Due to the 100% Steam CPU bug which Valve still hasn't fixed (See http://steamcommunity.com/app/221410...6056740854031/ ), any benchmarks done right now are not really accurate. I could get a 20+ FPS difference by messing up with Steam process niceness, and that doesn't even fix the issue, it just lowers it.
I would really like to see AMD windows vs linux benchmarks, especially on low-end hardware. I know people that say that they play TF2 (on windows) on hardware similar or lower than my AMD E-350, and the linux version with the very latest AMD drivers and with lowest graphics is unplayable.
I'm definitely interested to see how old of an integrated Intel GPU can play these games at 30 FPS, since there are tons of cheap chipsets that include them. Making my own little arcade box for Steam would be a fun weekend project.
If you're going to use Ubuntu, please make sure the latest updates (Undirect fullscreen windows) is enabled, or test in Gnobuntu, at least on some tests, please!
Canonical claim they have fixed the bug but yet enabling this option gets you 10% better performance!
Also, use a Performance governor instead of On Demand. Otherwise, SS3 cannot utilize the processor correctly.
Things are in beta, they supposed to be experimental, == not accurate, but ok.
Generally, if you use HD4000 you can play the game in Windows, but in Linux, because it is barely playable. I will have some benchmarks with my Macbook comparing: Windows 8, Mountain Lion and Ubuntu 12.04 using a timedemo for Team Fortress. Personally, I set minimum 40FPS as my limit for playability.
Generally, if you use HD4000 you can play the game in Windows, but in Linux, because it is barely playable.
That's funny, since my experience with new games so far (I have just recently got to play a new Unity game on Linux), fglrx + Radeon HD4890 works worse than a stock HD Graphics 2500. And this is with a rather old Mesa release, too.