Obviously the game was targeted at Xbox. The Game cube port was built as a feasability excercise, so that we could offer publishers two sku's instead of just one to make the product more attractive.
We were scheduled to finish work on the Xbox version around the end of August (it would have shipped for Xmas). Had we found a publisher soon enough we could have completed the GC version shortly afterwards.
In terms of cost, I'd rather not comment since I'm not certain about the figures that were thrown around.
In terms of what was and wasn't done, most of the car graphics were in game though not all cars had their performance characteristics set up. None of the upgrade paths for the cars were complete. The championship progress (similar in a lot of ways to WDC) was incomplete. Arcade mode and time attack worked, although no time had been put into balancing the races. There was a basic network play mode that supported 6 player Lan, and it should have scaled pretty well to internet play.
Graphically most of the major work was done, there were a few things we wanted to add, Headlight projections on the road, an additional diffuse shadow under the car, and reflective wet roads.
We would have reduced the complexity of the assets in order to keep the GC version running at 60 rather than dropping the frame rate to 30fps, the XBox datasets were too large to fit in the GC's memory anyway. Probably we'd have used our 12000 poly LOD cars and scaled back the backgrounds to make them fit.
There was some question as to whether the GC could run the full physics model for all the cars, although vectorising the ground collision code would probably have addressed the CPU performance issues that were present in the quick port.
If your looking for a performance comparison, as I stated above there is a large performance disparity between the two graphics chips, I'd rather not start a flame war by putting hard numbers on it.
Again as I've mentioned here before, it's debatable how much of a visual difference that disparity will give you.
