Oh no, another shader model debate!!! SM2.0b in the X800 is simply a SM3.0 without dynamic branching, both supports 512 shader instructions on each component doing a total of 1,536 max. SM2.0b even supports the SM3.0 Geometry's Instancing!!! But the main difference is that SM2.0b supports only static branching, only 1,536 will be the maximum, not even techdemos like ATi's Ruby exceeds the 300 shader instruction count. SM3.0 uses dynamic branching which extends the shader count lenght to 65,536 but will incurr in a huge performance hit cause is hard to predict floating point data, fortunately in X1XXX hardware, is minimal compared to the GeForce 7 Series which will incur in a larger performance hit. A X1650PRO iis only 5MHz faster than the X1600PRO and they have only 4 pixel pipelines, with 12 pixel shaders. It looses badly against the X800 PRO which has 3 times more pixel pipelines and the same amount of pixel shaders, the X1650 is as fast as a X700PRO or X800SE, sometimes faster, sometimes slower. If you talk about the X1650XT then is a different story, is as fast as a X850XT PE, sometimes faster, sometimes slower. Overall, the X800 should be faster, the X1650PRO pixel pipelines deficit can be overcome by the fact that it has 12 pixel shaders, but the X800 has 12 pixel shaders too!!
Facebook
Twitter