Původně odeslal GaborSch
The analysis of some ISO 1600 raw files shows, that the noise of the 50D is a *tiny bit* higher than that of the 40D. The difference is really irrelevant; perhaps analyzing more images would reveal, that they are equal, or even that the 50D is a tiny bit better, but there can be no talk about the expected big advantage of the 50D over the 40D. Of course the smaller sensel size of the 50D makes it "relatively better", but that does not help on the dynamic range.
The last real ISO is 3200 (vs 1600 with the 40D). However, I don't see any true gain in 3200 compared to 1600, it is just like 1600 on the 40D.
In other words, ISO 3200, 6400 and 12800 are not different from adjusting the intensity in raw conversion (6400 and 12800 are numerical derivatives anyway).
However, the ISO 6400 and 12800 files show something new among the Canon raw files: it looks like if some noise reduction had been carried out on the raw data; but I don't find the effect on the noise.