I've noticed that in most 3D applications, OpenGL or DirectX 3D acceleration works great with only a few items in the scene. But as you add more items, or high-polygons items, the hardware quickly loses all of it's advantage. When you run out of on board RAM on these cards, then the card gives up entirely and you get reverted to software rendering. Victoria with a pair of 3000x3000 texture maps, and you're already exceeding what most gaming cards can handle. Game 3D cards are totally optimized for getting high framerates in high-speed, low-polycount games. I doubt very much if any of the current nVidia or Radeon cards would help speed up rotation of a millennium figure scene much, even after triangulation.