Maybe if AMD did a 7nm RX570 successor, it would shine in performance and efficiency. But they pushed this architecture as far as they could and ended up with this junk.
Mind you, if 7nm is very expensive and makes these GPUs unprofitable, a "maxed-out" limited edition was the better choice than a sensible mid-range one...
I doubt that the node shrink will benefit the smaller chips more than larger ones. Usually upper mid-range to high-end chips achieve the highest performance per watt.
But still, the success of chips on the 7nm node depend on two primary factors; the node and the architecture. I would hope Navi have some architectural improvements/tweaks, but I'm not expecting anything radical.
This Radeon VII card is simply a stop-gap and a "backup plan" due to delays of Navi. Back in October we heard that AMD got engineering samples in their labs and they were looking "good", and then three months later at CES they didn't have a single fully working sample, and whipped up Radeon VII instead. That is a good indicator that something is seriously wrong with the "bigger" Navi chip. That might not even be the node at fault.
But let's mention CPUs once again. This card is such bad news for Zen 2.
There was already a leak that 8-core 7nm Ryzen will retain the ~100W TDP level. Everyone was like: "Naaah, AMD has already done this before. 8-cores on 7nm won't go past 60W and the future 12- and 16-core will also have 100W TDP". But looking at Radeon VII, I do believe in the 100W power draw. Which means 16-core Ryzen will pull 160W just like Threadripper. Not a big deal, but also not the miracle we've been promised.
I wouldn't read too much into that, since engineering samples can bee all over the place in terms of energy consumption. It can actually go both ways; do you remember the unrealistic Polaris engineering sample?
7nm will probably be good eventually, but until EUV arrives, it will continue to cause limitations. We can hope that Zen 2 is carefully crafted to deal with these production problems, but there will have to be some kind of trade-off between density, performance and yields until "7nm+" arrives.