- Documentation
GPU Compute Service
GPU Compute Service
User Guide
Installing NVIDIA drivers
Installing NVIDIA Tesla drivers
Quickly installing Tesla drivers in created instances
Installing NVIDIA Tesla drivers
Quickly installing Tesla drivers in created instances
Quickly installing Tesla drivers in created instances
To ensure that a GPU instance runs properly, you need to install the appropriate infrastructure software in advance. For NVIDIA Tesla series GPUs, the following software packages are needed:
- Hardware driver program: used to drive GPUs.
- Libraries: required by upper-layer applications.
For general-purpose computation scenarios, like deep learning and AI, or graphics acceleration scenarios like OpenGL, Direct3D, and cloud gaming, GPU instances can offer high-performance computing capabilities or display smoother graphics only with the Tesla driver installed.
Method 1: Use the cloud assistant to batch install drivers
List of recommended driver versions
You can use the cloud assistant to install NVIDIA GPU drivers and Fabric Manager. Supported GPU driver versions provided by the system are listed below. If you need a custom driver version, provide the download address of the corresponding driver installation package.
Usage notes
- After the GPU driver is installed, the system restarts the associated instance by default. Restarting the instance is recommended, since this ensures that the GPU driver takes effect. If you wish not to restart, specify the IS_REBOOT value as false.
- If you stop or restart the instance before the GPU driver is finished installing, driver installation will be interrupted. After startup, the instance will neither have the GPU driver installed nor resume the installation.
- If the GPU instance can not access the address of the driver installation package you provided via the intranet, this may incur public network fees. Download time will be affected by the public network bandwidth.
Procedure
tip
- The instance must be in the Running (Running) status.
- In the top menu bar, select the region where your business is located.
- In the left-side navigation pane, select "Maintenance & Monitoring > Cloud Assistant" and then select the "Public commands" tab.
- Find the target command for installing the Tesla GPU driver, and then click the "Create a job" button. Command-related parameters and related descriptions are listed below. For more parameters, refer to Creating jobs.
- After the job is created, the GPU driver and NVIDIA Fabric Manager are installed automatically.
Method 2: Select a public image with a pre-installed driver when you change the instance operating system
- Execute the nvidia-smi command to verify whether the driver is installed.
Last updated: 2025.12.03 16:28:05