BUT we are not talking about the noise on each pel, rather image quality scaled to a common resolution Let's say you have a 1024x1024 sensor and a 2048x2048 one (same chip size & similar tech). Now, imagine you're doing some sort of binning (before demosaicing) on the second chip, 2x2 pels for 1. The SNR (of the end result) will statistically be the same, the sensitivity will be the nearly same BUT the details will be better as you don't have to count in the bayer CFA interpolation error (and maybe an AA filter) on your CCD done to avoid moire and other color artifacts. With the binning, you almost have a true RGB pel and will thus lose less details in postprocessing. If you run down the numbers, yes, the image will be a little less noisy on the low res sensor, but with significant loss in details (which in turn make human perception less bothered by noise). Of course, if your camera/phone manufacturer does not use this information or requires it to be implemented in software, there is little benefit in the extra pixels, but that's not the question of SNR or pel count, but rather the manufacturer's decisions. And now we're oficially as off-topic as it gets