then you still get a proper comparison between them, even if the timings are useless outside said benchmark