Sorry guys but you are all wrong. It appears I have an old Mac Pro on which I have stuck an AMD 580.
Then I could compare using Handbrake and selecting same video with Videotool encoding checked RX580 versus Mac Mini running as well Handbrake: what does take around 55s with the Mini takes less than 30s with the RX580. As my Mac Pro has Xeons, Quick Sync is not available.
And this is also what the guys having an Imac Pro can find.
When I decide to select the external GPU for sharing, I must not detect Apple is overdriving my decision!! If Quick Sync is SO GOOD, then I would be smart enough to keep it! This is as simple as this: we must get what the option selection is supposed to provide.
But what does upset me, I do not know how but 1 day before I was rendering the same scene and by monitoring the two GPUs, they were equally used and the encoding time was 2mn09. Yes, I got the figure cause I did many times to confirm.
And yesterday without having changed any set up, the Iris Pro is 100% loaded and sharing time is now almost 6mns!!!
Values do speak and for me there is a HUGE bug.
Only thing I know, claiming eGPU is wonderfully supported by Apple Apps is just a joke.
And I repeat again: just type FCPX / eGPU / not supported and you will get tons of stories similar complaining a massive decrease in performances after making either macOS or FCPX changed. And all are using AMD’s GPU in their enclosures as well.
So reading above, Quick Sync does it better just let me laugh and wondering if some are just reporting marketing propaganda from Cupertino!!