Earlier this year (2012) Nikon introduced their full frame D800 and Canon introduced their full frame EOS 5D MkIII, both aimed at the same market segment. In the EOS 5D MkIII Canon used a sensor developed from that of the 5D MkII. It has a few more pixels (22.3MP vs 21.1MP) and a more efficient design with gapless microlenses to capture more light. For the D800 Nikon used a sensor (reportedly a Sony design) with a much higher pixel count of 36.3MP.
The higher pixel count and correspondingly smaller pixel pitch of the Nikon D800 would be expected to result in higher resolution, and of course it does. The D800 sensor has a pixel pitch of 4.88 microns, while the Canon EOS 5D MkIII sensor has a pixel pitch of 6.25 microns. All else being equal (e.g. the strength of the low pass anti-aliasing filter), resolution is an linear inverse function of pixel pitch, so you'd expect the D800 to show resolution 28% higher than the EOS 5D MkIII.
In fact, after performing resolution tests using the D800 and 5D MkIII, that's pretty much what I saw. This is really no surprise.
The above images show the resolution under comparable conditions for the Nikon D800 and Canon EOS 5D MkIII. In each case the RAW (Nikon NEF) image was optimized and converted to a JPEG using the manufacturers supplied RAW conversion program. The 2.5 lp/mm line group in the Canon image is approximately matched by the 3.2 lp/mm line group in the Nikon D800 image. Theory says the D800 should resolve about 28% more than the 5D MkIII and 2.5 lp/mm x 1.28 = 3.2 lp/mm, just as expected.
Every so often you'll see on the web that a camera like the D800, which has a theoretical maximum resolution of around 102 lp/mm, can only deliver enhanced resolution with a few of the very best lenses shot at their best aperture and only with the camera and lens mounted on a tripod with the mirror locked up etc. While this isn't nonsense, it's a serious case of exaggeration. The EOS 7D has a sensor with a pixel pitch of 4.3 microns has a theoretical resolution of up to 116 lp/mm. Nobody has been complaining that they can't utilize the full resolution of the 7D sensor.
The other claim for the Nikon D800 sensor is that it has considerably higher dynamic range at low ISO than the EOS 5D MkIII sensor. Dynamic range is mostly a measure of shadow noise. Higher dynamic range translates into lower noise in the dark areas of the image. It can also indicate non-linear sensitivity in the highlights, but it's usually mostly related to the shadow noise.
Dxomark.com compared the dynamic range of the Nikon D800 and the Canon EOS 5D MkIII and found the following.
This seems to show that at lower ISO settings, the D800 has up to 2.5 EV more dynamic range than the EOS 5D MkIII. But what would this mean?
In a normal in-camera JPEG shot at ISO 100, you probably wouldn't see any difference between a D800 and an EOS 5D image because the normal dynamic range of a printed (or screen displayed) image is much less than the dynamic range of either the Nikon D800 or Canon EOS 5D MkIII. Where higher dynamic range would show up is in images which have been processed to bring up the level of the shadows. An extreme example of this would be a highly underexposed image. Another situation in which you might wantto bring up shadow detail would in in an extreme backlit subject.
To test this I shot images of my resolution test chart with from 1 to 9 stops underexposure. The examples shown here were shot at 5 stops less exposure than the meter indicated. In both cases the unprocessed resulting image looked pretty much like a blank black frame. However processing the RAW (NEF) image to reveal the shadow detail brought back the image in both cases as can be seen below:
It's pretty clear that the D800 image is a lot cleaner than the Canon EOS 5D MkIII image. Both images were processed in their respective camera manufacturers RAW/NEF conversion software. Similar results were obtained using 3rd party conversion software. No deliberate noise reduction was applied to either image. While perhaps some of the difference could be accounted for me not fully optimizing the image, it's still pretty clear that the D800 sensor output in the deep shadows at low ISO is considerably better than the output of the Canon EOS 5D MkIII sensor. This is also consistent with the DxoMark findings that the D800 sensor showed about 2.5 stops more dynamic range than the EOS 5D MkIII sensor at the ISO 100 settings. The tests I did pretty much agree with this, with the 5D MkIII shadow noise looking at leat 2 stops higher then the D800 shadow noise.
Though I haven't detailed the tests here, at higher ISO settings (ISO 400 and up) the dynamic range of the EOS 5D MkIII equals or exceeds that of the Nikon D800 and the 5D MkIII shows lower noise than the Nikon D800, especially at very high ISO settings of ISO 12800 and up.
I think it's pretty clear that the Nikon D800 sensor beats the Canon EOS 5D MkIII sensor, at least in terms of resolution and also in dynamic range at low ISO settings. The output of the D800 sensor at low ISO is remarkably clean.
What gives the D800 sensor higher DR at low ISO settings? That's a tough question to answer because Nikon aren't likely to reveal the details of the sensor design and construction. The trick is to get as much of the light falling on the sensor as possible to actually hit the photosensitive areas. This is done by using microlenses over the sensor by all sensor manufacturers (including Canon and Nikon). It's also desirable to make the photosensitive area of each pixel as large as possible by minimizing the chip "wiring" that overlays the photosites. It's widely assumed (and in some cases proven) that many recent Nikon DSLRs use sensors made by Sony. Their sensors (EXMOR) have the analog-digital converter integrated directly into the sensor chip, along with noise reduction circuitry. Whatever they are doing seems to work pretty well!
Does this mean that the D800 is a better camera than the Canon EOS 5D MkIII? Not necessarily. Judging a camera based only on the sensor resolution and low ISO dynamic range would be a bit like judging a car on only top speed and 0-60 acceleration times. There's a lot more to a car - and a camera - than that.
I can't whether the D800 is a "better" camera than the EOS 5D MkIII because that depends a lot on the user, so there's no "right" answer. Each have their strengths and weaknesses. I've shot with both and both did a fine job and produced images of the highest quality. I'll leave it up to others to "pick a winner". If I wanted a camera in this class, I'd probably go with the Canon EOS 5D MkIII because I have a lot of Canon lenses, I'm familiar with the Canon control interface and switching systems makes little sense. The EOS 5D MkIII also goes to higher ISO ratings (ISO 102400 vs ISO 25600) and shows lower noise than the D800 at high ISO settings I'm pretty sure the EOS 5D MkIII would do everything I needed it to do - and do it very well. The AF system is excellent and the camera is very fast, I was very pleased with the images I shot and the way the 5D MkIII handles during the time I had the camera to test.
If I had no current cameras or lenses I might be tempted to try the D800, not only for the higher resolution and DR, but for the built in flash and because it's around $500 cheaper. However for high ISO shooters, the EOS 5D MkIII might still be a better choice. The Nikon D800 will also AF to f8, while the EOS 5D MkIII will only AF to f5.6 (though the 5D MkIII has more AF zones and more cross type AF zones so overall AF may be better). Again I was pleased with the images I shot while I was testing the camera. The Nikon interface was a little alien to me, but not so alien that I couldn't figure out how to use it!
What the comparison does seem to suggest though is that Canon need to work on their sensor design if they want to be "king of the hill" when it comes to DSLRs. While the sensor used in the 5D MkIII is certainly adequate in terms of base ISO dynamic range, there is now serious competition from some Nikon (and Sony) DSLRs. To be state-of-the-art or to lead the pack It seems like Canon need to come up with something inovative rather than just make small changes to the current sensors. The Canon DSLR imaging sensors are good, but they no longer appear to outperform imaging sensors used by Nikon in their latest DSLRs.