Large eddy simulation (LES) suffers from two primary sources of error: the numerical discretization scheme and the subgrid stress model (SGS). An attempt has been made to determine optimum combinations of SGS models and numerical schemes for use in performing practical LES for engineering-relevant problems. A formal quantification of numerical error present in finite-volume/finite-difference simulations was conducted. The effect of this error was explicitly added to a pseudospectral LES solver, and the modified pseudospectral solver was used to compute LES of decaying turbulence. In this way SGS modeling error and numerical error could be separately assessed. Verification of results was carried out using a commercially available finite-volume solver (FLUENT). Results showed that some combinations of SGS model and discretization scheme are more suitable for performing LES than others. Favorable combinations from the above findings were tested for an axisymmetric jet at Mach number 0.2. Results indicate good agreement with prior findings.