The interesting thing about the sensor in these things (Mitsubishi M64282FP) is that it has an analog output, so it's capable of much higher bit depth than in Nintendo's implementation of it. There are various other (mostly robot vision) projects using it that you can find on the Internet, and you'll see that the images are far less pixelated.
I also think it's rather amazing how much the tech has progressed - 15 years ago, the cost of a 128 x 123 (there's 5 black lines probably for BLC) - 16KP - monochrome array fixed-focus sensor module would be around the same as a 5MP AF colour module today, of the type used in smartphones.