A nodal multigroup neutron diffusion method for modern computer architectures has been developed and implemented in the ILLICO code. Vectorization and parallelization strategies that are successful in speeding up modern nodal computations on commercially available supercomputers have been identified and applied. Realistic three-dimensional benchmark problems can be solved in the vectorized mode in <0.73 s (33.86 Mflops). Vector-concurrent implementations are shown to yield speedups as high as 9.19 on eight processors. These results demonstrate that modern nodal methods, such as ILLICO, can preserve essentially all of their speed advantages (demonstrated on scalar computers) over finite difference methods. Several ways of treating two-dimensional reactor problems with nonsquare (“jagged”) boundaries as rectangular domain problems are presented and their effectiveness evaluated. They result in nonnegligible performance improvements and can be devised so as to preserve the physics of the initial problem.