Published on 14th Nov 2005, written by Dave Baumann for Consumer Graphics - Last updated: 29th Jul 2007
Splinter Cell: Chaos Theory
Splinter Cell: Chaos Theory still utilises the Unreal Engine, like its predecessors, however it has been significantly updated and alongside a Pixel Shader 1.1 path now also includes a Shader Model 3.0 path with additional options available with that. For more details of the rendering options and benchmarking using Splinter Cell Chaos Theory see our Benchmarking Guide here. In this test we'll be using the Shader Model 3.0 path, with the additional rendering that is available, as all boards are SM3.0 capable, and again the "normal" rendering will include the HDR path enabled.
Splinter Cell (FPS)
640x480
800x600
1024x768
1280x1024
1600x1200
7800 GTX 512MB
139.8
130.2
102.5
72.6
54.4
7800 GTX 256MB
131.2
113.0
85.4
59.4
44.2
6800 Ultra 256MB
96.4
76.1
56.6
38.6
28.5
7800 GTX 512MB % Faster Than:
640x480
800x600
1024x768
1280x1024
1600x1200
7800 GTX 256MB
6.5%
15.1%
20.1%
22.1%
22.9%
6800 Ultra 256MB
45.0%
71.0%
81.3%
87.8%
90.6%
With this Splinter Cell: Chaos Theory test the GTX 512MB board is again up to 23% faster than the 256MB board, and the average rendering performance is well above 60 FPS at 1280x1024, whereas it is just below 60 with the 256MB board.
Splinter Cell, 4x AA & 8x AF (FPS)
640x480
800x600
1024x768
1280x1024
1600x1200
7800 GTX 512MB
143.9
133.4
103.3
74.3
54.4
7800 GTX 256MB
135.5
114.4
85.5
60.6
44.0
6800 Ultra 256MB
99.1
78.7
57.8
40.5
29.5
7800 GTX 512MB % Faster Than:
640x480
800x600
1024x768
1280x1024
1600x1200
7800 GTX 256MB
6.2%
16.6%
20.8%
22.6%
23.6%
6800 Ultra 256MB
45.2%
69.4%
78.6%
83.6%
84.3%
With 4x FSAA and 8x AF enabled the performance differences between the two GTX's are again in a very similar range to what we've seen previously, with the gains for the GTX 512MB sticking closest to its fill/fragment rate differences in relation to the GTX 256MB board.
Note: Due to issues with HDR and FSAA in this title, NVIDIA's drivers disable control panel FSAA when this title is run, so 8xS performances cannot be shown.