Nvidia Unveils Virtualized GPU Supporting Multiple Users
Nvidia has announced the "first" virtualized graphics processing unit (GPU) for virtual desktop infrastructure (VDI) deployments in a datacenter.
The announcement was delivered on Tuesday by Jen-Hsun Huang (video), Nvidia's CEO, speaking at the company's GPU Technology Conference in San Jose, Calif. The new Nvidia VGX platform is notable for allowing multiple users to share a GPU on their virtual desktops. The technology could further enable bring-your-own-device scenarios in workplaces because power users with resource-intensive graphics workstations would be supported, along with office workers using productivity apps.
The technology is based on Nvidia's Kepler GPU technology. Huang said during the keynote that Nvidia had been working on this GPU virtualization solution for five years. The VGX platform enables every virtual machine to have a GPU, and Huang added that it should "accelerate VDI across the enterprise."
The platform has three aspects and can be added to typical server hardware in a datacenter. The first aspect is a "VGX Board," which provides support for multiple users via "four GPUs and 16 GB of memory" using a PCI Express interface. Next, there's the "VGX GPU Hypervisor," which is a software layer that integrates with various hypervisors to enable the GPU to be virtualized. Lastly, the "User Selectable Machines" aspect provides management control, so that organizations can set the degree of graphics support on virtual desktops, ranging from "standard" to "professional" for three-dimensional graphics support.
The VGX platform can support up to 100 users "from a single server powered by one VGX board," according to Nvidia's announcement. Forrester Research analyst David K. Johnson has seen the system demonstrated.
"I saw 100 unique, active graphics-intensive HVD [hosted virtual desktop] sessions running on a single 2u rack server," Johnson said, in a blog post. "That was a first for me."
Gartner analyst Gunnar Berger, also briefed on the Nvidia VGX platform, commented in a blog post that the technology could represent "a potential shift in the virtual desktop market." While he noted that putting a GPU into a virtual machine isn't new, what is new is the ability to share that GPU with multiple virtual machines. He said the technology could provide some capital-expenditure benefits to organizations delivering virtual desktops.
The Nvidia keynote address included Sumit Dhawan, group vice president and general manager at Citrix, for the announcement. Citrix was a key partner in the effort, but the Nvidia VGX platform likely will support various hypervisors, including those from Citrix, VMware and Microsoft.
Nvidia plans to make the VGX platform available for deployment sometime later this year. It will be available via Nvidia's hardware and VDI partners.
Kurt Mackie is online news editor for the 1105 Enterprise Computing Group.