NCP-AII Question 43
Select 4You are configuring a multi-node GPU cluster and need to install and configure NVIDIA NVLink™ Switch for high-speed interconnect between GPUs across different nodes. During the installation, which of the following steps are required to ensure the NVLink™ Switch is properly configured and operational?
- A
Ensure the GPUs being connected support NVLink™ and are compatible with the NVLink™ Switch.
- B
Install the NVIDIA NVSwitch driver and validate its compatibility with the installed CUDA version.
- C
Manually configure the NVLink™ Switch topology in the system BIOS settings.
- D
Enable Peer-to-Peer (P2P) communication between GPUs via the NVIDIA driver.
- E
Verify the NVLink™ Switch status using NVSMI (NVIDIA System Management Interface) or equivalent tools.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
To successfully install and configure NVIDIA NVLink™ Switch, it is essential to ensure GPU compatibility, install necessary software like the NVSwitch driver, and enable Peer-to-Peer communication. Additionally, verifying the status using tools like NVSMI ensures the setup is operational. Topology configuration is handled automatically by the system and does not require manual intervention through BIOS.
- A. Correct.
Correct: The GPUs must support NVLink™ and be compatible with the NVLink™ Switch for proper functionality. Compatibility is a critical pre-requisite for installation.
- B. Correct.
Correct: Installing the NVSwitch driver and ensuring compatibility with the CUDA version is necessary for the NVLink™ Switch to function correctly.
- C. Incorrect.
Incorrect: NVLink™ Switch topology is not configured manually through BIOS but rather automatically detected by the system and drivers once installed correctly.
- D. Correct.
Correct: Peer-to-Peer (P2P) communication must be enabled via the NVIDIA driver to fully utilize NVLink™ interconnect capabilities.
- E. Correct.
Correct: Verifying the NVLink™ Switch status using NVSMI or equivalent tools ensures the configuration is operational and functioning as expected.