Dell Servers Solution Resources : Concept Guide : Page 3

Dell, EMC and other trademarks are trademarks of Dell Inc. or its subsidiaries

NVIDIA GPU Virtualization software transforms a physical GPU installed on a server to create virtual

GPUs (vGPU) that can be shared across multiple virtual machines. The focus in this paper is on the

use of GPUs for compute workloads using vComputeServer profile introduced in GRID 9. We are

not looking at GPU usage for professional graphics or virtual desktop infrastructure (VDI) that will

leverage Quadro vDWS or GRID vPC and vAPP profiles. GRID vPC/vApps and Quadro vDWS are

client compute products for virtual graphics designed for knowledge workers and professional

graphics use. vComputeServer is targeted for compute-intensive server workloads, such as AI, deep

learning, and Data Science.

In an ESXi environment, the lower layers of the stack include the NVIDIA Virtual GPU Manager, that

is loaded as a VMware Installation Bundle (VIB) into the vSphere ESXi hypervisor. An additional

guest OS NVIDIA vGPU driver is installed within the guest operating system of your virtual machine.

Using the NVIDIA vGPU technology with vSphere provides options during creation of the VMs to

dedicate a full GPU device(s) to one virtual machine or to allow partial sharing of a GPU device by

more than one virtual machine.

IT admins will pick between the options depending on the application and user requirements:

• Partial GPUs: For AI dev environments a data scientist VM will not need the power of full

GPU

• GPU sharing: IT admins want GPUs to be share by more than one team of users

simultaneously

• High priority applications: dedicate a full GPU or multiple GPUs to one VM

Figure 2. GPU enabled VM instances using GPU Pass-

Though and GPU Virtualization (vGPU)