I have seen authors running simulations in reduced units like T* = KT/epsilon, \rho* = \rho * sigma^3 etc .. Is there any added advantage by describing quantities in reduced units rather than using real units like Kelvin, g/cc etc.
Reducing units is similar to dimensionless analysis. Momentum, heat and mass transfer equations are commonly solved (analytically or numerically) in dimensionless forms. In numerical solution, the dimensionless form stabilizes the solution through the normalization of independent variables. As example, in a non-steady heat transfer differential equation expressed in hours solved by finite differences, a time finite difference of 0.01 hours may represent an enormous difference (with respect to solution convergence and stability) if the equations are expressed in seconds with time difference of 0.01 seconds. Then the equation is expressed dimensionless form (a typical dimensionless time is t.a/l^2 where t is time, a heat diffusivity and l characteristics dimension for conduction), and therefore the finite difference of time is in terms of dimensionless time which is normalized with respect to heat transfer phenomenon. A correct election of reduced units has similar effect of independent and dependent variables normalization, producing a more stable numerical scheme.