Benchmarks - Pixel Shader (3DMark)

Here we'll take a look at the relative pixel shader performances using a variety of the 3DMark feature tests.

With the PS1.1 "Pixel Shader" test from 3DMark2001SE the "X" series of boards are all behaving about where we would expect them to be relative to one another, at least in high resolution where the X800 XT isn’t quite so system constrained. What we do notice though is the large gap the 9800 PRO has to the other boards – it’s even lagging behind the X600 XT, which it shouldn’t do by any theoretical measure. PS1.1 shader programs are very short in relation to PS2.0 or beyond, hence the performance of this test is likely to be quite dependant on the texturing requirements – as the "X" series of boards here have optimised texture filtering properties whilst the 9800 PRO doesn’t, what we may actually be seeing here is not performance differences due to shader processing capabilities, but in how the boards are handling the texture sampling.

Under 3DMark2001SE’s PS1.4 “Advanced Shader” test we see the performance trends are closer to the theoretical performances – however the X series correspond closer to their relative bandwidth differences than fill-rate (Pixel Shader) differences. The 9800 PRO is the one that bucks this trend as, although it has a higher bandwidth than both the X700’s it is scoring below them – again this could be due to the texture sampling differences between the boards with the X700’s using less bandwidth during texture operations.

3DMark03’s Pixel Shader 2.0 test is showing similar performance trends as the Advanced Shader test, at high resolutions, however here the 9800 PRO’s deficit to the two X700’s is a little greater.