Supported Hardware for CUDA Acceleration
Maple supports Compute Unified Device Architecture ("CUDA") acceleration on the following operating systems:
You will need to have an NVIDIA(R) CUDA-enabled card and the most recent NVIDIA drivers installed on your computer.
NVIDIA provides a list of CUDA-enabled cards at http://www.nvidia.com/object/cuda_learn_products.html. Any of these cards may be used; however, your specific hardware may impose certain limitations.
On Windows, you can install the standard NVIDIA drivers.
On Linux, you must install NVIDIA closed-source drivers. Most Linux distributions do not install these drivers by default.
On Mac, note that macOS 10.14 and higher do not support CUDA.
NVIDIA provides a rating, called the compute capability, to describe the capabilities of their hardware devices.
Hardware with a compute capability of 1.2 or less does not support double precision (float in Maple) computations. Hardware with a compute capability of 1.3 or higher supports double precision computations; however, in current hardware, double precision computations may be slower.
The amount of memory on the card may limit the size of the computation that can be accelerated.
Troubleshooting Display Driver Resets
Download Help Document