Originally posted by hithere
I think there's a misconception here: Users aren't benchmarking a card alone when they use "real-world" benchmarks (games). You contend that games treat different cards differently, therefore there can be no valid comparison. But what if the user doesn't care which card is "superior"? All the user wants to know is which card will deliver the best performance in his/her favorite game, no matter how that game "treats" either manufacturer.
If users don't care which card is "superior", then they are obviously not benchmarking the card. They are looking for which card is best for which game. Such a user must have money. Card A is best for Game X while Card B is best for Game Y (where best means fastest performance). Right off the bat, users need two cards.
Real users don't want to keep switching cards based on what game they happened to be playing at the time. So they need an apples-to-apples comparison, a benchmark that is not biased.