Ok…I don’t really remember either very well…but it’s still the same size pixels and even if the algorithm is processing them faster…that wouldn’t make it better that I can see, just faster…and the statements made re eye AF are that DX sometimes locks on when FX is having trouble and processing the same AF sites faster…is that really going to result in getting eye AF when FX didn’t work? Dunno.
Thom Hogan said this in an article that I can no longer find as stuff has moved around the site a bit.
The thing I suspect you may be missing based on comments above is that while you said the AF system is looking at the same pixels for FX/DX, that's probably not the case if, as Thom has said and camera behavior seems to confirm, the AF runs off the EVF. It would mean the AF has fewer than half as many pixels to analyze when in DX mode.
Even if it's not a question of speed, what that means is that there are fewer things for the AF to get distracted or confused by in DX. We all know mirrorless AF love to grab busy backgrounds, for instance, but in DX mode it would mean much less busy background for it to accidentally grab onto. A rough analogy may be to say that in DX mode the "signal to noise ratio" of subject vs. other stuff is higher.
EDIT: I just realized what is potentially a major point of confusion in terms of all of this. When you said that it's looking at the same pixels, I was thinking of the frame DX vs FX, but you may have been thinking of the AF area box which has the same number of pixels either way. However, if I understand correctly I don't think the subject detection works this way. I think subject detection is operating on the entire frame at all times. One point that confuses some new users of the Z AF system is that the subject detection will focus on an eye
anywhere in the frame even if the AF box isn't on it. If you have the tiny 1x1 custom AF box on a cat's tail and the head is all the way at the other end of the frame, it might put the focus on the cat's eye even though the eye is nowhere near the box. it seems to work by detecting the subject anywhere in the frame and then checking for whether the AF box is on a part of that subject. Thus, in DX mode the subject detection algorithms have about half as many pixels to analyze as in FX no matter which AF area box one is using.