I wondered about this too. I would guess by rendering two pixels out of every three horizontally, instead of 1 out of two in traditional CB rendering. And then alternating which pixel acts as the hole (is not rendered) every vertical line.
Nvidia posts details on Denver processor
Here:
http://blogs.nvidia.com/blog/2014/08/11/tegra-k1-denver-64-bit-for-android/
It does dynamic translation in software like the Transmeta chips used to do.