The sad thing about this benchmark is that the next time Phoronix does a * vs. Linux benchmark (* being OSX, Windows, BSD, etc) Ubuntu will still be used as the reference Linux distribution, even though Fedora out performed it across the board. (Not only graphical performance due to X.org version - but also C-ray performance [page 4]).
Now undoubtedly there are better performing distributions out there that can give both Fedora and Ubuntu a run for their money (In my experience [and I'm an avid Fedora users], using XFCE instead of GShell 3.2 has a major effect on performance - at least on low end machines) - isn't it high time the reference Linux distribution will be elected based on, errr, performance instead of install base size?
Actually, Ubuntu might not have even the largest install base, a lot of Ubuntu users switched to Mint because "it's like the old Ubuntu, no Unity". Certainly distrowatch stats show Mint overtaking Ubuntu and also a lot of my Ubuntu using coleagues made the switch, though it's hard to estimate the overall user base.
20W on an EEE PC? I thought those things were meant to be quite efficient. Nouveau driver all round; does that W510 have switchable graphics so you can test with intel graphics too?
Yeah, that's pretty awful. I have an Atom N270 based netbook that idles at around 8-10W, but it's a mobile cpu with 2,5W TDP and one of the first generation Atoms. The Atom 330 is a desktop part with 8W TDP and the chipset and a separate nvidia gpu on top of that, so that probably explains the terrible near 20W idle power.
I boot my system every few days. The fact that it takes 10 seconds longer for Fedora is not significant. I also have UBUNTU.
When all is running for graphical user interface, the response times are not significantly different. I cannot type as fast as a screen can change. So this evaluation has little meaining to me.
What is significant is how each software performs when acting under heavy load. That is, run a background job doing some I/O and some compute, and now measure foreground speed for such things as large file copy, or a set of compilations. And while the background is running, to then measure foreground response.
My evaluation is that after stabilization, both Ubuntu and Fedora are essentially offering the same performance for the same software (libreoffice as an example, or firefox, etc.)