" On March 27 Dxomark.com issued a press release that stated that the D800 matches the D4 (and ALMOST matches the D3s) in ISO performance. This is true - but only if you do what dxomark.com does - take that whopping 36 MP file and reduce it to 8.4 MB and produce a print (and compare it to a print from a D4 that has been reduced in resolution as well). The fact is that on the D800 you've collapsed over 4 pixels into one, and on the D4 you've collapsed two pixels into one, before looking at noise. Resolution reduction removes noise, so any camera with a sensor beyond 8.4 MP (dxomark.com's standard) and that is being tested using dxomark.com's published protocol has SOME noise reduction performed on it BEFORE they test it. And because higher resolution cameras are reduced in resolution MORE, they have had MORE noise reduction performed on them.
Scroll down to my entries of March 24 and 25 (below) for a more complete explanation and to see samples of what YOUR ISO 3200 D800 raw files will look like (and what D3s files look like) when you look at them at 100% (with NO noise reduction). Or, simply download this file (JPEG file: 2.5 MB) and view it at 100% on your computer. But hey - maybe my eyes are fooling me and at ISO 3200 the D800 and D3s samples in that file look identical. ;-)
I have absolutely no bone to pick with dxomark.com and quite like what they do. But in this case their test is producing a result (and they are really championing it - why?) that is misleading a lot of folks. The D800 DOES have astonishing ISO performance for a 36 MP camera, but it is no D3s or D4. In real world terms the D3s beats it in ISO performance by 1.3 to 1.5 stops. " - BRAD HILL