in LR you need to increase the exposure of the photo from the D3x with about 0.5 stops (to match the brightness).
I think you meant to increase the 0.5 stop to the D3S.
Right, the image is already underexposed with 1.3 stops, so this correction gives it only 0.8 stops underexposure. (Considering that these are both Nikons from the same range are really doubt that they have an ISO sensibility that differs with 0.8 stops; I simply believe that the photos were taken in a different amount of light.)
On my screen the D3s wins with a little though, but not much.
When I discussed quality above I was discussing "high end" prints and limits of human vision.
Right, but if the resolution of the print / display increases, the image with fewer pixels must be upscaled more, which means that its noise is increased plus it would show blurry details because it has no real details (they are interpolated).
This example was meant to show that the theory that less pixels means less noise is an illusion. Even if we ignore the technological advancements of D3s, is it really important that tiny difference in noise that you get from halving
the number of pixels (note: only in this illusory world where the better quality of the D3s is strictly given by the fewer pixels)?
I can see no practical application where fewer pixels would mean a technically better image
(= not better pixels). The only advantage there is, is in terms of more FPS and less storage space per image, but if the camera (with more pixels) implements pixel binning, these advantages are very limited (it really depends how fast the camera does pixel binning).
Of course, one could always go to http://www.canonrumors.com/forum/index.php/topic,255.msg3911.html#msg3911
where I point to a guy who did take the same photo in the same light with a 1D4 and a D3s (the photos have the same exposure settings and brightness).
Those images have to be normalized for the physical size of the sensor (and black level). Some people ignore that comparison because they do not understand that sensor size is THE technical reason why DSLRs and medium format (and whatever other uber-sized format) exist: more light for the same exposure AND noise. When such images are seen at the same physical
size as the ones coming from smaller sensors, that extra light actually clears the image of some noise.
(Alas, the guy's conclusion is wrong because he compares the noise at pixel level, not at image level.)