Fluid Flow LiveGX Solver

The LiveGX solver is the default solver for fluid and fluid-thermal simulations in Refine. The LiveGX solver in Refine leverages GPU solver technology (currently available in Explore) combined with body-fitted meshing to provide both fast and accurate quantitative fluid results such as mass flow and pressure drop.

Depending on your system’s available GPU resources, you have the option to choose between the GPU and CPU solvers. Use the GPU solver for a quick solution. If the demands of a simulation exceed the graphics card resources available, select CPU instead to access the system RAM, which is generally larger than the GPU memory. The CPU solver allows you to run larger cases, but it will take more time. Solve time can be improved by utilizing more CPU cores (HPC license required) by changing the Max number of processors. Whereas, the GPU solver will be faster, but the case size will be limited by the size of the GPU. If using the GPU solver, an approximate guideline is that 1 GB of GPU memory is required for every 1 million elements in the model.

Shared memory fallback (implemented by NVIDIA for stable diffusion for some of the drivers) allows an application to use shared memory if GPU memory is exhausted. Applications that previously crashed when running out of memory can now continue to run, but at lower speeds. Simulation may run slower when solving a model that exceeds the GPU size.

The LiveGx solver is accessible under Simulation Options in the Simulation tab ribbon. Click Additional Fluid Flow Options (Refine) to see the options in the expanded panel. GPU and CPU are available with Use LiveGX solver.

Note: The LiveGX solver uses a different wall distance algorithm when run using CPU or GPU. This difference may result in slight variations in convergence behavior and results between a CPU and a GPU solution.

Along with the LiveGX solver, tetrahedral and polyhedral meshing options can be used to efficiently capture a wide range of physical features to both optimize GPU memory usage and accurately simulate fluid flow and heat transfer. Additionally, automatic boundary layer definition and local and global fidelity controls can aid in solution accuracy.

Note: Use LiveGX solver is switched on by default in the physics Settings panel.

Known Limitations

  • Results variable Vortices Lambda 2 is not supported.
  • The Spalart-Allmaras turbulence model is not supported.
  • There may be differences between detail monitors and standard monitors since the detail monitor value is a face-based result while the standard monitor value is a nodal based result.
  • Compressible flows:
    • Discovery provides compressible flow modeling capabilities for subsonic internal flows. However, there may be cases where the flow becomes transonic or supersonic due to a flow constriction in the fluid domain. External aerodynamics applications, supersonic and hypersonic flows, and compressible liquids should be avoided.
    • For simulations including multiple fluids, all gases in the simulation will be treated as compressible using variable density with the ideal gas law. All liquids in a simulation with multiple fluids are treated as incompressible, regardless of whether compressibility is turned on.
    • Buoyancy is not compatible with compressible flow. Turn off the Include buoyancy selection in the Gravity HUD to solve a compressible flow simulation.
    • Mass flow rate specification at a flow outlet is not supported in compressible flow simulations. You can specify pressure at the flow outlet.
    • Velocity specification at a flow inlet is not supported in compressible flow simulations. You can specify pressure at the flow inlet instead (or mass flow rate), or switch to incompressible flow.
    • Velocity specification at a flow outlet is not supported in compressible flow simulations. You can specify pressure at the flow outlet.
    • In Refine, if a domain with a compressible fluid is excluded from the simulation, after subsequent inclusion it is incompressible.