I only have one gripe with the methodology used in the testing of Cedega and Windows. It surely was a pitty that a key game such as UT2004 couldn't install either in Cedega or Vista, since it would have been very interesting to see how efficient was Cedega at converting DirectX code into OpenGL code and how would that reflect in the native clients of both platforms. Since the Linux version uses the OpenGL renderer, and the Windows version uses by default the DirectX renederer, it would have been in our best interest (for the purposes of Wine and Cedega to begin with) to contrast how much of a differece was there with native Direct3D code compared to translated code (Wine/Cedega) and native OpenGL code (native Linux).
Surely enough the version of Vista used was known (or I believe it was known) to have many problems with OpenGL applications (be it games or imaging applications), and that clearly shows throughout the results yielded by the tests ran in Vista. I wonder how a more recent build with updates applied would fair? (Not that I'd be willing to actually fork the cash for it, though).
Since the crop of games chosen for this test were all running OpenGL, it IS interesting to note the performance difference between native and translated (native Linux Vs. Wine/Cedega) leaning towards the translated code being "faster" (if only by a few FPS). It would definitely be interesting to see the same tests run with a broader variety of hardware (commodity hardware, and higher end hardware), say an array of Geforce 6600 / Radeon X700; Geforce 7600 / X1600; Geforce 8600 / Radeon HD2600 and their beefier cousins 8800 and 2900XT/3870, and probably (instead of the Quad or Octal core setup) a more common Dual Core system. Just a thought, though.