The G1 is ok for resolution but falls short by quite a bit in the noise tests. Even at ISO 100 the noise is clearly visible. I like to shoot at ISO800 pretty much all the time and on my 450D it's possible. The 450D at ISO800 looks better than the G1 at ISO 100 in the dpreview tests.
Paul.
I retract what I said about the G1 being impressive in terms of noise performance. But looking again at the dpreview tests, I must say that I cannot draw any conclusion whatsoever. Unfortunately they don't show any RAW noise comparisons (except for a graph which is not very meaningful). Imaging Resource doesn't have any RAW comparisons either.
From looking at the JPEG comparisons I can see that the G1 does hardly any noise reduction in-camera. The noise in its JPEGs has a relatively natural look, rather than the clumpy look characteristic of NR. There was a time when I would have liked that... but these days, I always use RAW, so it doesn't matter to me how good a camera's JPEGs are (except for chimping).
The only really proper way to compare noise would be:
1. First compare sharpness at ISO 100. The softer a camera is, the more sharpening it will require, and sharpening increases noise.
2. Measure RAW sensitivity on all three color channels, in terms of quantum efficiency.
3. At each ISO, measure read noise (in electrons) and gain (in electrons per ADU).
4. Measure the amount of pattern noise in the bias frame (in modern cameras, there will be virtually no pattern, i.e. no banding)
5. Formulaically combine these numbers into four "noise performance indexes" for each ISO: indicating Chromatic Noise, Noise in Dark Areas, Banding in Dark Areas, and Noise in Midtones and Highlights.
This would give a truly objective means of measuring noise performance in digital cameras, meaningful to people who shoot in RAW, and yet AFAIK no review site does it.