Sigh ... Sometimes I feel like I'm in 2002 all over again
A large portion of the Gamecube vs. XBox debate, as far as which produced better graphics, has centered on the differences between a fixed functionality GPU and a programmable GPU.
Back in the late 1990s most of the best GPUs had a fixed functionality pipeline that was augmented with pixel combiners. Many of these GPUs were able to produce many of the effects that were present in Gamecube/XBox games but game developers didn't take advantage of them because every GPU had a different set of features, and each manufacturer had different ways to take advantage of them.
The Geforce 3 was the first (IIRC) nVidia GPU that included programmable pixel and vertex shaders and Microsoft selected this GPU (and modified it) for the XBox, while Nintendo worked with a small GPU company that developed fixed functionality GPUs for flight simulators (ArtX).
The Gamecube was able to generate polygons that had 8 texture layers, each using a different texture effect from their built in pipline or using simple pixel combiners, and still render 10 to 12 million polygons per second. From what I remember, the XBox could only do 4 texture layers at a time but the programmable shaders allowed more complicated effects per layer.
If games were designed to take full advantage of the Gamecube's GPU they could not be replicated on the XBox because the XBox would require two texture passes and dramatically lower polygon throughput below what the GPU could do. At the same time, if games were designed to take full advantage of the XBox's GPU they could not be replicated by the Gamecube because the Gamecube's ability to produce advanced shaders was limited to the capabilities of pixel combiners.
