Of course the Canon 1Ds MKIII is a 21MP camera and the Nikon D3 is a 12MP camera. This made for a difficult if not unfair comparison in many respects. How to you frame the shot? Same image size or same resolution? I don't know. Both approaches are flawed.
In the end I said, screw it. I don't have all the answers. Let's just shoot a head and shoulder and see what we get. Whichever way I do it someone in the online world with greater tribal loyalties than common sense will get pissed off and claim foul, so I'll just do a test that satisfies my curiosity, and people can natter about it one way or the other if they wish.
Michael, I find it odd that you conducted the D3 vs 1Ds3 comparison in such a casual manner when a much more rigorous method of comparing across formats had already been established in Measuring Megabytes – The Great Luminous Landscape 2006 State-of-The-Art Shootout
The method was as follows:
1. Frame for same image size--as this is how people will use the camera. No matter what the resolution people should always try to fill the frame with the subject, this is just correct photography technique, no?
2. Resize all images to the same resolution for display--and this should not be the native resolution of either or any of the cameras being compared, because resizing is a lossy operation and no camera should gain an unfair advantage by not being required to go through this step. E.g. in the 1Ds3 vs D3 shootout, choose a resolution that is middle ground between the two cameras and size up the D3 image and size down the 1Ds3 image.
The article goes into further detail about RAW development, curves, etc. but on the matter of sharpening, I would add that if it came to sharpening 'to taste', I found it helpful to do one of the following:
-leave the noisiest image untouched and sharpen other images until they are as noisy, then see which image is the sharpest; or
-leave the sharpest-looking image untouched and sharpen other images until they are as sharp-looking, then see which image is most noisy.
(I find that it is easier to gauge subjective noise levels than sharpness, so I've mostly settled on the first method.)
Here's a comparison set done using the first method: