Games Benchmarks - Splinter Cell: Chaos Theory (DirectX)
The Spliter Cell series of games have always been based on the Unreal Engine but also always sporting a more advanced renderer. The third instalment uses many modern features such as high dynamic range lighting, parallax bump mapping, attenuated soft shadows. Our tests we performed in PS 2.0 mode first without HDR and then with HDR with tone mapping and 16 tap anisotropic filtering. Antialiasing is not possible to use in conjunction with HDR.
| Splinter Cell: Chaos Theory (FPS) | 640x480 | 800x600 | 1024x768 | 1280x1024 |
| X850 XT PE | 103.5 | 96.7 | 79.3 | 57.2 |
| 9800 PRO | 53.1 | 43 | 32.4 | 22.9 |
| X850 XT PE % Difference to: | 640x480 | 800x600 | 1024x768 | 1280x1024 |
| 9800 PRO | 94.90% | 124.90% | 144.80% | 149.80% |
Splinter Cell: Chaos Theory is a modern game with modern features and it shows. The X850 XT PE starts at 640x480 with nearly double the framerate of the 9800 PRO and it never becomes CPU limited although we do see that the rate at which performance drops as we ramp up the resolution is higher than the Radeon 9800 PRO - in this case it could well be the difference in Vertex Shading capabilities is making for the large performance differential at low resolutions. While the X850 card remains playable in all the resolutions tested, the 9800 PRO stops being recommended for a playable experience at 1024x768. Splinter Cell is a slower paced game and the minimum recorded frame rate at that resolution in our timedemo run was a hair under 20FPS so the 9800 PRO can be considered only just barely playable here.
| Splinter Cell: Chaos Theory, 16x AF (FPS) + HDR | 640x480 | 800x600 | 1024x768 | 1280x1024 |
| X850 XT PE | 95.4 | 80.8 | 62.1 | 43.4 |
| 9800 PRO | 42.7 | 33.5 | 24.8 | 16.7 |
| X850 XT PE % Difference to: | 640x480 | 800x600 | 1024x768 | 1280x1024 |
| 9800 PRO | 123.40% | 141.20% | 150.40% | 159.10% |
HDR and 16xAF do take their toll and we can generalise that these two quality settings yield performance equal to the next-highest resolution without them. Like in the pure mode above, the rate of performance drop from resolution to resolution is steeper with the X850 than with the 9800 PRO. It's difficult to pin-point the reason but it could just be a consideration of using an average figure: at lighter loads the maximum recorded frame rate might get higher comparatively than at higher resolutions thereby inflating the figures represented here; since the 9800 is already so strapped for speed this is not as noticeable thus leading to more linear performance drops (if you note, the 9800 PRO seems to be consistently losing around 9 fps from a resolution rung to the next, while the X850's drops are consistently larger and larger).
The X850 remains playable on all resolutions with room to spare for this slow-paced game while the 9800 stops at 800x600.
