Note the following limitations for the GPU accelerator:
The GPU is used for the factorization of the matrix of the linear system, and is effective only for large problems (3D flows, Newtonian, generalized-Newtonian, or viscoelastic).
The GPU is not compatible with the classic direct solver.
Note: As of Ansys Polyflow 2021 R1, the classic direct solver is no longer supported.
Polyflow supports NVIDIA's CUDA-enabled Tesla and Quadro series workstation and server cards. The machine(s) being used for the simulation must contain at least one supported NVIDIA GPU card with CUDA cores. The following cards are supported:
nVIDIA Tesla P100 (both workstation and server)
nVIDIA Tesla K20c (both workstation and server)
nVIDIA Tesla K40c (both workstation and server)
nVIDIA Tesla K80 (server with GPU cooling solution only, 2 GPUs consume 2 ANSYS HPC licenses)
nVIDIA Quadro K5000, K5200, M6000
nVIDIA Quadro K6000 (both workstation and server, will be slow in graphic processing when used for GPU acceleration)
Note that the previous list is not exclusive; it is worthwhile to attempt using the GPU accelerator capability even on low-end cards.
The MUMPS solver can exploit GPUS. However, the performance gain has been found rather limited for standard problem sizes. Improvements are under investigation. Moreover, the use of GPU by MUMPS on cluster is not available.
For Windows, the driver version must be 333.11 or newer (the latest version of the driver of the specific GPU that equips the computer is recommended). For optimal performance on Windows, the TCC (Tesla Compute Cluster) driver mode is recommended when using Tesla series GPU cards. Some limitations exist when using this driver mode (for example, no support for Windows XP). Check your GPU card documentation for more details on how to set this driver mode and the existing limitations.
For Linux, the driver version must be 331.75 or newer (the latest version of the driver of the specific GPU that equips the computer is recommended).