Last year, our research group at the Wind Energy Division of Uppsala University was granted computing time on the Tetralith supercomputer to perform CFD simulations of the atmospheric boundary layer (ABL) and wind turbine wake flows. While our expertise has traditionally focused on Large Eddy Simulations (LES) using the Finite Volume Method (FVM), we have increasingly turned toward the Lattice Boltzmann Method (LBM), which offers a highly efficient and scalable alternative for large-scale flow simulations. Our initial allocation request focused exclusively on high-fidelity CPU-based simulations, and we have been very satisfied with the Tetralith service. Alongside CPU time, we also received several thousand hours on the Dardel-GPU system. While we are grateful for this allocation (which we did not specifically apply for) our GPU-LBM solver is compatible only with NVIDIA GPUs, preventing us from using the existing Dardel-GPU hours. We would therefore like to request an allocation on the newly available DARDEL-GH system. These resources will support several ongoing and future research efforts involving VirtualFluids (VF). Originally developed at TU Braunschweig, VF has recently drawn interest from multiple research groups, including ours at Uppsala University. In recent years, we have contributed to the implementation and validation of key atmospheric and wind energy modelling features (such as turbine, wall, and turbulence models) significantly improving the solver’s suitability for advanced wind-energy research reinforcing Uppsala University’s position at the forefront of the field. Examples include: • https://doi.org/10.1017/jfm.2023.390 • https://doi.org/10.5194/wes-5-623-2020 • https://doi.org/10.1063/5.0065701 The first planned LBM project, which will make use of Dardel-GH resources, is briefly described below. As wind energy development scales up, accurate ABL simulations are increasingly important for predicting wind speed, turbulence, and wake effects that influence turbine performance and project bankability. While LES remains the most accurate tool for such studies, its high computational cost limits its use outside academic research. LBM provides an attractive alternative due to its inherently parallel structure and excellent GPU performance, enabling high-resolution simulations at lower computational cost. Although prior studies have highlighted the potential of LBM-based LES in wind energy applications, a comprehensive comparison with established Navier–Stokes-based solvers is still lacking. This project aims to fill that gap by benchmarking an LBM-LES solver (VirtualFluids) against a well established Navier–Stokes-based GPU solver (AMR-Wind) for a small wind farm of four turbines and will consist of two parts: 1. Physical accuracy comparison: wind and turbulence profiles, energy spectra, integral scales, turbine power, and wake development. 2. Computational performance assessment: Node Updates Per Second (NUPS), runtime, and energy consumption. Although the project deadline (January 15) limits simulations to isothermal conditions, we plan to extend the study in early 2026 to include thermal stratification, enabling a more comprehensive evaluation. Future work will include large wind farm simulations and modelling of realistic atmospheric conditions.