GarbageManatee said:
Basically I would like someone to give me a hierarchy of ATi cards from best down to the 9800 pro.
This will be next to impossible to do as Radeon cards have been changing their architecture in ways to be better suited to specific conditions.
For example, a 9800 Pro will utterly SPANK an X1600 Pro in texture rich, high overdraw games.. but the X1600 Pro will meet or exceed the 9800 Pro in many shader-rich games such as Oblivion (assuming HDR off on both cases).
The reason being is- in previous generation products, the architecture would create GPU pipelines that would provide similar pixel shader pipelines as they had normal texturing pipelines. Newer Radeons are leaning away from this and putting a stronger emphasis on pixel shader performance than pure texturing/fillrate performance.
A a very simplified example, the 9800 Pro has 8 pixel/texel texturing pipelines and 8 pixel shader pipelines for executing shader code. The X1600 Pro only has 4 pixel/texel texturing pipelines yet 12 pixel shader pipelines. So, this in effect makes this architecture only yield half the per-clock ability to render textures, but a good 25-50% improved ability to process pixel shaders.
A card like the 9800 Pro with a heavily shader-rich scene will not reach it's fillrate potential because the pixel's it's rendering have shader-code (and bottlenecks) that need to be calculated prior to rasterizing the scene. Most all new games (DX9.0 and above) have heavy use of shaders so no matter what fillrate you have, you may still have low performance due to complex scene shaders to provide better looking visuals and more realistic surfaces in the games.
The X1900/X1800, for example- only has 16 pixel/texel rendering pipelines, but a whalloping 48 pixel shaders. Compare this to NVidia's line which can have 20 or 24 pixel/texel pipelines but also 20-24 pixel shaders. Older, fat texture games will see improvement on the Geforce 6 and 7 series cards, but shader rich games like Oblivion will see improvement on Radeon X1000 series cards.
When looking at upgrading your videocard, you should dig to see how many texture pipelines, pixel shaders, vertex shaders and memory interface (128 bit, 256 bit, 512 bit, etc.etc.). From here, you can leverage clock speeds, shader ops, texture render ops (ROPS), etc.etc. per clock. You'll see there is no "clear" winner, but instead price/performance brackets and products from both IHV's that can yield improvements in specific games (be they high-texture or complicated shaders).
Hope this helps!
