As applications become increasingly graphics heavy, VDI administrators must know how to take advantage of virtualized GPU technology.
Virtual graphics processing units (vGPUs) are not a brand new concept; NVIDIA vGPU has been around since 2012. The technology renders complex graphics for virtual desktop users who work with graphics-intensive applications such as AutoCAD or Photoshop. With vGPU, all the graphics processing happens on a VM host server rather than on users' devices. The endpoints simply decode the delivery stream.
NVIDIA vGPU was born out of the changing definition of what an office is, according to Jeff Sporn, senior business development manager at NVIDIA. Users expect to be able to work from just about anywhere, and that means IT must be able to deliver even the most graphics-intensive apps to wherever workers need them. In a session here at this month's VMware User Group UserCon, attendees learned the ins and outs of the NVIDIA vGPU offerings, including the hardware and software components.
Get to know NVIDIA hardware options
NVIDIA has a slew of graphics processing cards -- the physical computer chips admins install in the server to render graphics -- including the Tesla M10, M6 and M60. The M10 focuses on user density and supports 128 desktops per server. The M6 is unique in that it's compatible with blade servers. And the M60 aims to maximize graphics performance.
Picking between the different graphics cards comes down to what resources a company's users need, Sporn said.
What are NVIDIA's software options?
From a software perspective, NVIDIA offers three options: NVIDIA GRID Virtual Applications, Virtual PC and Virtual Workstation. Virtual Applications is designed for organizations that want to deliver full-fledged Windows applications to users with Citrix XenApp or Remote Desktop Session Hosts. Virtual PC emphasizes application, browser and video user experience. And Virtual Workstation aims to provide graphics-intensive apps on a variety of device types from a variety of locations.
How well do you know virtual GPU technology?
GPU virtualization offloads graphics rendering from a user's device to the data center. Using a virtual GPU makes VDI possible for graphics-intensive apps. So how much do you know about this technology? Take this quiz and find out.
In August, these software options added new monitoring features that give VDI admins access to granular vGPU metrics. Admins can find out how many resources, such as 3D or frame buffer resources, each user consumes. Admins can use the information to assign profiles to users based on the apps and workloads they run.
"If a user is complaining about performance, one way to check it is to see what their profile is like," Sporn said.
NVIDIA vGPU supports Blast Extreme
Blast Extreme is VMware's proprietary remote display protocol. The company designed it as an alternative to PCoIP, which does a lot of its graphics rendering through CPU rather than GPU. Blast sends that work to the GPU instead. Teamed with NVIDIA's vGPU hardware and software offerings, it reduces the time between a user's click and the graphics showing up on the screen by 51 milliseconds, reduces bandwidth usage by 89% and increases density by 18%, Sporn said.
Implementing Citrix's XenDesktop vGPU feature
Compare Citrix vGPU with VMware vGPU
When is vGPU a good fit?