Zoomblab: So many different factors motherboard, memory, audio, desktop, display server, compiler, file system ...
Wow, I haven't seen so much illiteracy in my life ever before. FPS has never depended on anything in your list, given that you use the same hardware platform.
- A good motherboard comes with a good chipset. In the old days the chipset had a memory controller influencing your framerate, nowadays this is only the handling of the BUS. As modern chipsets tend to all be good, the influence is indeed negligible.
- Before textures are uploaded to the GPU they are first loaded in system memory. Depending on the game faster memory and a faster file system may have an influence on the framerate, although this texture content is usually pre-loaded.
- On an nVidia card you have certain contexts. This can be compared to multitasking on a CPU: every task (context... well, sort of) wants some time to calculate, and the scheduler takes care that every task gets it's time. A desktop can (and will) do GPU calculations without being visible. Yes, the desktop env. does affect your framerate.
- The display server is more or less responsible for sending commands from userspace to the kernel module. A more optimal display server (userspace driver, OpenGL library) will send less commands with less CPU overhead. Yes, the display server affects the framerate too.
- A more optimized compiler will produce less CPU overhead in the kernel, the driver, the OpenGL library, the display server, the game itself.... Depending on the bottleneck of the system, this could or could not have an influence on the framerate.
Given the hardware is equal (not entirely clear from the article) there still are plenty of factors that influence the system performance, other than the kernel.
I understand these are out of the box experiences from each OS. In that regards the benchmarks are relevant in my opinion.
Each OS could be tailored further as mentioned in prior posts using lighter Window Managers and such. Although the overall effect is still present. I personally have never used desktop effects yet since they do not provide any enhance to my work flow. I always make sure to disable default enabled services that I would never use. I also use E16 or kde4 over any other desktops. I may see better performance overall with my style of setup but with all of the choices in Linux why skew the results of using non out of the box scenario's?
Honestly I would think most people that use Ubuntu are using Ubuntu's setup. If not they are probably using kubuntu or some other derivative. If out of the box experiences don't matter to you then the best thing you can do is to try you're own suggestions and bench each OS in a similar fashion on you're hardware to see if what you changed from default will really make that performance difference.
Users of both kernels are going to pretend to be mildly offended in an effort to appear smarter than you ;-)
All kidding aside, my concern is that sensationalist fluff pieces such as this will diminish the credibility of Phoronix's other worthwhile and noble efforts. Much like you've done with the "linux power management issue", your revenue-generating musings tend put-off those individuals with a ability to analyze your results and fix problems. PTS aside, if you strip off all the BS on the front page, Phoronix has become little more than a release-announcement-aggregator and the musings of a Mike McGee clone with an American accent.
Games on *nix:
Most *nix gamers play a single game (Eve or Wow for example). What matters to them is that their gaming experience is rock solid stable for long durations, and offers the full breadth of immersion that the creators of the game had intended. Crashing when alt-tab is pressed, losing mouse focus, and stuttering sound/graphics are examples of issues that break immersion and drive gamers away from these platforms. While many of us care little if our frame rate is 120, 80, or 60 fps, I can state with certainty that all gamers care quite a bit about those brief periods of zero FPS that send gamers flocking back to consoles and windows.
It would be nice if you gents put in some real effort and wrote an article about "playing games on Linux versus BSD", and actually took some time to play a couple games, join a guild, hop on mumble, and wrote an objective piece on the immersion breaking issues that you experienced. It will certainly generate better forum chatter than the last three pages of "Yeah-huh/Nu-huh", "You did it totally wrong", and "Ubuntu sucks".
You've got a driver with more than 95% of the same code (pretty much the same) delivering results that are pretty much the same (5-10%) with games using pretty much the same abi's. It may be that a couple of the key abi's used in openGL transaction for the nvidia driver really are a bit faster in the BSD's, or some user space programs are. Anyways a lot of differences and a result that could very well be insignificant (Ubuntu isn't the fastest distro around). The only really valid claim from these benchmarks is the nVidia makes a damn good driver for several different operating systems and has done so for quite some time.