If I know my system is ferromagnetic or antiferromagnetic. Now I want to do calculation with magnetic moment. So when should I do calculation with collinear or noncollinear tags.
The cost to do non collinear is biggest, so one uses known intuition
to choose cheaper calculation.
If the system is going to be ferromagnetic (known either through experiments, or by intuition of nature of exchanges) collinear is sufficient.
If the system is expected to be antiferromagnetic, local coordination geometry should also be noticed. Usually, if there is no competing exchange, and the magnetic lattice (the lattice system formed by just the atoms with magnetic moment) is bipartite, the antiferromagnetic state is also going to be collinear, such as Neel state. So in this case also a collinear calculation should suffice.
If however one expects non-bipartite lattice, or competing exchange, then one needs to explore non-collinear calculation.
For example, if the system has triangular lattice structure, and expected exchanges are only nearest neighbour, the ground state of such system would have moments at 120 degrees from their neighbours. Such state can't be accessed within collinear setup.