Well I've been depressed because the specs of the Nikon 800 at 36mp is so much better than the Canon 5D Mark III at 22MP.
But I felt much better than reading the latest Popular Photography, although what it says disagrees with most of the internet sources. They claim the D800 certainly has has the resolution but nothing is free and it is much noisier even at Iso 800. The 5D3 has low noise up to 12800 but gives up quite a bit in resolution. So I don't have to necessarily sell all my lenses 🙂 but read on, it isn't so bad.
[Updated 4th July from Scott]
Well it turns out Scott's research disagrees a bit with Popular Photography. Basically says that noise is about the same for the two cameras up through ISO 800. Jpeg processing is too aggressive according to DPreview.com on the Canon so you see image degradation even at low ISOs. Hopefully they will fix that in firmware updates. But the bigger issue is that while you see more noise, when you print, it all becomes about the same because while there is more noise, these photos are way above the resolution of the average piece of paper.
The net is that Dpreview.com gave both cameras the same 82 point rating and after going through all the analysis in great detail I can see why. Basically, if you look at RAW performance and not JPEG (because JPEG is really more about what in camera noise reduction is applied), then you see that on a noise basis, the practice matches the theory which is that the Canon 5D Mark III has a lower noise curve all the way up through ridiculously high ISO 25,600. However, in the practical range from ISO 3200 down, they are neck and neck and in fact, the Nikon appears to have less of the most annoying noise (shadow noise).
Pretty incredible for a 36MP sensor. The more difficult news is that getting an actual photo which uses the extra 1/3 in pixels is very hard. We are close to the optical resolution of most lenses, so the DPreview folks found that even in the studio in controlled conditions, they had to only use the middle aperture and then use mirror lockup, a tripod with lots of weight on it, a fast shutter speed, etc., etc., to show the potential of 36MP. Put another way, this is really for the really dedicated studio photographers or Ansel Adams types with tripod and sandbags of weight using professional lenses in the best conditions.
The main advantages for the Mark III are pretty small. They pushed most of the video features up into the $15K Cinema EOS (ugh, lack of competition is painful) and the main benefit of a$3K purchase is in reality the much better raw noise levels, autofocus system and going from 4fps to 6fps for sport shots. Conclusion is that with these big full frames, 22MP is really good enough and you need a high quality L series professional lenses to see it. In conclusion, the main reason to upgrade from the 5D2 to the 5D3 is going to be for sports shots and for lower light photography where ISO 3200 really does look as good (as an aside, dpreview.com shows that the Fuji X-1 Pro has truly astoundingly low noise). The main drawbacks are that unlike the Sony's there is no full time focusing in video, so it is definitely not a no-brainer to upgrade.
Might make more sense, to keep the 5D2 as I know its limitations and it is only 50K shots through its 150K shutter life and focus on either a truly high quality compact camera (Fuji X-Pro 1, Nikon V1, Sony Nex or maybe the future Canon) or perhaps a dedicated raw shooting video camera that does a Cinema EOS at the right price (like the Blackmagic which uses Canon lenses 🙂
If I were buying a new prosumer level camera right now though, I'd probably say that the Nikon D800 is the right deal, it is cheaper than the Canon 5D Mark III and has the high resolution bragging rights, but the big deal is that it is \$500 cheaper. I do think in the end that web is right, with these two cameras, we are at the limits right now of still photography. Not bad when you consider, even the test conditions or dpreview.com don't allow them to test for more than 4000 lph and the D800 is theoretically at 4900 lph.