GPUs are Becoming Critical for VDI Success: We’ve Got Numbers

By Aditya Kaul & Kees Baggerman
February 1, 2021 | min

The global pandemic has made many organizations reliant on virtual desktop infrastructure (VDI) technology in a way they never were before. With thousands of employees across the globe suddenly working from home, IT teams have turned to VDI to deliver a great user experience that doesn’t sacrifice security. 

However, supporting a much larger number of VDI users—with more diverse needs—creates new challenges. Many teams are asking themselves for the first time whether they should add GPU acceleration to existing and new VDI deployments.

To find out whether you can benefit from GPU-accelerated Nutanix VDI solutions be sure and read our GPU solution brief and our solution note: NVIDIA vGPU on Nutanix

The Case for GPU Acceleration

There are some industries where GPUs have long been a necessity—for example Energy, Media, and any company that relies on computer-aided design (CAD). For applications like these, GPU-accelerated VDI can be a great alternative to expensive graphics workstations that have to be refreshed every few years. In other industries like healthcare, GPUs are also becoming increasingly critical to reduce latency and improve the user experience.

However, there’s also an emerging need for GPU acceleration among “knowledge workers” that crosses industry boundaries. VDI deployments that lack GPU acceleration may struggle to meet knowledge-worker needs now that the pandemic has pushed them into a work-from-home mode where they are separated from office workstations.

In recent years, the business applications that knowledge workers rely on have become more graphics intensive at the same time display resolution has increased. This can make it harder for CPU-based VDI deployments to keep up. There are also an increasing number of compute-intensive applications such as machine learning that benefit from GPU acceleration. 

Nutanix’s unique hyperconverged infrastructure (HCI) technology makes it simple to support GPU-accelerated VDI to flexibly meet the needs of knowledge workers and support the most GPU-intensive applications and power users.

Here’s a real-world example:

While working with a new customer, we did a proof of concept for NVIDIA T4 with vGPU profiles in combination with a modern Windows Server Operating System and Citrix Server VDAs. Previously this customer was using non-GPU supported legacy operating systems and their main line-of-business application was a web-based planning tool that consumed so much memory and processing power that the user experience on this shared VM was impacted for all the users on this VM. Running the planning tool would take one to two minutes, with the system consuming around 85% of the CPU processing power, rendering the system almost useless for the rest of the users on the system.

Moving to Nutanix with an NVIDIA-powered GPU solution, they found that this web application was using the added GPU capabilities, which helped drop the CPU utilization for the same planner process to around 10% CPU utilization, without impacting any of the other user sessions. This resulted in a much better user experience, overall better utilization of the resources, and fewer VMs. Accordingly, less hardware was needed to run the same workload.

Why Nutanix for GPU Acceleration?

Whether you have hundreds of users or tens of thousands, Nutanix VDI solutions deliver an excellent user experience and superior security. You can start small and grow to thousands of users without disruption, performance-related risks, or big upfront costs. Nutanix offers superior support for GPUs in both Citrix and VMware VDI environments, including:

  • More choice. Run your preferred NVIDIA GPUs on your preferred platform. Choose from Nutanix NX appliances, hardware from our OEM partners, or a wide range of third-party servers. 
  • Flexible licensing. With per-user licensing, usage is metered based on maximum concurrent users, simplifying procurement and license expansion.
  • Integrated hypervisor. Nutanix AHV makes GPU configuration extremely simple, allowing you to architect VDI infrastructure that improves performance and agility while eliminating separate hypervisor licensing costs and reducing TCO. 
  • Support for high-density and high-performance workloads. Nutanix AHV incorporates advanced technologies such as multi-vGPU and vGPU Live Migration to provide optimized support for knowledge and task workers, power users, and the most intensive applications.
  • NVIDIA. Nutanix offers advanced support for the NVIDIA® Quadro RTX™ 6000 and 8000 (passively cooled variants) on our NX-3155G and on OEM platforms, delivering the latest hardware-accelerated ray tracing, deep learning, and advanced shading capabilities. 

Benchmarking Nutanix GPU Performance

Nutanix characterized the performance of our GPU-accelerated platforms using the nVector benchmarking tool to demonstrate the advantages of GPU acceleration. nVector was developed at NVIDIA to simulate the workload that results from different end-user profiles. As described in NVIDIA’s announcement blog, nVector measures “key aspects of the user experience, including end-user latency, framerate, image quality and server utilization. This delivers better insights and feedback on the actual end-user experience, enabling IT to architect and size the VDI infrastructure based on relevant utilization thresholds.”

Our test configuration consisted of:

  • Nutanix platform: Nutanix NX-3155G-G7, dual Gold 6248R CPUs @ 3MHz, 768GB, six (6) SSDs and two (2) NVIDIA Quadro RTX8000 passively cooled GPUs. 
  • Client VMs: Windows 10 with 4 vCPUs and 8GB
  • Client vGPU profiles: T4-1B and T4-2B (1GB and 2GB framebuffer, respectively)

Nutanix tested the impact of:

  • 48 client VMs running with the T4-2B profile (equivalent to using two 4K displays)
  • 64 client VMs running the T4-1B profile (equivalent to one 4K display)

We compared these with the same client configuration without vGPUs running the same “knowledge worker” workload.

As the charts below indicate, vGPUs reduced CPU utilization substantially compared to the same clients without vGPUs.

This corresponds to an average reduction in CPU utilization of 31% with 48 VMs, and 29% for the 64 VM testing. 

More importantly, we observed a 28% reduction in latency (as illustrated below for the 64 VM testing) with a modest improvement in framerate.

Are GPUs the Right Choice for You?

If you’re reading this blog, it’s quite likely that you’re already wondering whether your VDI environment needs GPU acceleration. Perhaps end-users are complaining about excessive latency, or poor performance of applications overall. 

Nutanix GPU-Accelerated VDI solutions can help in both cases. Reducing latency by almost a third provides a noticeable improvement to users, while a 30% reduction in CPU utilization (versus the same server without GPUs) means that a single server may support more users for greater density (subject to GPU limits). The combination of a reduction in node count, flexible per-user licensing, and our integrated AHV hypervisor can make a Nutanix solution extremely cost effective.

And, while our testing focused on the “average” knowledge worker, deploying Nutanix HCI with GPUs gives you a flexible resource pool that can also support the most graphics- and compute-intensive applications.

Find Out More

Visit to learn about all our Nutanix NX platforms.

© 2021 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo, and all Nutanix product and service names are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. All other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s). This post may contain links to external websites that are not part of Nutanix does not control these sites and disclaims all responsibility for the content or accuracy of any external site. Our decision to link to an external site should not be considered an endorsement of any content on such a site. Certain information contained in this post may relate to or be based on studies, publications, surveys and other data obtained from third-party sources and our own internal estimates and research. While we believe these third-party studies, publications, surveys and other data are reliable as of the date of this post, they have not independently verified, and we make no representation as to the adequacy, fairness, accuracy, or completeness of any information obtained from third-party sources.