Why would you output 50MP at 16 x 3 = 48 bits per pixel after up-sizing from the Bayer pixel elements? And there is NO reason to output 16 x 4 = 64 bits per pixel as both G pixel elements are green and you just have one final G value per up-sized RGB pixel. Just output the friggin raw file @ 16 bits per pixel element to avoid bloating it up by a factor of 3x. Or output a lossless compressed raw at < 16 bits per pixel element. Also, who outputs the entire 3:2 screen for video (assuming you're talking about video)? They always chop it down to something wider, like 16:9 so a huge chunk of the pixels aren't stored (or even needing any processing at all) in the video stream anyway.
--
Canon outputs the full Y channel (i.e. Luminance) for Canny edge (kinda like SOBEL) detection for it's eye focus so that is 50 megabytes right there at 8 bits and 100 megabits at 16 bits. For the JPEG and HEIF compression they ALSO sub-sample RGBG down to YCC so that each channel can be done separately for the FULL 3:2 aspect ratio output.
AND since Canon can do FULL RAW (that is likely RLE or Run Length Compressed), you are looking at a MINIMUM of 50 megabytes of data for the RAW BAYER pattern at 8-bits or up to 100 megabytes for the 16-bit Raw Bayer pattern. The various Interframe video codecs MUST HAVE the full RGB or YCbCr frames in order to figure out what is an I-frame and what are Delta frames for long GOP (Group of Frames) video compression.
Even at 16:9 DCI 8K 8192 x 4320 pixels YCbCr at the 4:2:2 10-bit most people record at, you are STILL looking at a minimum of 44.2 megabytes PER FRAME for the full Y-channel sample. 11.1 Megabytes for the Cb channel sample and another 11.1 Megabytes for the Cr channel sample so for EACH FRAME during video compression you are looking at 66.4 megabytes uncompressed PER FRAME which then gets encoded down to the proper I-frame or Delta frame information.
Sooooooo....no matter WHAT HAPPENS... Canon and Sony are processing AT LEAST 66.4 megabytes per frame for STILLS OR VIDEO !!! That requires a FAAAAAAST ADC/DAC/DSP chipset egro which is WHY the Canon DIGIC and Sony BIONZ are 1.0 GHz to 1.5 GHx ARM Quadcore Cortex A5 cpu cores .....OR.... the later 8-core ARM chips!
How do I know this? Cuz I looked at the Canon BIOS microcode so I can mod it to my hearts delight. You should see my menu options. They look like something out of a horror novel with movie references and tiny pirate skull icons! I'll post a photo of my modded Canon 1Dc menu options later. I should note that our parent company frowns on me modding the newer 1Dx2's, 1Dx3's and R5's microcode!
V