VMware has updated their white paper on 3D graphics acceleration, called Deploying Hardware-Accelerated Graphics with View Virtual Desktops in Horizon 6. It offers clear, concise descriptions of the most common use cases and scenarios.
Virtual Shared Graphics Acceleration (vSGA) provides great consolidation ratios because it enables a single Graphics Processing Unit (GPU) to serve a large number of people who have an occasional need for an immersive graphics user experience with advanced graphics applications.
At the other end of the user spectrum from vSGA, Virtual Dedicated Graphics Acceleration (vDGA) gives exclusive use of a GPU to designers, research scientists, and others who have traditionally worked on dedicated—and expensive—standalone graphics workstations. There are several advantages here: In contrast to the older computing model, vDGA puts the vast computing power and memory of the data center at the disposal of the individual user. Because the GPU is virtualized—it runs in the data center—the user can access it from “virtually” anywhere, given a suitable network connection and a client device. Configurations based on NVIDIA GRID GPUs support multiple GPUs per graphics card and multiple graphics cards per server, so that there are also some consolidation benefits here.
The optimal solution for advanced graphics acceleration is the Virtual Graphics Acceleration Unit (vGPU) model. With performance that can approach the level offered by vDGA, and a consolidation ratio (low cost per user) like that of vSGA, vGPU is proving to be an increasingly popular solution for enterprises that support large numbers of engineers, designers, and other users who require advanced graphics applications.
Of course, this paper also offers installation and configuration instructions and tips for troubleshooting. But wait, that is not all! Although the concepts are clear, the sizing and performance figures are based on the very powerful NVIDIA Kepler series graphics cards. We are happy to announce that we are working closely with NVIDIA to test View graphics on the even faster and more powerful NVIDIA Maxwell series cards, also known as GRID 2.0.