The performance improvements for AMD are mainly due to Global Foundries opening up a 14nm fab.
This is pretty incorrect. You don't get 52% IPC uplift just from a process node shrink. The Bulldozer family was a double whammy of bad for AMD because it was a bad design choice as well as them being stuck on an older, less power efficient node.
Had they released Zen chips on their old node sizes, they would have still realized the IPC gain, but would have had to work with lower clocks and higher power consumption. They're now competitive with Intel on performance/watt, that comes from the node shrink, but they're also competitive on performance/clock, which comes from the new architecture which doesn't have such boneheaded decisions baked in like a shared FPU between two Bulldozer "cores"
AMD hasnt really designed any bonehead chips. Ever. They just havent had access to parity FAB's.
Bulldozer was AMD's Netburst moment. It failed hard vs Intel on everything except specific multithreaded integer workloads, and even then only beat Intel at much higher power consumption. Every tech reviewer on the planet who knows what they're talking about has shouted it from the rooftops. The Core architecture pulled ahead of AMD's Phenoms and they never recovered till Zen.