Geometry Performance

Here we'll use a couple of tests to look at the Geometry performance of the S18 Nitro, first using GL_REME which looks at fixed function Transformation and Lighting.




GL_REME (Vertices/s) Pure transform 2 point lights 8 point lights 2 directional lights 8 directional lights
S18 Nitro 46,899,294 17,633,087 5,262,591 25,496,048 8,411,925
X600 PRO 35,109,940 19,148,557 8,173,049 31,347,200 24,221,972
 
S18 % Faster Than: Pure transform 2 point lights 8 point lights 2 directional lights 8 directional lights
X600 PRO 33.6% -7.9% -35.6% -18.7% -65.3%

Here we see that the Pure Transformation performance of the S18 Nitro is higher than that of the X600 PRO, but not to the degree that the theoretical performances would suggest. When lighting is applied the S18 Nitro looses out to the X600 PRO.

With Rightmark D3D we can compare the the fixed function performance to Vertex Shader performance.




D3D RightMark (FPS) Fixed Function T&L VS1.1 VS2.0 VS2.0 Static FC
S18 Nitro 22.5 12.4 11.3 10.2
X600 PRO 20.6 19.3 19.6 12.6
6600 46.1 17.1 16.7 10.4
6200 45.5 17.1 16.7 10.5
 
S18 % Faster Than: Fixed Function T&L VS1.1 VS2.0 VS2.0 Static FC
X600 PRO 9.2% -35.5% -42.5% -19.4%
6600 -51.3% -27.1% -32.4% -2.7%
6200 -50.7% -27.0% -32.5% -2.9%

With its four vertex shaders and higher clock speed the S18 Nitro is able to outperform the X600 PRO in fixed function Transformation and Lighting geometry processing, but is well behind both the GeForce 6's which may still have some hardware dedicated to this task. When the tasks switch over to Vertex Shader programs the S18 Nitro has the lowest performance of them all, despite it having the highest theoretical performance - whether this is down to the underlying vertex shader architecture, un-optimal driver shader compiler/optimisers or both is unknown.