I think many people could have noticed that the IQ of the K20D is actually better than that of the K-7 with more image details and less noise. Here is an example for quick comparison:-
http://www.dpreview.com/reviews/pentaxk7/page29.asp
Do note the K-7 images above are with Medium Noise Reduction on whilst the K20D are Without. I don't agree with the author that the different noise levels of the comparison images. I think the noise reduced K-7 images are quite close to those of the non-NRed high ISO images from the K20D, which clearly contain more details (and of course). Maybe I would even say the K20D's test images even have slight edge for having a little less noise - go measurebate and verify yourself!
Back to 2006, I had written a technical article on the case study of an imager with 2 data output channels against a 4th one. Although it is about CCD imagers (Sony ICX483AQA and ICX493AQA, used in the Nikon D200 and the K10D respectively), the same technical issues and concerns apply in case of the older Samsung 14.6MP CMOS imager (two channels for data readout) against the new one that used in the K-7 (four channels). Just read my that article for more technical details and my remarks:-
http://ricehigh.blogspot.com/2006/12/secrets-of-k10d-part-1-of-3-heart-ccd.html
So, after all, the 2-channel or 4-channel approach is just a give-and-take. If the imager designer wants more speed, he would scarify the quality, and vice versa. The case of CMOS sensor could be even worse as amplification on pixel information is not by individual data "clocked-out" channel, but in more discrete form primitively with individual circuit component required per pixel, say, the signal (pre-)amplifier. On top of these, aggregated data capture/output channels and devices are still required. As such, more noise are resulted. Again, just remember: no electronic component and circuitry can be made equal. The unequalness and asymmetry or whatsoever is called just means inconsistency, which is the source of noise that finally appears in the recorded picture.
Last but not least, the image processors like the DIGICs and PRIMEs etc. can do the tricks by "Noise Reduction", but noise is eliminated altogether with image details, which are lost, as a result (which is yet again what I don't like about the Canon APS-C DSLRs like the 50D and the 7D etc.). So, again, Full Frame is the only way to go for 135 form factor DSLRs, including our K-mount one, if *both* low noise and higher image details are desired including the maintain of a comparable higher pixel count. Btw, I hope that the latest rumour of Pentax/Hoya has planned for a Full Frame DSLR is true and if it really comes by fall 2010, then it would be a real dream for all old film K-mount Pentaxians like me!