The above tests are invalid. For whatever reason the D2x was exposing the scene at 1 stop less exposure than was the D50. As matrix metering was used on all three cameras and raw conversion was done with Adobe Camera Raw, the exposure differences are really quite puzzling and might help explain the bad rap that the D2x gets for high ISO noise. In addition, the D50 got 1/4 more exposure. Looking at all three files as tiffs after conversion in ACR (with everything turned off), the D50 had the nicest levels.

The purpose of this test was to explore the noise levels of the three cameras when using 3,400 Kelvin light - rather than studio flash or daylight. 

Click on the above images to expand. Then click on lower right of expanded image to expand again to full size All three images were shot at the same time from the same location. A Lowell studio light with 40"x40"' soft box at 3400 Kelvin provided the sole light source. All three cameras used the Nikon f2.8 17-35mm lens set at aperture of f4. A tripod was used.

In order to be able to directly compare the D2x (with its 12 megs) against the D50 and D70 (with 6 megs) I did have to do some re-sampling in PS CS2.  After cropping the same identical section I re-sampled all three cropped images to a 4x5 inch 300 dpi print size (as you might before preparing a file for publication). Granted the cropping was rather severe, but it was necessary in order to see the noise. It is interesting to note that noise from all three cameras simply was not noticeable when printed full image on 8.5" x 11" paper.

Before anyone gets upset with any apparent bias I might have, let me say I use the D2x, my wife uses the D50 (loves the smaller size and weight), and my 45 year old son owns the D70. I might also add that the D2x was a hands down winner at ISO 400 and below. It also produced better color, better resolution, better speed, better auto white balance, etc, etc. Also note, that I did NOT do the test with jpg photos. My informal testing, however, would seem to indicate similar ranking.

My conclusions:

None. The tests are in error.