Im not sure the article is entirely accurate - the RDRAM on the PS2 is not embedded on the EE, the VRAM is embedded on the GS (which accounts for the huge bandwidth). As I understand it, in simple terms, the PS2 favours flexibility in how you achieve a result (programmable VUs), while the X-Box has a large array of nice rendering features & effects available from the word go.
Also Im pretty sure the way the scene is rendered on the PS2 and X-Box differs somewhat, graphics on the X-Box is solely down to the GPU where the world is constructed and rendered. The PS2 has this split over two components (the EE & GS) - the EE handles the geometery before the information is piped to the GS where it draws the scene.
Of course I might be horribly wrong, if thats the case, please put me right.
-bd