A new optimization strategy based on system performance optimization life cycle (SPOLC) is introduced for high-performance lattice Boltzmann simulations of three-dimensional decaying isotropic turbulence. This strategy improves the performance of turbulent flow simulations in periodic boxes at different resolutions using the lattice Boltzmann method (LBM). The strategy improves the performance by modifying the lattice Boltzmann model, mathematical representation, computational algorithm, software implementation, and computing hardware utilization. The modifications include: (1) Establishing the slice concept as a logical grouping layer added to the LBM, applying an aggregation–disaggregation mechanism, enabling two-dimensional (2D) operation on the three-dimensional (3D) model, (2) improving lattice data access pattern by using an alternative one-dimensional (1D) array for numerical representation instead of the 3D cubic representation, (3) major reduction in memory access iterations by switching from function-wise iteration method to lattice-wise iteration method by applying code fusion to the streaming, velocity and collision model functions and iterations, (4) applying process parallelization and data vectorization, (5) achieving a much more efficient utilization of modern compute units by increasing the adaption of stream processing model. Furthermore, a correctness validation process has been applied by conducting lattice-wise value comparisons between the proposed solution output and the original implementation output. Simulations of decaying isotropic turbulence at resolutions ranging from [Formula: see text] to [Formula: see text] using the LBM are carried out for these purposes. Calculations are performed on two systems with distinct specifications, to validate the effectiveness and portability of the SPOLC strategy. The calculation times are significantly reduced after applying the SPOLC strategy on S1 with the lattice Boltzmann relaxation time [Formula: see text] by over [Formula: see text] compared to S2’s original time, increasing to [Formula: see text] at higher resolutions. Different features of the flow fields are depicted and their characteristics are discussed. Thin tubes are visualized, and the energy spectra are studied. All fields are initialized by a forced turbulent field simulated in a previous study using the LBM.