Virtual desktop users want to be able to run all the same applications that they ran on their PCs, but the user experience with high-end graphics is typically inferior -- at best.
NVIDIA, best known for delivering graphics for gaming, has set a course to improve virtual desktop user experience for all types of applications, whether from the cloud or with virtual desktop infrastructure (VDI).
Last year, the company delivered a platform that virtualizes the
Can virtual desktops keep pace?
In an ongoing SearchVirtualDesktop reader poll, over 76% of respondents said they believe that VDI adoption will increase as performance improves.
What NVIDIA will deliver next could come to light when the company discusses its roadmap for virtualized GPU technology during the Citrix Synergy 2013 conference later this month.
Whatever the company delivers will be pure gravy because the virtualized GPU last year was the final piece that virtual desktops needed, industry watchers said.
"You can do the full spectrum with virtual desktops now," said Gunnar Berger, an analyst with Gartner Inc., based in Stamford, Conn. "By bringing in the GPU, if people want to use [Windows] Aero, they can do that.
"You get everything, from the low-end user experience to the extreme high-end engineering workloads," he added.
That flexibility is increasingly important as more of today's apps require 3-D graphics. For instance, Windows 7's Aero Glass requires a GPU to operate. But pooled virtual desktops that don't use a GPU don't deliver those types of apps.
Limitations are a problem for end users who want to do everything on a virtual desktop that they can on their PC.
"If you mess up user experience, you mess up the whole project," Berger said.
You can do the full spectrum with virtual desktops now.
However, VGX isn't something that all VDI shops can add to their deployment, because companies' existing server hardware may not support it. It also costs more, Berger said.
"It isn’t something you would add to 2,000 desktops; you need to have a server that supports the VGX platform," he said.
For low-end uses, companies can instead use Intel Corp.'s Ivy Bridge chip, which has the GPU built in, Berger said. AMD also offers a server GPU geared toward remote desktops, called AMD FirePro.
What NVIDIA GRID and VCA offer
Dawnrunner, a San Francisco-based film production company, recently tested some cloud-based desktops and on-premises remote desktops. While they technically succeeded, the creative work the company does required much better performance than was possible with those approaches, according to James Fox, CEO of Dawnrunner.
"It was frustrating at times," Fox said. "We have people remoting in from all over ... Transferring terabytes of files across the cloud was too cumbersome."
The poor user-experience issues killed their cloud and VDI project at that time.
Then, they tested VCA, NVIDIA's soon-to-be-released GPU-based system that businesses can deploy in-house to run applications such as those from Adobe Systems Inc. and Autodesk. It sends graphics over the network to be displayed on a client computer, and the remote GPU acceleration gives users the same graphics experience they would get from a PC, according to NVIDIA.
Dawnrunner's team, which previously used powerful Dell Precision laptops with GPUs built in, found VCA to be just as good.
"I did side-by-side comparisons and they kept pace; the rendering was as fast as our principal workstations," Fox said.
NVIDIA VCA is offered in 8 GPU or 16 GPU configurations, with pricing starting at $24,900, plus an annual software license of $2,400.
The other new system, NVIDIA GRID, is a platform for large enterprises that is based on VGX software. It combines the remote capabilities of NVIDIA GRID GPUs, management libraries and NVIDIA GRID Kepler boards, which allow hardware virtualization of the GPU.
Dell, Hewlett-Packard and IBM all offer NVIDIA GRID-based servers that support GPU-accelerated virtual desktops and professional graphics applications that can be delivered remotely to any device. GRID pricing varies depending on the server specs.