| lilbroex said: There is a difference between numerical performance comparisons and graphical results. Everyone is factoring "out" the fact that it uses modern technologies that are simply impossible on the older hardware components that they are trying to retroactively compare it to. |
Benchmarks are good for very rough performance ranges, but don't necessarily suggest what you can achieve with the cards ...
As you mentioned Tessellation is an important factor. In theory, you could achieve similar visual results to most HD console games while rendering 25% to 50% as many polygons because a large portion of the polygons do not contribute significantly to the images produced. The reason for this is that the average 3D model that is being rendered is at a higher level of detail than can be displayed on the portion of the screen they represent.
I think the following is a good demonstration of what I'm talking about:
The close up characters are (probably) 10 times the size of the Banjo Player, who is (probably) almost 10 times the size of the far guy walking up the stairs. While many game engines have some form of Level of Detail system, few can actually render the nearest characters at a high enough detail or the middle and far characters at a low enough detail to really use detail efficiently without Tessellation.
With the next generation of graphics engines, I would expect hardware of the same (rough) performance level to produce graphics which were more detailed simply because there was less detail wasted where it couldn't be seen.
