The Hardware Acceleration node () can be added under a Time-Dependent Solver node for hardware acceleration using a supported GPU for time-dependent simulations using the discontinuous Galerkin (dG) method.

A CUDA Toolkit must be available The path can be specified in the installer and changed on the Computing > GPU Acceleration page in the Preferences window, where you can also verify your CUDA installation.

Select the Accelerated solver for discontinuous Galerkin checkbox to use an optimized solver that can be run on either a CPU or a GPU. When enabled, the default is to run calculations on the GPU using single-precision floating-point arithmetic.

The Compute residual on GPU checkbox is selected by default. When enabled, calculations on the GPU can be run with either single or double floating-point arithmetic. You can control the precision using the Use single precision on GPU checkbox (cleared by default to use double precision). Clear the Compute residual on GPU checkbox to run the accelerated solver on the CPU. When disabled, all calculations are performed using double floating-point arithmetic.