Nvidia and VMware announced software to allow customers to virtualize GPUs in order to speed up AI and other deep learning workloads.
The virtualization can happen either on a customer’s premise or in the VMware Cloud on AWS, the companies announced on Monday as part of the VMWorld conference in San Francisco.
They announced Nvidia’s Virtual Compute Server software or vComputeServer for VMware vSphere for enterprise data centers. It will be available from Cisco, Dell, HPE, Lenovo and other OEMs. The software will be licensed for each GPU.
The partnership means that customers of VMware on AWS will be able to access a highly scalable cloud service made up of Amazon EC2 bare metal instances that are accelerated by Nvidia T4 GPUs and Nvidia’s vComputeServer software.