Originally posted by Na Horuk 50MP sensor would bring along all sorts of problems. The SR would have to be fine tuned, the lenses would need to be improved (coatings, materials, will not be cheap and might necessitate bigger lenses), CPU would have to be faster, buffer bigger, cards would need to allow faster write speeds as well as more space.. even computers would struggle - how many of us would have to buy new hard drives, faster computers to deal with 50MP raw photos?
We can see that with the Nikon 36MP it already struggled with the previously mentioned things, but is still manageable (even if it requires steadier hands). If you look at K-3 files they already feature so much detail and resolution.. unless they are making massive prints that will be viewed from close up, or cropping a lot, i can't imagine who would need 50MP. Except maybe the marketing department, to show consumers of the ignorant variety that "we are winning the MP war!"
Those are actually all pretty straightforward issues that can more or less be solved by throwing money at the problem. Overall a little extra onboard RAM won't bump the camera cost too much overall. 50MP is "halo product" territory and a marginally higher unit cost is swallowed by the profit margin on a $4000 camera. And for computers, 16GB of DDR3 ram is like $150 - peanuts to someone who is buying a brand new $4000 camera. More RAM will make everything nicer anyway - having 16GB is great day-to-day, going back to machines with 4GB sucks.
Samsung is doing great work on the "faster CPUs" front as well - most image processors are more or less a "fixed pipeline" with a few knobs that can be twiddled in software, but Samsung's DRIMe-V is jumping to a really flexible architecture (sounds FPGA-like) that lets them get a lot more out of it. They've got it capable of dumping bursts of 28MP RAW images at 240 fps, so I think 8fps at 50MP is plausible for continuous shooting. What I suspect they're doing there is adding strong lossless compression to the RAWs, reducing the filesize they're writing, which directly translates into "faster effective card speed". You can get a size reduction factor of 2 with reasonable effort, maybe 3 if you really threw lots of processing power at it. If you implement it as an FPGA, it doesn't necessarily need to consume a lot of watts. They're doing a similar thing on video - H26
5 rather than H264 lets them squeeze more and larger video onto the card much faster. They can do UHS-3 class performance on a Class 10 card, relative to H264. And if Samsung finds some really cool new way to do things - they can drastically change the operation of the processor with firmware updates.
All of this takes a more powerful and flexible CPU. Most camera manufacturers don't really care about that crap and just dump a variation on the Fujitsu Milbeaut processor in their cameras because it's easy and cheap. This includes Canon, Nikon, and yes, Pentax - all just tweaked Milbeauts. That architecture can't really be reconfigured on the fly, if functionality (compression codecs, etc) isn't baked in then it can't be added after the fact. That's the tradeoff you make - the camera is marginally cheaper, but you need to start running dual cards, UHS class cards,etc to compensate for a crappy processor.You've also removed potential for major feature updates with firmware - not that anyone except Fuji seems interested in doing that if it might cut into sales of next year's model. So overall you'll be buying another camera sooner and spending more on cards/etc. Penny wise, pound foolish.
To me the bigger concern is diffraction - the megapixel count keeps climbing, which means smaller pixel pitch, which means a low diffraction limit. The unspoken problem here is that you need a lens that can deliver 50MP worth of resolution by f/5.6. Of course this is a function of pixel pitch, and the 24mp APS-C cameras ran into this problem first, this is just bringing the pixel pitch of a 24mp APS-C to a FF sensor. Long term this will continue to be a problem and the real long term future (10-15 year timeframe) lies in MF-sized sensors. There have been persistent rumors about Nikon and Canon prototypes with these large sensors. Pentax is actually ahead of the game here since it has the 645 lineup. I tend to view that as a long-term, strategic product line.
90s-vintage lens design won't cut it for much longer, but the lens problem isn't totally insurmountable. Modern computer design and liberal use of exotic materials and aspheric surfaces do make a big difference. I am astonished at what Sigma has been putting out lately - they are just spanking Canon and Nikon with their Art series. Better resolution at half the price (eg 35/1.4). Faster lenses than anything anyone else has put out (18-35 1.8 for APS-C, rumored 24-70 F2 for FF) and also super sharp wide open. Zeiss has realized that diffraction is going to be a key factor too. This is exactly what the Otus series are aimed at in a long-term sense - letting people get >50MP out of their >50MP sensors at wide apertures, before diffraction starts eating up the resolution.