Benchmark Results: Games
We used Unreal Tournament 2004 (UT2K4), Doom 3, Prey, and Enemy Territory: Quake Wars (ET:QW) as our gaming benchmarks. While these titles may be old, they are some of the newest retail first person shooters to have native support in Linux. We set all four of our games to the highest detail settings with the maximum anti-aliasing (AA) and anisotropic filtering (AR) values. All games were run at a resolution of 1600x1200 with v-sync disabled.
In UT2K4, we recorded a two minute timedemo in the large, outdoor Tokara Forest level with 15 bots (16 total players).
For Doom 3 we used the timedemo included in the game.
The Prey timedemo (also two minutes) was recorded at a point early in the single-player game containing various enemies, scripted events, portals, and changes to gravity.
The timedemo for ET:QW was recorded in an instant action deathmatch in the Island level with 16 bots, also running for two minutes.
Due to the surprising UNiGiNE results, we re-ran the benchmarks on Lucid with desktop effects disabled. After an initial run to load the level, we performed five iterations and averaged the results to achieve our final FPS scores.
Under default settings, Lucid Lynx was able to pump out a few FPS more than Hardy in the newest game, ET:QW, and the oldest, UT2K4. The Hardy Heron, however, bit back by delivering a few FPS more than Lucid in the other titles.
When we disabled desktop effects on Lucid, the release's gain over Hardy in UT2K4 and ET:QW increased. But Lucid also overtook the elder LTS in Prey, and narrowed the gap in Doom 3 down to mere fractions of a frame per second.