CNN Architecture

What is the difference between {applying activation function(ReLu) after Conv2D then MaxPooling}
and {applying Conv2D followed by MaxPooling then activation function }.

Conv2D → ReLu → MaxPooling

Conv2D → MaxPooling–> ReLu

thanks in advance

The result is the same. Based on a few test runs, applying relu before maxpooling seems to run slightly faster than applying relu after maxpooling.