The Boltzmann-Fokker-Planck equation has been applied to treat charged-particle slowing down in solids. The discrete ordinates (SN) methods, with exact kernels (I*) or traditional truncated Legendre expansions (SNPL), have been used to investigate well-defined benchmark problems related to atomic displacement cascades. For an overall higher accuracy, it is found that an exact kernel transport calculation is equivalent, in terms of CPU cost, to a SNPN approach in one spatial dimension. Moreover, if the related cross-section processing methods are compared, it is shown that the calculation of the scattering kernels needed by the I* method requires only as much CPU time as the standard P0 matrix evaluation.