Overview

This document provides insights into leveraging NVIDIA Virtual PC (vPC) for knowledge workers. It gives recommendations based on NVIDIA’s nVector knowledge worker benchmarking and covers common questions such as:

• Which NVIDIA GPU should I use for my business needs?

• How do I select the right NVIDIA virtual GPU (vGPU) profile(s) for the types of users I will have?

• What advantages of running NVIDIA vPC versus traditional CPU-only virtual desktop infrastructure (VDI)?

Knowledge worker workloads will vary per user depending on many factors, including applications, the types of applications, file sizes, the number of monitors and their resolution. The test procedures used NVIDIA nVector, a tool for capturing real-world metrics, as the testing framework for executing a typical knowledge worker workload that simulates application workflows. Since the number of monitors and their resolution directly impact sizing, our testing to support this document explored multi-monitor setups with various screen resolutions. Tests were executed on CPU-only virtual machines (VMs) as well as VMs accelerated with NVIDIA vGPU.

It is recommended that you test your unique workloads to determine the best NVIDIA virtual GPU solution to meet your needs. The most successful customer deployments start with a proof of concept (POC) and are “tuned” throughout the lifecycle of the deployment. Beginning with a POC enables customers to understand the expectations and behavior of their users and optimize their deployment for the best user density while maintaining required performance levels. Continued maintenance is essential because user behavior can change throughout a project, as well as each individual’s role within said organization. For example, a user that was once a light graphics user might become a heavy graphics user when they change teams or are assigned a different project. Management and monitoring tools enable administrators and IT staff to optimize their deployment for each user.

The NVIDIA virtual PC (vPC) software edition enables the delivery of graphics-rich virtual desktops accelerated by NVIDIA GPUs. NVIDIA vPC allows sharing the same GPU across multiple virtual machines, delivering a native-PC experience to knowledge workers while improving user density. Because tasks typically done on the CPU are offloaded to the GPU, the user has a much better experience, and more users can be supported.

Virtual GPU profiles determine the amount of frame buffer allocated to your virtual machine. The vGPU profiles supported on NVIDIA GPUs with NVIDIA vPC software include:

• 1B profiles: 1024 MB of frame buffer (supported on Ampere and Ada GPUs only). They are ideal for high-density deployments and support single or dual HD displays.

• 2B profiles: 2048 MB of frame buffer. They are suitable for multi-monitor and higher-resolution use cases.

• 3B profiles: 3072 MB of frame buffer. Introduced with vGPU 19.0, this new profile expands the vPC variant beyond the previous 2 GB limit to address the growing demand for GPU resources in modern virtual desktop environments. The 3B profile significantly increases the number of productivity applications that users can run simultaneously. Testing shows it can support up to twice as many applications on Windows 11 compared to the 2B profile. It’s an excellent choice for customers who want to maximize GPU acceleration for a wide range of knowledge-worker workloads while maintaining strong density and responsiveness.

Using NVIDIA’s nVector testing framework, we conducted extensive testing on various configurations with both profiles to give IT admins an idea of what to expect when they scale within their own environments. Because users work in applications with varying levels of utilization, performing a POC with your workload against the testing done within this document is recommended.

NVIDIA vPC delivers an engaging user experience for the digital workplace. Users can be most productive using modern applications and work the way they want, from anywhere. When combined with GPUs such as the the A16, NVIDIA vPC delivers up to 50% better performance over CPU-only VDI ¹, enabling scalable, cost-effective virtualization with performance that rivals a physical PC.

NVIDIA’s performance engineering team developed the nVector benchmarking tool, which simulates knowledge worker workflows at scale. These workflows are representative of typical productivity applications used in enterprise VDI environments:

• Microsoft Word

• Microsoft Excel

• Microsoft PowerPoint

• Google Chrome (64-bit) for web browsing and video streaming

• PDF viewers such as Microsoft Edge

The benchmark replicates real-world tasks such as content creation, editing, switching between applications, scrolling, zooming, and exporting to PDF. Web browsers stream videos and load interactive websites. When running at scale, nVector randomizes these tasks across multiple virtual machines to simulate multi-user environments realistically.

Figure 1 Characteristics of NVIDIA’s Benchmarking Tool

The above table shows the workflow of each user. The graph shows cumulative increase in the number of users running workloads through time. Multiple users are tested at a time to simulate scale, with start and end times staggered to be more representative of real VDI environments.

As of this release, the benchmark is delivered through Login Enterprise powered by NVIDIA nVector, combining the broad testing framework of Login VSI with GPU-aware metrics and workload precision from nVector. All performance measurements in this guide reflect this integrated benchmarking platform.

The integration of NVIDIA nVector with Login Enterprise is set to transform performance testing by leveraging GPU acceleration and other crucial metrics when running compute-intensive workloads, ultimately enhancing the user experience. This combined solution offers several key advantages, including enhanced scalability for testing thousands of virtual users and ensuring the infrastructure can manage peak loads with minimal latency. It also provides improved consistency in user experience by continuously monitoring the impact of network latency on graphical performance. In addition, organizations will benefit from comprehensive analysis, gaining deep insights into both client and server-side performance to pinpoint bottlenecks and optimize resources effectively.

Figure 2 illustrates the high-level architecture of an NVIDIA virtual GPU. NVIDIA GPUs are installed within the server, accompanied by the NVIDIA vGPU manager software installed on the host server. This software facilitates the sharing of a single GPU among multiple VMs. Alternatively, vGPU technology allows a single VM to utilize multiple vGPUs from one or more physical GPUs.

Physical NVIDIA GPUs can support multiple virtual GPUs (vGPUs), which are allocated directly to guest VMs under the control of NVIDIA’s Virtual GPU Manager running in the hypervisor. Guest VMs interact with NVIDIA vGPUs similarly to how they would with a directly passed-through physical GPU managed by the hypervisor.

Figure 2 NVIDIA vGPU System Architecture

In NVIDIA vGPU deployments, the appropriate vGPU license is identified based on the assigned vGPU profile for each VM. Each NVIDIA vGPU behaves similarly to a conventional GPU, featuring a fixed amount of GPU memory and supporting one or more virtual display outputs or heads. Multiple heads can accommodate multiple displays. The vGPU memory allocation is managed by the NVIDIA vGPU Manager installed in the hypervisor, utilizing the physical GPU frame buffer at creation and retaining exclusive use of that GPU memory until termination.

All vGPUs sharing a physical GPU have access to its engines, including graphics (3D), video decode, and encode engines. For optimal performance and critical paths, a VM’s guest OS leverages direct access to the GPU, while non-critical management operations utilize a para-virtualized interface to the NVIDIA Virtual GPU Manager.

Footnotes