One of the main reasons I dislike using simulation racing games for comparisons of processing power is they often have substantial "replay mode" functions that produce far superior images than are actually possible in game.
The best way to do screenshot comparisons is to find similar games for both systems, find similar environments for both games, and capture images using the same method. There are some challenges to this though, while gaming sites captured images from Dreamcast and PS2 games off of the RCA cable output, and some sites captured Gamecube games from progressive scan output, the XBox allowed people to take screenshots of games straight out of the image buffer which produced far cleaner images.
edit: As a general rule, if you see PS2 or Gamecube images which are crisp and clear they're usually marketing shots (that tend not to reflect the in game experience) or people who captured them off of an emulator.
AMEN
The Gamceube doign what it does best. Processing large volumes of gemoetry and shading "really" really" fast and fluidly.