Actually I think he got this one wrong, I believe he said Xbox pushed 12 million polygons @ 30fps, that means Xbox was pushing ~400k Polygons a frame while GCN's best was ~334k per frame.
I'm not surprised if true. I had all 4 systems that gen and it seemed pretty obvious to me that power wise they went Dreamcast<PS2<Gamecube<Xbox. I've heard the Dreamcast is more powerful than the PS2 but it died an early death and couldn't show off its power. I don't know how accurate that is.
Gah.
Ok, let's tackle this. Not every developer of that generation released polygon figures, hence there's no way to really know. PS2 did 10 million polygons per second in some games, not more; GC did quite a few 15 million polygon per second games, it was its thing (Metroid Prime series were regarded as 15 million, as was F-Zero GX); I've heard Xbox went as high as 15 million polygons per second at 30 fps on Rallisport Challenge.
We have no figures for RE4, but that one runs at 30 fps and has ganados with 5.000 polygons everywhere, plus detailed closed environments and Leon and Ashey accounting for 10.000 polygons each. That's huge for that generation standards; so on a per frame basis it should be using a lot. It's simply down to choice basically, they were doing 60 frames because they could, on the GC; on the Xbox though, not so much, and I'll explain; not only did Xbox have more hit texturing than GC, hence lower textured polygon throughput; it also had a pipeline limitation of textures per pass, sitting at 4; GC did 8 textures per pass and turns out everyone wanted to have that that generation; but none besides GC had it. So that meant Xbox in order to do it had to use the polygon trick, which is the same as saying you rendered twice, so it hampered your effective polygon throughput (and nice framerate) in half, yet it was done all the time.
PS2 also did multiple passages often and some times (it was actually designed for that, so hit wouldn't be huge in doing so) the announced polygon throughputs sometimes took that into account, meaning a game pushing 10.000 polygons on it might be doing as low as 5.000 with a mere two passages. Hence polygon pushers on both PS2 and Xbox avoided multiple passages (VF4 looks somewhat simple, despite being high polygon as hell), most games that gen saw more advantage in going that route though, hence how Halo 2 actually pushes less polygons per second than Halo 1 (which was a 10 million polygon game) and Chronicles of Riddick or Doom 3 on Xbox did it as well; hell, Splinter Cell games were doing it.
So no Xbox had the advantage at places, but didn't do so on polygon throughput or texturing capabilities (a byproduct of it's polygon texturing capabilities)
As for DC, DC was more feature rich and that meant it supported things PS2 couldn't do or had too much hit doing, it also supported texture compression, tile rendering and was a texturing beast, 480p for all games (something that had to be implemented on PS2 and most devs wouldn't bother with). PS2 was crapshit texturing, hence most games used 4-8 bit textures, which was a huge downgrade from DC days (and again, without texture compression lending a hand, hence why they were also being downgraded in color depth), then again, DC max throughput was 5 million polygons per second on Le Mans 24 Hours (and it simply couldn't go higher due to RAM limitations, not to mention the GPU probably couldn't handle it); as previously said, PS2 did 10 million polygons per second, so there's a palpable difference between both systems.
So no, DC is not more powerful than PS2 and would have hard time competing throughout that generation; but PS2 did some things spectacularly wrong in comparison to DC; hance it pretty much only won because it came two years later. (moore's law states performance should double every 18 months and that was certainly true and easily achievable back then)
