What is a virtualized GPU, and what is it good for? How likely are remote or mobile device users to want or need this technology?
A graphics processing unit (GPU) performs calculations to quickly render images. Even more advanced, a virtualized GPU provides graphics to remote users without doing the actual rendering on the physical device.
NVIDIA released the Kepler family of graphics cards, which were architected to support remote access to high-end graphics rendering and services. The idea is to provide access to remote users who need complex graphics rendering work done. It does this by allowing the VM's server to interact with a local graphics resource and rendering the graphics quickly so that users can take advantage of them without having to render complex graphics on their own remote devices.
Virtualized GPU technology is particularly compelling for individuals who remote into a VM from a low-powered PC, but even more so for tablet or smartphone users. One way to accomplish this is to hook up the plumbing inside a Windows-based VM environment. Or you could use VMware Fusion for Mac OS users, special-purpose software and APIs for Linux, Android and other runtime environments.
NVIDIA's GRID boards make it possible for multiple remote users to share one or more high-end GPUs, to virtualize graphics rendering within the virtual machine. The technology is designed to minimize lag for remote users, even when rendering complex graphics. The Kepler architecture in those boards supports high-performance H.264 encoding for video streaming and can make numerous CUDA cores and physical GPUs available to a pool of remote users. NVIDIA virtualized GPUs work in the following environments:
- Citrix XenApp 6.5 with the OpenGL add-on, XenDesktop 5.6 Feature Pack 1 and XenDesktop with NVIDIA VGX software (also with XenDesktop 5.6 FP1 with HDX 3-D Pro);
- Microsoft RemoteFX in Windows Server 2012; and
- VMware View 5.2 with vSGA.
Some of the most interesting uses still haven't materialized, especially when it comes to tablets and smartphones. But today, some of NVIDIA's tools support Direct3D, OpenGL, OpenCL, CUDA and other graphics libraries available to conventional desktops, through a remote link to a VM. This permits remote access to various computer-aided design software, high-end Adobe apps and other GPU-accelerated programs inside a VM.
Virtualized GPU technology is evolving quickly and will keep upping its capabilities for some time to come.
Related Q&A from Ed Tittel
Disconnected VDI means remote users can access their desktops from anywhere, but there are some downsides.continue reading
VDI requires new hardware and software, so make sure you get some VDI training and certifications under your belt before you deploy virtual desktops.continue reading
This week, our experts explain why input/output operations per second, or IOPS, matter when measuring virtual desktop performance.continue reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.