If both perform similar, 970 winning some, RX-480 winning others, I suspect most people will go RX-480 as it is like a newborn baby, GCN is great for improving in performance over time.
AMD proved this, 390 struggled to keep up with 970, whereas now a 390 can comfortably challenge and beat a 970 as they gain well with driver updates due to GCN.
RX-480 will see similar gains, it also is more optimised for DX12 and of course its price point is aggressive and its other benefit is the low power and no requirement for PSU upgrade.
Our 390 pricing is extremely aggressive but they are nearly all gone, once 390/390X is cleared the RX-480 will be the clear go to card in the £200 bracket, with only the 970 competing at present. 1060 is too much an unknown, it should compete but the price could be £200 and it could also be £300, only time will tell.
There are a lot of new architecture features on the RX 480 that will only be utilised on newer games where as the 970, and the 1070/1080 for that matter have very few DX12 specific features to gain performance from in future games.
Fury X, even Hawaii has hardware that gives it an advantage in DX12 over Nvidia architectures, RX480 has another whole new set of features on top of that by the looks of it. A year, even 6 months from now, maybe even a couple games before then could utilise features RX480 has that no other cards do and push it significantly beyond a 980. Look at the few games where a 390x goes from competing with a 980 to competing with a 980ti in DX12. RX480 will likely gain even more in those specific titles and more DX12 titles are likely to show that level of performance gain over time.
Another key thing to think about, Pascal(outside of the GP100) offers extremely few new hardware features and nothing that seems to gain on DX12 performance compared to Maxwell. Effectively the entire die is being utilised today in current games, which is useful but worth keeping in mind. There was effectively in the past die space taken up by features on AMD cards that at launch were taking power but not actually helping performance, but 6 months, a year, 2 and 3 years later being used and improving performance. Effectively RX480 will be underutilised at launch where 1080 is effectively at it's peak. Either every new game that comes out, game devs will utilised new features on the RX480 and performance will improve but this won't happen on the 1080. We've seen it generation after generation with AMD. Their architectures are forward thinking and push game devs to use newer more advanced features.
I'm not particularly saying what Nvidia are doing is bad, just a different strategy... though I will say this, if both companies made architectures with no forward thinking architecture at all and purely for performance in current games, I think the drive forward towards new effects and features would be slow to non existent.
Without DX12 features being in GCN... we would have had no Mantle, no Vulkan, no DX12. AMD pushed for these because they needed the API industry to move forwards to better utilise their architectures. But on that note, Vulkan and DX12 being available already means there is already the API infrastructure there to enable devs to utilise these hardware features much much quicker on Polaris/Vega than with Tahiti/Hawaii/Tonga/Fury.