Interesting. With DX12 and other higher video demands the performance of the RX9 480 improves relative to that of the GTX 1060. I recently bought an RX 9 480 4gb version so that caught my eye. I suppose the greater bandwidth of the bus comes into play there.
Or just the DX12 API doing its thing and showing improvements over the arcitecture that works best with it. I doubt memory bandwidth has much to do with it at all, particularly 1080p and 2560x1440. In the first title, you should notice that the % difference doesn't get smaller between resolutions. That is one way to tell a difference in memory bandwidth.
Here is the breakdown...
If it was memory bandwidth, the higher resolutions should close the gap on the 1060. It didn't. You can see its the API, in that title, which did most of the work.
It's also interesting that there is very little difference between an i3 4330 and an i7 6700k in performance in that game in DX11. And it must be somewhat CPU intensive because the AMD CPUs fall way behind. It must depend a lot on single core performance, as I know a lot of games still do.
Single core, many cores... it's like 30% behind in IPC. So long as the game engine has the number of threads it needs, it's all up to ipc/clocks at that point. Techspot used to take this a step further and take and amd chip and intel chip to 2.5 ghz and up to 4.5 to see how the game responded.