By rendering graphics on a VM's host server rather than on a physical endpoint device, GPUs enable complex graphics...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
without sacrificing performance. As markets for innovative visual interactions grow, there are more use cases for virtualized GPU power, and admins need to figure out how to take advantage of it.
GPU power can match resource consumption
Virtual GPUs (vGPUs) exist in many forms and uses, from virtual office desktops to on-demand 3D simulations and virtual modeling.
Many administrators and businesses that use virtual desktop infrastructure (VDI) for normal office tasks might assume that virtualized GPU power is not a significant value proposition, but GPUs are actually more important than ever.
Even low-end VDI desktops with Microsoft Word, Excel and Outlook benefit from GPU acceleration. Over the past five years, applications and web browsers have increased their resource consumption, but GPU acceleration can offset this bloat and the resulting weak performance.
GPU resources aren't a problem on a modern desktop, but in the world of VDI, the non-GPU-enabled server doesn't have the GPU resources. The net result is that all of those GPU-related system calls to redraw screens and buffer video have to be done in software with the server CPU, which creates a suboptimal experience.
For example, Windows 10 consumes 30% more resources than Windows 7, and previous options to reduce this no longer exist. Video display in Windows 10 is now an all-or-nothing proposition. If your VDI farm doesn't have GPU acceleration yet, this should convince you to seriously consider it.
The problem is that companies, even larger ones, don't completely understand the use cases for virtualized GPU power. GPUs enable a more efficient use of resources by saving CPU cycles that other virtual desktops can use. This creates a host of new opportunities, a more significant improvement than the marginal one implied by those who assume GPUs merely enable better graphics.
A lack of GPU resources means fewer resources for other users and suboptimal performance. GPUs pass hardware-assisted functions directly to VM hardware. The reduced workload, as shown by several case studies, creates the ability to fit more virtual desktop sessions on a single host. The cost for these hardware-accelerated GPU cards is inexpensive, totaling less than $6 per user, per month.
Outperform and out-save
The demands of desktops have radically changed in the past couple of decades. SaaS is now commonplace with services like Salesforce, YouTube and Facebook -- to name a few. Simple layout and functions no longer cut it.
There has been an explosion of GPU capabilities above and beyond the desktop. Nvidia recently released a supercomputer in a box clocking in at almost 2 THz. Though GPUs might not clock such ultra-high speeds, they work in a massively parallel way.
Thousands of GPU cores running at 700 MHz can outperform normal CPU cores, which inherently handle one instruction at a time. Though hacks, like hyper-threading, can improve CPU cores, this still causes suboptimal performance and is far below what GPU cores can offer.
As a lot of gamers know, however, GPUs are expensive. The sheer number of cores lends itself to compute-intensive activities, such as those required for AI, but the retooling process requires specialized programming languages and build processes. There are several APIs that can interface with GPUs, including Compute Unified Device Architecture, and expose the GPU's power to the user.
GPUs aren't only limited to VDI. Many large cloud providers now provide GPU-enhanced VMs as part of their offerings. In a recent conversation, a professional in bioinformatics revealed to me that mainframe work that had historically cost over $1 million in compute can now be done with GPU-assisted cloud instances for less than $100,000.
The move to fully virtualized environments
Companies with compute-intensive requirements, such as those using computer-aided design and computer-aided manufacturing, are moving from discrete workstations to virtual environments. Toyota, for example, now designs all its new cars within a fully virtualized environment. VDI offers compelling benefits, including the ability to create a virtual car and tour it in a virtual environment with a level of detail that allows you to see down to the ripples in the leather seat. This saves time and money and avoids the process of building and reworking full-scale mock-ups.
It also adds a completely new way to exhibit and sell products. Buyers for architecture firms, for example, can see their houses in three dimensions and resolve issues that wouldn't be visible in architectural drawings.
This virtualized GPU power also creates changes on the company's back end. Saved CPU cycles lead to saved costs in hardware, floor space and cooling. There are also savings on the front end by way of compute, network and cooling costs. This is a win-win situation, even before considering VDI's security and ease of management.
GPUs have a bright future
What are your options for starting with a GPU-assisted environment? Nvidia has a device called the Tesla M10 that provides four GPUs and 32 GB of frame buffer -- i.e., video RAM. The cost for an M10 is about $2,300.
The vendor will provide a device driver that plugs in to the hypervisor and allows the virtual desktops to access the GPU hardware. Licensing is involved in this process, so be sure to do your research.
We are only beginning to discover the use cases for virtualized GPU power. From developers to designers, people are using vGPUs to build cars, design homes and accomplish many more innovative projects that would have been cost-prohibitive only a few years ago.
GPU virtualization's uses extend beyond graphics-heavy apps
Examine the ways Citrix HDX 3D Pro supports GPU virtualization
Explore three methods of running graphics-heavy apps