The benchmarking script used for testing Chaos Theory can be found here at the bottom of the linked page. The Lighthouse demo included with the game was used for all benchmark tests, which allows readers to perform comparison testing if they feel so inclined. The batch files were edited to configure the various tests, and all scores were obtained from the .xls file that the game’s timedemo code creates and writes to once each demo is completed.
The first benchmark tests were used with the script configured to have the additional features (HDR, tone and parallax mapping, and high quality soft shadows) disabled in its batch file. The Radeon X800 XT was tested using the 1.1 and 2.0 shader profiles and the GeForce 6800 GT using the 1.1 and 3.0 profiles; the new 2.0 profile added by the 1.04 patch is, again, limited to ATI hardware only.
There’s essentially no performance difference between the 1.1 and 2.0 profiles with the X800 XT that would be discernible during actual gameplay, though the latter profile lagged a few frames behind at each resolution. This negligible performance hit is certainly worth the removal of the banding and blockiness of the specular lighting that’s present with the 1.1 profile. The 6800 GT’s performance variance is, however, the opposite of what the X800 displayed, with the higher shader profile resulting in faster benchmark scores; this is most likely the result of the 3.0 profile’s code path being able to render the game with fewer passes, though the difference lessens as the fill rate demands increase with resolution changes.
The next set of tests was conducted with HDR and tone mapping enabled, again using the 2.0 and 3.0 shader profiles. Anti-aliasing is disabled by both profiles once HDR is enabled. These two selectable features are not available under the 1.1 shader profile, so scores for the ATI and NVIDIA boards are from the 2.0 and 3.0 profiles, respectively.
Without knowing exactly what enabling HDR with the 2.0 profile is changing in the rendered output, it’s hard to reach any firm conclusions on the performance numbers these tests produced. Nevertheless, while the 2.0 profile saw a strong performance drop it’s not comparable to the dip in framerate that the 3.0 HDR option results in with the 6800 GT. Worth noting is that there’s a significant visual quality difference between the two higher shader profiles with this setting, as seen below in the screenshots section of the article. Whether or not the game is playable at 1600×1200 with these features enabled is up to the individual’s tolerance for low framerates.
The second set of tests was run with parallax mapping and high quality soft shadows enabled.
These two features don’t incur as steep a performance hit as HDR and tone mapping, though the 2.0 profile still sees the delta go from 10% to roughly 15% across the resolutions. The 3.0 profile sees a slightly worse performance drop at the lower resolutions, but this delta increases with the resolution until the difference is closer to 20%.
The final group of testing was done with all four additional features enabled.
As expected, these scores are the lowest among all the sets of benchmark tests, with 1600×1200 most likely being unusable for the majority of players. And as shown earlier, the 3.0 profile takes a much steeper frame rate drop with HDR enabled than the 2.0 profile.