if you want to do impulse response function ordering of VAR is significant especially if you use cholesky decomposition. In structural Var model ordering might not be that significant.

If no restrictions are imposed, and you are using Cholesky decomposition, then the norm is to follow a causal ordering. The variables that come at front end of the VAR affect earlier variables simultaneously or with a lag, however, variables at the end of the VAR affect the following ones only with a lag. In summary, variables at the front end are weakly exogenous. After all, you need economic theory to justify your order.

p.s if you are doing SVAR with sign restrictions to identify the shocks in your system, then ordering have little effect if any on the results.