The guy who made that site did the graphs all wrong. What you are supposed to do to make really biased looking benchmarks is to start the scale super high to any difference between speed looks like a huge difference. Companies do this all the time.
For example, lets say CPU1 gets 100fps and CPU2 gets 90fps. For the graph, the minimum fps shown will be 80 and each notch will represent 1fps. CPU1 which gets 100fps will be 20 notches. CPU2 which is 90fps will be only 10 notches.
Just by looking at the graph it looks as if CPU1 is twice as fast even though it's not.