I've not dug into this enough yet but all prior generations of Nvidia GPUs have had caveats when adding steps to unify main memory. It certainly isn't "one addr

I've not dug into this enough yet but all prior generations of Nvidia GPUs have had caveats when adding steps to unify main memory. It certainly isn't "one address space" as far as the OS is concerned, perhaps Nvidia is saying this is sufficiently well enough abstracted in CUDA 8. The true land of milk and honey is prophesied in Volta when paired with Power 9 and CAPI, first appearing in the Summit and Sierra systems.