Also a quote I'd like to excerpt:
This gives rise to a few thoughts ...
Clearly, the D3 has a chance to be a much better performer as far as shot noise is concerned. The sensor is not only bigger, but potentially more sensitive, as the lowest ISO is 200, compared to the 1dmk3's 50 (actually more like 70, probably, even though it is called 50). If this minimum of 200 is for good reasons, like high quantum efficiency and a high fill factor, then it will be collecting a larger percentage of photons incident on the sensor (a bad reason would be a tiny photosite fill factor with large microlenses to compensate, filling up the tiny wells quickly). The Canon 1 series are not stellar performers in the quantum efficiency department.
If the prints compared were from very well exposed images, with no significant shadow areas, then I could easily see the D3 outperforming the smaller, less sensitive Canon sensor, as shot noise is generally the dominant noise at higher tonal levels.
As you drop down into the shadow areas, however, read noise becomes more dominant, and this is where Canon has historically had a significant edge in DSLRs. This is the more mysterious part. Has Nikon reduced read noise to levels as low or lower than Canon? Currently, Nikon read noises range from about 15 to 60 (D40 and D2X) ADU at the pixel level at ISO 1600, while Canons range from 3.2 to 7.2 (1Dmk3 translated to 12 bits, and 400D). Canon's noisiest currently has a stop less read noise than Nikon's cleanest pixels.