Scaling based on node size has been limited for a long time. It feels like it started at 13nm that every shrink was getting less and less performance uplift, but it's likely it's been going on for longer and we just had much more room for improvement on chip design.