The one thing that still puzzles me is, how that unified memory thing works and actually handles contention between GPU and CPU without slowing either down. The

The one thing that still puzzles me is, how that unified memory thing works and actually handles contention between GPU and CPU without slowing either down. There's not much information out there how it actually works. Anybody has any details?