@Laserback
It depends on the bus speed of the RTG card more than the chip used in that card:*e.g. CV64 has faster bus than CV3D. On Quake I guess cpu crunching time and bus speed are more balanced. Anyway it also depends on the way the rendering is gone. If you write directly to RTG mem instead of doing a WritePixelArray() like some games do you get a nice boost.
@utri007
There's a huge difference between ZorroII and ZorroIII in terms of speed, open WarpOS Voxel demo (you don't need PPC IIRC) and try it on a ZorroII system and compare the performance with a ZorroIII/PCI one. In Quake the bottleneck is probably the cpu+ram but if you use a faster cpu you should get nice speed up using Zorro3, PCI & Permedia2 cards.
BTW, there was a patch that syncronized screen refresh with custom chip refresh that helped with gfx cards, it would be nice to test it with these type of games.