Dell PowerEdge R6525 EMC PowerEdge Servers with NVIDIA GPUs and VMware vSphere - Page 5

Use cases for NVIDIA GPUs in vSphere 1 Use cases for NVIDIA GPUs in vSphere The use cases discussed here are divided into VDI and non-VDI. NVIDIA GPUs are further classified into those used by multiple users and those operating in dedicated mode. This technical white paper briefs about configuring the NVIDIA GPUs with vSphere for various. The features available in vSphere with NVIDIA GPUs on the Dell EMC PowerEdge servers are: • vDGA (Virtual Dedicated Graphics Accelerator) • vGPU (Virtual GPU) • VM DirectPath-I/O GPU 1.1 VMware vDGA vDGA provides direct pass-through to a physical GPU. This method provides the following: • Unrestricted and dedicated access to the GPU. • Best performance to the user as the GPU device is dedicated to a single VM which accesses the GPU directly. • Limits the GPU usage to a single VM and prevents the use of Motion feature. In this method, the GPU device is passed through to the VM. The relevant GPU driver must be installed inside the VM guest operating system. No special drivers are required to be installed in ESXi. 1.2 NVIDIA grid vGPU NVIDIA vGPU allows sharing one or more NVIDIA GPUs to multiple VMs. This method provides the following: • Direct access to the physical GPU on the ESXi host across multiple VMs. • Options for multiple vGPU assignments to a single VM. • GPU-enabled VMs can be migrated to remote hosts with GPUs. In this method, GPU profiles are created based on the physical GPU and those profiles are mapped to the VMs. This method requires software components installed in both the ESXi host and the VM. The VM that has the GPU profiles attached requires a GPU driver. The ESXi host also requires that the vGPU manager software is installed. The vGPU option is used for VDI and virtual workstations. 5 Dell EMC PowerEdge Servers with NVIDIA GPUs and VMware vSphere | Technical white paper