NVIDIA virtual GPU (vGPU) software enables multiple virtual machines (VMs) to have simultaneous, direct access to a single physical GPU, where applications operate directly with the NVIDIA driver stack and NVIDIA GPU hardware. This gives VMs unparalleled graphics performance application compatibility and the cost-effectiveness of sharing a GPU among multiple scaled workloads.
This chapter covers how NVIDIA vGPU solutions fundamentally alter the landscape of desktop virtualization and GPU accelerated servers. NVIDIA vGPU enables users to execute these solutions with various workloads of all levels of complexity and graphics requirements. This chapter also describes the NVIDIA vGPU architecture, the NVIDIA GPUs recommended for virtualization, the NVIDIA vGPU software licensed products for desktop virtualization, as well as key standards supported by NVIDIA virtual GPU technology.
The promise of desktop and data center virtualization is flexibility and manageability. Initially, desktop and data center virtualization was used, as flexibility and security were the primary drivers of cost considerations. The democratization of technology has reduced the total cost of desktop virtualization, thereby expanding market accessibility and driving growth with NVIDIA as a key facilitator. This, along with advances in storage and multi-core processors, make for a good and/or competitive advantage regarding the ownership cost.
The biggest challenge for desktop virtualization is providing a cost-effective yet rich user experience. There have been attempts to solve this problem with shared GPU technologies like vSGA that are cost-effective. Still, those technologies do not provide the rich application support needed to succeed and ensure end-user adoption. Dedicated GPU pass-through with vDGA, VMware’s terminology for passthrough graphics, provides 100% application compatibility but is cost-effective only for the highest end-use cases due to the high cost and limited density of virtual machines per host server.
Due to the lack of scalable, sharable, and cost-effective per-user GPUs that provide 100% application compatibility, providing a cost-effective rich user experience provided a challenge for broad use cases in desktop virtualization. Meanwhile, high-end 3D applications did not work in a virtualized environment or were so expensive to implement with vDGA that it was reserved for only the most limited circumstances.
Today this is no longer true. Thanks to NVIDIA vGPU technology combined with VMware Horizon, NVIDIA vGPU allows flexibility where multiple virtual desktops share a single physical GPU. This continuing innovation provides the 100% application compatibility of vDGA pass-through graphics. Still, the lower cost of multiple desktops sharing a single graphics card gives a rich yet more cost-effective user experience. With VMware Horizon, you can centralize, pool, and manage traditionally complex and expensive distributed workstations and desktops more easily. Now all your user groups can take advantage of the promise of virtualization.
The high-level architecture of an NVIDIA virtual GPU-enabled VDI environment is illustrated in figure below. Here, we have GPUs in the server, and the NVIDIA vGPU manager software (VIB) is installed on the host server. This software enables multiple VMs to share a single GPU, or if there are multiple GPUs in the server, they can be aggregated so that a single VM can access multiple GPUs. This GPU-enabled environment provides unprecedented performance and enables support for more users on a server. Work typically done by the CPU, such as GPU-ready workloads or graphics and encoder tasks can now be offloaded to the GPU. Physical NVIDIA GPUs can support multiple virtual GPUs (vGPUs) and be assigned directly to guest VMs under the control of NVIDIA’s Virtual GPU Manager running in a hypervisor.
Guest VMs use the NVIDIA GPUs in the same manner as a physical GPU passed through by the hypervisor. In the VM itself, vGPU drivers are installed, which support the available different license levels.
NVIDIA vGPUs are comparable to conventional GPUs in that they have a fixed amount of GPU memory, but they feature virtual displays instead of physical ports. Managed by the NVIDIA vGPU Manager installed in the hypervisor, the vGPU memory is allocated out of the physical GPU frame buffer when the vGPU is created. The vGPU retains exclusive use of that GPU Memory until it is destroyed.
These are virtual heads, meaning that there is no physical connection point for external physical displays on GPUs.
All vGPUs that reside on a physical GPU share GPU engines, including the graphics (3D) and video decode and encode engines. The right side of Figure 1.1 shows the vGPU internal architecture. The VM’s guest OS leverages direct access to the GPU for performance and fast critical paths. Noncritical performance management operations use a para-virtualization interface to the NVIDIA Virtual GPU Manager.
NVIDIA virtual GPU software is compatible with NVIDIA GPUs and supported on vSphere with a vSphere/ESXi vSphere Foundation License, formerly known as Enterprise Plus. Determine the GPU best suited for your environment dependent upon application use, optimization for performance or density, and professional visualization via GPU acceleration.
Please refer to NVIDIA GPUs for Virtualization for a complete list of recommended and supported GPUs. For a list of certified servers with NVIDIA GPUs, consult the NVIDIA vGPU Certified Servers page. Cross-reference the NVIDIA certified server list with the VMware HCL to find servers best suited for your NVIDIA vGPU and VMware vSphere environment. Each card requires auxiliary power cables connected to it (except NVIDIA T4). Most industry-standard servers require an enablement kit for the proper mounting of NVIDIA cards. Check with your server OEM of choice for more specific requirements.
NVIDIA virtual GPU software divides NVIDIA GPU resources so the GPU can be shared across multiple virtual machines running any application.
The portfolio of NVIDIA virtual GPU software products for desktop virtualization is as follows:
NVIDIA Virtual Applications (NVIDIA vApps)
NVIDIA Virtual PC (NVIDIA vPC)
NVIDIA RTX® Virtual Workstation (RTX vWS)
To run these software products, you need an NVIDIA GPU supported by vGPU software and a license that addresses your specific use case bundled with a vSphere/ESXi vSphere Foundation License.
NVIDIA vGPU software allows you to partition or fractionalize an NVIDIA data center GPU. These virtual GPU resources are then assigned to VMs in the hypervisor management console using vGPU profiles. Virtual GPU profiles determine the amount of GPU frame buffer allocated to your virtual machines (VMs). Selecting the correct vGPU profile will improve your total cost of ownership, scalability, stability, and performance of your VDI environment.
The NVIDIA vGPU software solution offers unmatched flexibility and performance when correctly paired with the proper vGPU software Licenses and NVIDIA GPU combination. These vGPU software solutions are designed to meet today’s modern enterprises’ ever-shifting workloads and organizational needs. Refer to the NVIDIA Virtual GPU Positioning Guide to select the best vGPU software license and GPU combination based on your workload.
Please refer to NVIDIA GPUs for Virtualization for a product overview of each NVIDIA GPU.
The NVIDIA Virtual GPU Software User Guide contains Information on Supported Graphics APIs and Support for OpenCL and NVIDIA® CUDA® applications.