You can accelerate the Algebraic Multigrid (AMG) solver inside Fluent by using General Purpose Graphics Processing Units (GPGPUs) in offload mode, provided that suitable hardware is available on your compute machines. When set up, you can use GPGPU acceleration for AMG computations in a Fluent session on linear systems with up to 5 coupled equations. Using GPGPUs requires HPC licenses. Licensing details can be found in HPC Licensing in the Ansys, Inc. Licensing Guide.
GPUs that are supported in the current release are posted on the Platform Support section of the Ansys Website.
When starting the parallel Fluent session the following conditions must be met:
The number of solver processes per machine must be equal on all machines.
The number solver processes per machine must be evenly divisible by the specified number of GPGPUs per machine.
GPGPU acceleration for the AMG solver is subject to the following limitations:
When using Fluent in Workbench, the number of GPGPUs cannot be set through:
Properties of Setup Cells
Properties of Solution Cells
Properties of Solution Cells through RSM update
Properties of Parameter Sets through RSM update
GPGPU acceleration of the AMG solver will not be used in the following cases:
The population balance model is active.
The Eulerian multiphase model is active.
The system has more than 5 coupled equations.
In order to use GPGPUs in offload mode to accelerate the AMG
solver, you must specify how many GPGPUs are to be used per machine by using the
Solver GPGPUs per Machine (Offload Mode) setting in the Fluent Launcher or
by using the -gpgpu= ngpgpus command line
option. For details, refer to the following sections:
| Starting Parallel Ansys Fluent Using Fluent Launcher |
| Starting Parallel Ansys Fluent on a Windows System Using Command Line Options |
| Starting Parallel Ansys Fluent on a Linux System |
Once the Fluent session is running, you can view and/or select the available GPGPUs on the system using the following TUI commands:
-
parallel/gpgpu/show Displays the available GPGPUs on the system.
-
parallel/gpgpu/select Selects the GPGPUs to use in offload mode. Note that you can only select up to the number of GPGPUs that you specified on the command line or in Fluent Launcher when starting the session.
By default, GPGPU acceleration is applied automatically to coupled systems and not to scalar systems because scalar systems typically are not as computationally expensive. However, if desired you can enable/disable GPGPU acceleration of the AMG solver for coupled and scalar systems by using the following text command menu:
solve/set/amg-options/amg-gpgpu-options/
The menu contains commands for each supported equation type, allowing you to enable/disable GPGPU acceleration, choose between AMG and FGMRES solvers, and specify various solver options.
Note: When using this text command menu, you must ensure that GPGPU acceleration is enabled for only coupled equations or only scalar equations, rather than a mix of both.