|
Imo the frames per second benchmarks are not quite suited to give a good advice on future proofedness (at least not them alone). Because this proofedness also comes with the support of future calculation technologies that are to be implemented in future DirectX versions. For example the support of more advanced geometry shaders. Often these shaders are not tested sufficiently with todays games (and many graphic cards simply do not support them). Since todays games - for the sake of backward compatibility and market relevance - do not realy use the most advanced features of the latest graphics cards (utilizing them requires costly development) it is hard to predict future proofedness based on FPS tests on these games. So any FPS test actually just shows what the graphics card can perform using the older calculation methods.
It says next to nothing about the capabilites of future calculation needs in games.
One of the major future improvements will be in-GPU surface subdivision/tesselation and massive texture based vertex-displacement technologies. The calculation process will then be something like this:
1st) Use a low polygon base mesh to do the boned animation.
2nd) Tesselate the transformed low poly mesh using surface subdivision. Good algorithms will smooth out all the edges in this step. This step does add lots of additional vertices and thus makes the low poly base mesh a high polygon mesh.
3rd) Since this smoothing process will make the mesh look very undefined in certain places, the texture based vertex-displacement will give it the coarse contours back where it is needed (think of this like bump mapping but on vertex level). This vertex displacement is only usefull on very high polygon meshes.
4th) For the very fine details bump mapping and all those old fashioned image manipulation techniques will be used.
Today there is no game doing this. Most games rely on the 1st and 4th step but do not use the 2nd and 3rd. They do other stuff instead, that somewhat makes use of the calculation principles used in the 2nd and 3rd step. But not enough to actually give a good picture of the performance-relevancy in the future.
|