@markr041

What about cameras that use all the pixels of the sensor - that oversample? Do they debayer from the full sensor (say 24 megapixels) and then downrez to 4K after debayering, in which case they are true 4K?

It totally depends on the camera and its internal signal processing. Rule of thumb is that you get the best results if you debayer and downscale in post, because desktop software (like Resolve) can use much more computing-intensive, higher-quality algorithms than in-camera chips. Therefore the trend towards RAW video cameras. But a camera that has high-quality codecs and internal processing does the job as well - only that you then end up in league of the Canon C300, Sony FX9 etc., way above consumer and prosumer video and mirrorless cameras.

According to Yedlin's test (linked further above), you even need medium format sensors to achieve true 4K...

I found a solution that works. It is simple and obvious.

If you own an old quad core computer like sandy bridge or ivy bridge i5/i7 with ou without overclock, it will be better to avoid shooting 4k 30p. If you shoot 4k 24p the playback in timeline will work fine without stutter.

I did lots of cuts only / intercut with two different files and the playback worked fine with the native 4k 24p files. Less frames per second means less processor usage for decoding in playback.

Avoid using 150mbps or 200mbps in camera and use 100mbps, this also helps. Use H264 compression in camera and avoid H265. Use 8bit recording and avoid 10bit and LOG. Avoid 422 and use 420 color. Avoid apply LUT. Use real time effects and grading.

Set the program monitor window to 50% size and 1/2 (half) resolution.

Close all other softwares.

This way you do not need to transcode to cineform or to other intermediate codec and also do not need to use proxy files, you can use the native MP4 or MOV files for editing, and you will not need lots of expensive SSD storage.

Another tip is to set the GPU refresh rate to the same frame rate of your footage, if you are working with 30p set the GPU refresh rate to 30hz, if you are working with 24p set the GPU refresh rate to 24hz. In my tests this helps for a better cpu usage in playback and also avoid stutter (not only the stutter from different frame/refresh combination, but it helps to avoid the stutter from real drop frames)

If you work with FullHD 1080p you can shoot 30p no problem because the playback will be ok with native h264 files.

Considering i5-2500k overclocked to 4,4ghz with 17,6ghz total can edit native 4k 24p h264 100mbps files so the minimum system for 4k 30p will be: i5-9600kf intel processor with 6 core overclocked to 5ghz with a total 30ghz, 4x4gb ddr4-266=16gb ram, asus prime z390M plus motherboars. This upgrade is not so much expensive and can solve the 4k 30p editing. Two layers of 4k 24p or 30p need a GPU with 768 cuda cores and 2gb ram.

If you need to do chroma key with two layers or picture in picture or long wipe dissolve compositions you can keep working with 4k 24p and transcode to cineform 10bit and it will work fine in the old quad core computer using 50% size / half resolution preview.

@apefos why don't you just use proxies in your NLE?

Interesting article about: "4k or not 4k for filmmaking?"

https://www.borrowlenses.com/blog/do-you-need-4k-an-independent-filmmakers-experience/