you simply can't compare 580 or 480 to Vega. They aren't the same cards. This isn't one giant launch of various forms of Pascal, where you can accurately extrapolate performance across nVidia's entire release this last year, from card to card.
They are different architectures.
To be fair, he is asking about the process, not the architecture. Vega will presumably be built on the same process as the 580, and whether there are any benefits in that regard over the 480/470 is a good question. I'd say the answer is no. Of course, as you point out, that's not the whole story.
Still, I have nagging worres about its efficiency (and performance in general) and it's simply how different the promotion for Vega is vs. Ryzen. With Ryzen, you got a fair number of demos (handbrake, blender, BF1) from AMD, along with power consumption during those tests relative to the Intel. With Vega, we get some half-assed demos - we got about 10 seconds of random gameplay in Sniper Elite 4, (with an FPS counter, but no average or minimum FPS) a cool demo of HBCC (but tells us little about overall performance) and then some footage (at least in a stress zone) of Prey running on two Vegas, all of them without any hint towards power consumption.
W.R.T numbers AMD showed during tech week, there are some issues with those too. While it beats the Titan Xp in Specview, (or at least 3 out of 4 tests, one conspicuously missing) Quadro drivers give
a massive performance boost in those tests (yup, that's a gutted GK110 Quadro smashing a 1080), so Vega FE's position relative to Quadros is shaky.
Just a lot of things that don't bode well.