Originally posted by audiobomber Strictly speaking, your statement is correct. A raw file is unaffected by camera settings.
This is not entirely true.
For instance, if you opt for dark frame substraction as a camera setting (referred to as "long exposure NR" or similar) then this will irrevocably affect the RAW data.
I also suspect that other NR settings also affect the RAW data but I haven't verified this as I just set every in-camera processing to "off" so that I have all options in post-processing.
BTW, it is well-know that Pentax applies some non-optional NR to RAW files with the K-5 beyond ISO 1600. If they have changed the strength of this "smoothing" (as DxOMark calls it) for the K-3 then this could be misinterpreted as a noisier sensor, even though Pentax just did the right thing and eased off on the compulsary NR.
Noise should always be judged with a view to the captured detail as well. It is not that hard to deliver a noise free image with all details smeared. In other words, one should not hold it against the K-3 if it retained more noise in order to maintain details.
Even DxOMark does not look at resolution in the context of evaluating noise, so clearly a fair comparison isn't that simple.