DxO Labs have recently launched their very interesting new website, the DxOMark.com, solely for ultimate measurebations on DSLRs (well, they say those are "measures" but I think "measur(e)bation" should be a better term in describing what they are doing! ;-)).
In this website, what you can see are only cold figures, but with no photos of any, except the images of those DSLRs under test! What I do like most are those cold figures are really cool, and most importantly, scientifically obtained and meaningful.
I bet this site will be the most popular measurebators' paradise after the old classic (but outdated) lens measurebation site Photodo.com (the old site).
Putting aside the new aggregate sensor performance mark they have created, the most important thing they have created and published is the comprehensive and up-to-date "Image Quality Database". In this database, you will see all the crucial performance figures, which are presented in a clear and professional manner but yet these are easy-to-read and also come with a concise and easy-to-understand explanation on each part for what have been done and what the figures represent. The "Compare cameras" facilitiy is really cool and it is very handy for comparing different cameras just with a few mouse clicks.
For example, they measure the Signal-to-Noise Ratio (SNR) for each available ISO speed instead of just measuring the noise level (as what DPR have been doing for years which IMO is just almost meaningless as I have said this for long). The SNR is measured on a 18% grey target which IMO is again very sensible. The DxO Labs do measure also the Dynamic Range for *each* ISO speed and so do the actual sensitivity of each ISO against as rated. They also measure the Tonal Range and Colour Depth which I must agree are two very important image quality performance parameters in reallife.
Besides, the ISO noise performance scale/mark is sensibly designed. They called this as the "Low Light ISO" in their front summary page for the camera performances and comparison. This mark is just the essential highest ISO number which the SNR has dropped to a predefined threshold where image quality has become unacceptable.
Here is the direct comparison between the Pentax K20D Vs Samsung GX-20 Vs Nikon D90. It would be quite interesting to note that the Samsung GX-20 actually has slightly better image quality in every critical IQ performance aspect and thus undoubtedly it has a slightly higher overall "sensor mark" over the K20D. However, as it can be seen the Nikon D90, which has the latest technology and enhancement, wins both the K20D and the GX-20 quite significantly (except the colour depth where very slight differences can be counted), no matter for the DR, SNR and Low Light ISO and etc. To be fair, the D90 wins also its big brother but predecessor D300 and thus it is clear that the D300 has no bright future and will be obsolete very soon as positioned by Nikon just in the recent half year, i.e., for APS-C DSLR, the hotcake is the D90 (which has the movie mode, too) and the true up-market advanced amateur model of Nikon is just the 135 Full Frame D700!
As for the sensor size, larger format sensor DSLRs always win smaller sensor format DSLRs, overall speaking. Just look how poor the results the 4/3 DSLRs are (Pana and Oly regardless, model regardless, latest or older) and then look back how good or better the results for 135 Full Frame DSLRs, I think there is nothing arguable, anymore.
Well, I think it's time for me to stop here and let all you folks to start measurebating and compare endlessly! Oh well again, welcome to the paradise of measurbation or JUST the Hell for Photographers (or even at the level below the Hell!)! Bye and see you later in this Hell, maybe forever!! ;-p (But please don't worry. Actually, there will be no problem - all those DxO professionals as well as those so-called DPR "technical writers" will be there with all you folks, and actually should be all of Us - which should including Me, of course!)