It was quite clear 4.7 was a dumbed down high efficiency model they put out in a rush to handle the capacity issues they were having at the time. I've experienced myself substantial degradation on basic reasoning tasks, which were fixed in 4.8.