The company offers its Grid server cards in two models: the K1 and K2. The Nvidia Grid K1 graphics board has four lower-powered Kepler-based GPUs and 16GB of DDR3 memory, which can support up to 32 virtual machines simultaneously (servers typically support two boards, so it would support a total of 64 users per server). The K1 is designed to host the maximum number of concurrent users on a server. The Nvidia Grid K2 graphics board, on the other hand, comes with two higher-end Kepler GPUs and 8GB of GDDR5 memory and supports up to eight VMs. The K2 is designed to deliver maximum density for users of graphic-intensive applications. Nvidia said its next-generation chips will allow for larger frame buffers, which effectively means its next-generation cards will probably have even more GDDR5 memory.
According to the announcement, Citrix has incorporated HDX GPU sharing and deep compression techniques into its latest edition of XenDesktop 7, which will enable users to take advantage of the hosted-shared form of desktop virtualization to deliver rich, graphics-intensive applications. When combined with Microsoft Windows Server RDSH, it allows for true GPU virtualization to enable the sharing of GPUs across multiple user sessions.
"With Nvidia Grid vGPU, even the most intensive graphics applications can be delivered by XenDesktop 7," said Bob Schultz, vice president and general manager of the Desktops and Applications Group at Citrix. "Now businesses can provide their users with the performance that they expect and need for engineering, design and video applications, while centrally securing and managing valuable intellectual property and sensitive information."
Nvidia and Citrix said the following types of businesses and verticals could benefit from this type of virtualized desktop graphics capability: architects and engineers using computer-aided design (CAD) tools; businesses using PLM tools for manufacturing applications; digital-content creators using video and photo editing software; and health care specialists using picture archiving and communication system (PACS) applications.
M7 Global Partners, a consortium of the top nine Citrix platinum-level IT providers in the United States, has endorsed the Nvidia Grid vGPU technology. M7, which has customers of all sizes across a wide range of industries around the world, plans to deploy Grid technology on servers from all the major vendors. Nvidia revealed that 175 customer trials with Nvidia Grid are already well underway.
As of this writing, the vGPU multisharing technology is limited to the Citrix XenServer hypervisor. We can only assume that this GPU virtualization technology will be coming to VMware vSphere and Microsoft Hyper-V at some point in the future.
This article, "Citrix and Nvidia accelerate desktop virtualization with virtual GPUs," was originally published at InfoWorld.com. Follow the latest developments in virtualization and cloud computing at InfoWorld.com.