GPU Driver

Before using a GPU-accelerated ECS, make sure that a GPU driver has been installed on the ECS for GPU acceleration.

GPU-accelerated ECSs support GRID and Tesla drivers.

To use graphics acceleration, such as OpenGL, DirectX, or Vulkan, install a GRID driver and separately purchase and configure a GRID license. The GRID driver with a vDWS license also supports CUDA for both computing and graphics acceleration.
- A graphics-accelerated (G series) ECS created using a public image has had a GRID driver of a specified version installed by default, but the GRID license must be purchased and configured separately. Before using such an ECS, check whether the desired driver has been installed on it and whether the version of the installed driver meets service requirements.
- To install a GRID driver on a GPU-accelerated ECS created using a private image, see Installing a GRID Driver on a GPU-accelerated ECS.
To use computing acceleration, install a Tesla driver.
- A computing-accelerated (P series) ECS created using a public image has had a Tesla driver of a specified version installed by default.
- To install a Tesla driver on a GPU-accelerated ECS created using a private image, see Installing a Tesla Driver and CUDA Toolkit on a GPU-accelerated ECS.

**Table 1** Acceleration supported by GPU drivers
Driver	License	CUDA	OpenGL	DirectX	Vulkan	Application Scenario	Description
GRID	Required	Supported	Supported	Supported	Supported	3D rendering, graphics workstation, and game acceleration	The GRID driver must be paid and requires a license to accelerate graphics and image applications.
Tesla	Not required	Supported	Not supported	Not supported	Not supported	Scientific computing, deep learning training, and inference	The Tesla driver is downloaded free of charge and usually used with NVIDIA CUDA SDKs to accelerate general computing applications.