1) Architecture. Understanding what the actual hardware is doing when you walk through a program. Knowing this can help answer questions like: Why does a variable go out of scope? When a data file might benefit from an index.

2) Algorithms. Learning the math behind why one approach is significantly faster can change the way you think about solving problems.

3) Approach. More advanced thinking around handling information. Why might you use a btree vs a linked list etc. Can I achieve the same result using less memory?