Just what do you think they're benchmarking exactly then on that website in my OP?
The entire gamut of typical types of operations that games utilize. Not the extremely small subset of operations that hashing requires.
What he said, basically it's a assigning a number (that you can compare to other computer setups) about how well it ran the game. It's essentially pushing the card how a game would and recording it's response. Naturally that is going to include a lot of things we never would use and be mostly irrelevant.