two things from the specs can make the difference
Hi-Z getting in the way of R200 http://forum.beyond3d.com/showthread.php?t=12535
or size of texture caches
RV250 and RV280 have a number of improvements over R200, including a doubled texture cache.
http://forum.beyond3d.com/showthread.php?t=2606
RV250 as 9000 Pro keeps up pretty well with the bigger R200 in most games and the same happens here. If Doom3 breaks hierarchical Z on R300, older chips are unlikely to be better off.
Another interesting ATI debilitating OpenGL experience is the KOTOR games, and their predecessor NWN which is on a similar engine. There is again NVIDIA development bias, but some ATI driver versions do run the games fine. If I remember right, the last D3D9 Radeon driver release doesn't work at all.
wrote:Truform. 😀
I think they're also the final ATI chip with error diffusion dithering...
How much support did Truform actually get, though? I suspect it probably got about as much support as the third texture unit in the original Radeons.
There's a Wikipedia article with a list of truform games.
If the diagrams of R200 press got upon release are true, then the texture cache is unified for all TMUs. In that case 2 kB would be unusually low value, trailing behind even Rage Pro. Could it be the cache was actually bigger, and that RV250 claim of bigger cache came from higher capacity per TMU and not increase of total size? Pumping up texture cache on a cost reduced part will less texturing power looks really bonkers to me.
Yeah the 2KB and 4KB references seem pretty tiny. I can't find any details. It's a shame that there was no documentation released for pre R300 chips.
Does anyone know if RV250 has better texture filtering than R200? I wonder if the pipeline overhaul affected it. R200 has some nasty limitations. R300 was a huge step forward.
