In general video editing is very CPU-dependent. This should be obvious -- that's why the Mac Pro is available with 12 cores and the upcoming iMac Pro with 18 cores. If it was "all GPU" they wouldn't need all those cores.
Yeah. Not true. I'm a professional colorist. I work in tv and film I have for almost 20 years. I work in a studio with 20+ edit bays and we pump out tv commercials and spots on a daily basis. Our technical expertise in this field is light years beyond these youtubers.
This Dave Dugdale is an amateur weddding video guy who probably clickbaits you to give him amazon discounts on gear. Looking through his website he is not an authority on anything but being a wedding DP.
First he's comparing apples and oranges when one compares premiere to davinci to Final Cut Pro x as it relates to rendering in CPU and GPU.
Resolve holds the entire image and any color nodes in its GPU to playback and then create a desired effect, where as Premiere uses the CPU then GPU to decode the frame and hold it in its memory for you to see, once you see it in Premiere the process is done, where Resolve is still holding all the images and nodes in memory. A different technology. Apple Final Cut Pro X uses proxy's and background rendering to achieve most of its speed. None of these technologies are 1 to1. A YouTube novice can say "hey option a exports in 1 minute and option b in 1.5 minutes, option a is better!" But it's not. It's two totally different technologies.
For example. When I grade feature films and I am using say a RAW 4K r3d file, this entire image is held in RAM, when I add nodes and effects, if I have enough GPU power it plays back in real or greater than real time. The CPU is only used in the same way a CPU is used to run a computer, or if you have a CPU dependent Source Codec. There are usually Consumer codecs like AVCHD or h264 variants or codecs with innerframe compression. Say you are using DPX files, their is barely any compression so your CPU is not a variation of your speed. In Premiere for example the CPU is not used to hold this image in memory. You decode the image, do the mathematical operations to see your results, and your final play back and/or render. A higher core or faster CPU will not make the GPU encoding process any faster. Since in Resolve this is all done in the GPU.
If you take my example of the feature films I grade and take that same r3d file add the same level of effects and color, you can not play that back in realtime, it has to be rendered, then if you change rendered again. When Premiere added the mercury playback using and your using CUDA, that is a similar technology as CUDA in resolve, but it's not program wide and doesn't work for all effects and operations inside the program.
A novice video editor like these youtubers doesn't work heavily enough in this program to have a valid opinion. He or she not using the program in a way a professional colorist is using it and they aren't even pushing it close to a level I would.
I usually have about 40+ color operations on complicated project and it plays back in realtime on the GPU. Put 40+ color operations an r3d file in timeline in premiere and you be waiting forever for the CPU to render this.
Also I work directly with Blackmagic a lot and the drivers for Pascal GPUs are not as fast on MacOS as the drivers for windows or even Linux. So if your comparing Pascal on MacOS it's not using its full power.
Also Final Cut Pro X is kinda a toy and it cheats. For you to do one to one comparisons with Davinci and Premiere you have to disable all the background rendering and proxy features. Rendering a 4K image that is actually a 2k of HD file, being held in proxy, is not a real comparison.
A simple process timing of a process in these tests is not a portrayal of these programs. A quick little YouTube snippet of an export or a process shows nothing. Ask real experts that uses these programs all day everyday to produce tv and commercial content under extreme pressure and real deadlines and you will know the truth.
To say a CPU is more dependent of GPU is just silly and not at all accurate. Also I don't know who the YouTube guy is or what is his angle but he obviously doesn't use any of the stuff professionally.
My take is don't trust YouTube click bait.