OK so not a million miles away then, though your cpu is probably a good 15% quicker if not more.
I'm sure if there is a problem with a-sync Nvidia will deal with it somehow...even if it means heavily optimising a driver for just this game
You can't optimise DX12 drivers in that way. they are supposed to be dumb and have minimal involvement. So Oxide disabled a-sync functions and let the rendering engine serealise it as with thier DX11 pathways for Nvidia hardware.
From the things said, it is the way the hardware architecture is setup. the Work dispatching units are not independent of a block of processing cores. so if one core needs to send work to another, it takes context switching to do this.
Where as with GCN, the work dispatching units are completely independent of the processing cores and each other. But they are also connected to each other thought L2 cache. so they can work independent of each other and send work to any free block of processing cores. While also directly communicating with each other.