This is a very interesting article about better ways of benching rendering speed in games:
http://techreport.com/articles.x/21516/1
Instead of measuring fps which by giving an average for a whole second hides some rendering problems, they measure the time of individual frames which allows them to see problems with graphics cards that while maintaining a good fps have some frames rendered very slowly which the human eye picks up on. Certainly highlights problems with Sli/Xfire and microstuttering.
http://techreport.com/articles.x/21516/1
Instead of measuring fps which by giving an average for a whole second hides some rendering problems, they measure the time of individual frames which allows them to see problems with graphics cards that while maintaining a good fps have some frames rendered very slowly which the human eye picks up on. Certainly highlights problems with Sli/Xfire and microstuttering.