First - forgive me the lousy formatting of this article. I'm really not good editing the HTML in blogposts and try to stick to the WYSIWYG editor of blogger.com most of the time.People keep asking about the resolution of the Fuji S5pro. Fuji sells it as a 12mpx camera, but its sensor only has 6 million so called "S-Pixels" and another 6 million so called "R-Pixels". The extra R-Pixels are ONLY there to increase the dynamic range (they kick in in the highlight areas and are less sensitive). But thats only the first speciality about Fuji's sensor (which is called "SuperCCD SR pro" in the S5pro). The second one is the layout of the "light buckets", and their shape. The pixels are not lined up vertically and horizontally, but diagonally, and they're not squares, but octagons.

Now, whats the benefit of that? The pixels are larger in this layout (less noise), and the total resolution that these 6 million pixels can resolve is actually higher. In order to really gain the extra resolution, the raw sensor data has to be blown up into a larger version to fit into the normal vertical and horizontal alignment of a "normal" image file (which of course doesn't consist of interleaved pixels that are aligned diagonally).
So what happens? The image data is blown up to 12 megapixels, and it appears soft and unsharp (a part of the softness is also due to the strong antialiasing filter in front of the sensor). Thats not an error or a mistake - 12mpx is above the resolution of the SuperCCD (this sensor does exist in Fuji's p&s cameras, too, just not in the "SR" version with the extra dynamic range).
So what is the resolution limit of the S5pro? The
folks that really know what they're talking about say that it is somewhere in the range of 8 to 9 megapixels.

I've prepared some example images to illustrate this with a real world example. I took a photo in the forest in the evening, from the tripod, using mirror-up and a cable release for the shutter to ensure maximum sharpness (I counted to three after the mirror release and then released the shutter, which roughly translates into something like 1.5 seconds of mirror prerelease - should be enough). I used the AF-D 1.8/50 Nikkor, which is the sharpest lens that I have. This is a 1 second exposure at aperture 8, the camera was set to ISO100. I'd say that the distance to the trees shown in the crop was somewhere in the range of 5 to 10 meters, I can't really say for sure anymore.
I stored the raw sensor data and imported it into Adobe Lightroom. I set ALL sharpening controls in Lightroom to ZERO and exported three versions of the image. Output sharpening was turned off, too. The first image (at the top) is a 400x400 crop of the full resolution 12mpx file. The second image is a same-sized crop, but this time from an ~8mpx export (3500 pixels on the long side). The third image is again a crop of the same size, but this time from a 6mpx export (3000 pixels on the long side).
Now, judge for yourself. For me, acceptable sharpness is reached in the 8mpx version. Just remember that
this is completely unsharpened raw sensor data when you compare the three images and that
you're seeing the actual 100% crop (400x400 pixels) right here in the post (unless your browser somehow scales the images). Adding a
little bit of sharpening makes the 8mpx version really "tack sharp".