Most of you are aware of this feature on DPReview and probably understand it pretty well (I am still learning). Looking at the comparison in the attachment, the 6D uses a 85mm EF f/1.8 lens and the others use a 50 mm EF f/1.4 lens. Is this comparison fair seeing they don't use the same lense and f/?.? How do I interpret the results with these differences in mind?
There are two types of noise comparisons:
a) per-pixel noise, and
b) equivalent print size noise
Since we are now living in at 15MP+ world (in terms of most current DSLR bodies), I consider the equivalent print size noise comparison to be more important. A practical example: my old Nikon D40 captured photos at ISO1600 that were excellent when printed at A4 size, but if you opened up the photo at 100% pixel view, you would recoil in horror at the noise
Keep in mind that a 300 DPI print at A4 size only requires roughly 6MP. This implies that your 15MP+ photo will be scaled down to roughly 6MP before printing. When you scale down an image, you are effectively averaging the values of adjacent pixels (roughly speaking), which causes the noise to "cancel out" a bit.
The noise comparisons at DPreview are 100% pixel view. The apparent image quality at this magnification (100% pixel view) is relevant when you plan to crop out only a small part of your image, or if you are printing
huge posters --- around A2 size for the Canon 18MP sensor. If you find yourself cropping to 100% a lot, then you are probably not using the right lens.
Keep this in mind when looking at the DPreview high-ISO samples. Under these conditions, the full-frame bodies will have a significant advantage, because the effective light-gathering area of a 24MP body's pixel will be roughly 2.25 times that of a comparable 24MP crop-sensor's pixel (at 1.5x crop), which will cause the per-pixel noise to appear much lower in the full-frame sample (about half as much per-pixel noise). This advantage is real, but following my argument above, it applies to a specific case (100% view) that is unlikely to correspond with your normal use (say, printing up to A4 size, or on-screen viewing).
DPreview usually makes available the raw photos for download. I would recommend that you download a few high-ISO samples, process them with your usual software (say, Lightroom), and print them at your typical print size. This will give you a different (perhaps more realistic) perspective of the apparent noise in high-ISO samples.
One last thing: the aperture of the lens used in the DPreview samples does not affect noise in a meaningful way since they are shooting in a studio, which means they can adjust the lighting or shutter speed to obtain an equivalent exposure across full-frame and crop-sensor bodies.
