NVIDIA has been on fire lately with strong company growth aided by particularly impressive numbers in its datacenter business. NVIDIA’s Q4 results showed datacenter revenue up 105% year-over-year contributing over 40% of the company’s overall 34% year-on-year revenue growth. Much of the datacenter revenue growth is attributed to the success of NVIDIA’s latest GPUs in hyper-scale cloud and large enterprise infrastructure for high-performance computing (HPC) and artificial intelligence (AI) based applications. However, the steadily increasing use of GPUs in the evolving virtual desktop infrastructure (VDI) market is likely another growing driver.
GPUs have been available for use in enterprise VDI deployments for years and were traditionally an add-on for graphics-intensive design and engineering workloads. Recently, however, they have become necessary in VDI solutions for many traditional knowledge worker roles due to the increasing graphics and video intensity of common productivity applications. NVIDIA has long been the leader in GPU use for VDI and is set to benefit as their usage becomes more prevalent in the VDI realm.
Due to multiple factors, a potential inflection point for GPU use in VDI deployments is forming:
- Enterprise adoption of Windows 10 drives a natural reconsideration of VDI, particularly with GPUs. Windows 7 is approaching Microsoft ’s end of extended support set for January of 2020. In looking at adoption of Windows 10, a lot of IT departments are reconsidering client computing options on a per organization/role basis — whether to refresh full PCs or deploy VDI. When re-evaluating VDI as an option for Windows 10, IT has multiple factors to consider (which I have previously written about in an in-depth paper).For one, typical knowledge worker applications now include video streaming, video conferencing, and multimedia website browsing, all of which have increased graphics use and consume substantial shared processing resources in VDI deployments. Secondly, Windows 10 and the common business productivity applications run on it have increased their graphics processing requirements by over 30% at the operating system level and more than 50% at the application level versus Windows 7. Addressing these demands with CPUs risks both hardware resource cost inefficiency and delivery of a poor user experience.User experience is the top factor in the success of VDI deployments. There is no easier way for a VDI deployment project to fail than when poor performance causes users to turn against the technology before IT can make adjustments to resolve the issues. GPUs better ensure that IT can quickly, efficiently address user performance needs. Key to NVIDIA’s leadership in this market is its GRID virtual PC capability for fine-tuning GPU resource allocation per user, which is complemented by its monitoring and management tools for ease of IT administration in optimizing resource utilization for performance and cost. NVIDIA Virtual GPU
- VDI bundles now regularly incorporate GPUs to simplify GPU-optimized deployments. In the early days of VDI, IT had to work through what stack of compute, network, storage, and software components to build, deploy, and manage to deploy a VDI solution. Hardware OEMs developed solution bundles with pre-configured and tested component stacks to address this issue. Introduced last spring, Dell EMC ’s VDI Complete even went as far as extending the single-vendor solution approach all the way to consolidated pricing, packaging, and support, but it lacked a GPU-enabled option. Dell EMC and VMware followed it up at last August’s VMworld with VDI Complete configurations including NVIDIA GPUs. Though enterprise customers often still prefer to customize aspects of their VDI solutions, these new offerings have made such strong progress towards simplifying VDI projects that many are reconsidering their options—with the added benefit of GPUs in mind.
- GPU use in virtual desktops is easier to trial than ever via cloud-based desktop-as-a-service (DaaS) offerings hosted on Amazon Web Services (AWS) and Microsoft Azure.
The DaaS delivery model is optimal for a variety of cases, such as small, remote office deployments or when an organization is addressing substantial, unpredictable staffing increases and decreases. However, a subtler benefit of the addition of GPU options is that it makes DaaS attractive for trial GPU deployments. AWS Workspaces offers GPU-enabled virtual desktops with NVIDIA GPUs. VMware offers Horizon Cloud on Microsoft Azure compute instances with NVIDIA GPUs. IT can now test virtual desktop use with target organizations to better determine when GPU-enabled virtual desktops can better address the performance needs of various knowledge worker roles. This, in turn, better informs purchase of on-premise VDI infrastructure at scale. This option is typically only pursued by the more innovative, resourceful IT leaders, but adoption is growing.
As more enterprises understand how widely and easily VDI solutions are being deployed with GPUs for Windows 10, they are likely to reconsider VDI for broader portions of their organizations. As deployments grow with adoption of Windows 10, I expect we’ll see more strong datacenter numbers from NVIDIA — not just from HPC and AI applications but from VDI deployments. We may see some of this at NVIDIA’s GPU Technology Conference (GTC) next week. I’ll be following with interest.