Validating and Updating the Nvidia Cuda Toolkit for VMware Bitfusion Clients

Products

• VMware vSphere ESXi

Issue/Introduction

VMware Bitfusion clients need to have a current, support version of the Cuda driver/toolkit available to use remote GPUs provided by the Bitfusion server.

Resolution

From the standpoint of the application running on the Bitfusion client system, the GPUs are local hardware and use the same drivers and frameworks as a non-distributed GPU-based AI/ML solution.

VMware recommends that the Cuda driver/toolkit remains current on Bitfusion client systems. VMware Bitfusion is compatible with Cuda toolkit versions 9.2 through 10.2. For more information about installing the Cuda toolkit, refer to the Running TensorFlow on vSphere Bitfusion Example Guide, which covers Cuda toolkit installation.

Additional Information

In addition to the TensorFlow guide references above, installation information for the Nvidia Cuda Toolkit is available directly from Nvidia.

• Nvidia installation documentation for Cuda Toolkit 10.2
• Nvidia installation documentation for Cuda Toolkit 10.0
• Nvidia installation documentation for Cuda Toolkit 9.2

Note: This article links to external content provided by a partner. The links in this article are correct as of 2020-07-21.