NVIDIA Virtual PC Best Practices

Solution Brief

TB-10862-001_v01 | February 2022

 Document History
TB-10862-001_v01
Version Date Authors Description of Change
01 February 16, 2022 UD, JL Initial release

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | ii

 Table of Contents
Executive Summary ............................................................................................................ 1
Graphics Recommendations ............................................................................................... 2
Recommended GPU: NVIDIA A16 ........................................................................................................ 2
vGPU Profiles for NVIDIA vPC Deployments ........................................................................................ 3
vGPU Configuration on the NVIDIA A16 GPU ...................................................................................... 4
Recommended ECC Memory Settings ................................................................................................. 4
Support for Multiple and High-Resolution Monitors........................................................................... 5
Server, VM, and Display Protocol Recommendations .......................................................... 6
Server Recommendations.................................................................................................................... 6
VM Configuration Recommendations ................................................................................................. 6
Golden Image Recommendations ....................................................................................................... 7
Display Protocol Recommendations .................................................................................................... 8
Windows 10 Settings for Optimum Graphics Performance .................................................. 9
Hardware Accelerated GPU Scheduling............................................................................................... 9
GPU Preferences for Software Applications ........................................................................................ 9
Monitoring....................................................................................................................... 11
Conclusion ....................................................................................................................... 12

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | iii

 List of Tables
Table 1 NVIDIA vPC User Profiles .....................................................................................................3
Table 2 Windows 10 GPU Preferences...........................................................................................10

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | iv

Executive Summary

Having a reliable virtual desktop infrastructure (VDI) is essential to any user's virtualized experience.
For NVIDIA Virtual PC (vPC), this means reliably built virtual machines that provide users a consistent
experience with flexible desktops to perform their daily work.
The NVIDIA vGPU solution is the industry's most advanced technology for virtualizing GPU hardware
acceleration. NVIDIA vPC improves virtual desktops for every user, with proven performance built on
NVIDIA GPUs for exceptional productivity, security, and IT manageability. NVIDIA virtual GPU software
divides NVIDIA GPU resources, enabling the GPU to be shared among multiple virtual machines
running any application.
This best-practices guide provides considerations for sizing your VDI desktop (such as vCPU, and
memory sizes), properly sizing a virtual GPU (vGPU) profile, and enabling hardware-based
acceleration to provide a rich user experience. The goal of this guide is to answer frequently asked
questions such as:
 What is the recommended NVIDIA GPU for my NVIDIA vPC VDI?
 What considerations should I make when evaluating NVIDIA virtual GPU (vGPU) profiles for my
users' workloads?
 What VM configurations should I consider when designing and deploying NVIDIA vPC?
Understanding users' workloads and their related frame buffer requirements can help you determine
the overall density requirements of your VDI desktop environment. Digital worker or knowledge
worker use cases will vary per user depending on many factors, including:
 Number and type of applications
 File sizes
 Number of monitors and their resolutions

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 1

Graphics Recommendations

To provide a rich user experience for VDI desktop users, select a GPU that is purpose built for high-
density, graphics-rich VDI and a vGPU profile that is optimized for VDI desktop users’ workloads. Also
ensure that the GPU’s error-correcting code (ECC) memory setting is compatible with NVIDIA vPC and
that the amount of frame buffer in your selected vGPU profiles is sufficient for the number and
resolutions of the monitors that you plan to use with each vGPU.

Recommended GPU: NVIDIA A16

Density-optimized GPUs are typically recommended for knowledge worker VDI desktops to run office
productivity applications, streaming video, and Windows 10. They are designed to maximize the
number of VDI users supported on a server.
The NVIDIA A16 GPU is purpose-built for high-density, graphics-rich VDI. Leveraging the NVIDIA
Ampere architecture, the NVIDIA A16 provides double the user density of the previous generation of
NVIDA GPUs while ensuring the best possible user experience.

Note: Although NVIDIA vPC is supported on other data center GPUs, it is essential to consider the cost per
card and total return on investment (ROI) when selecting a GPU for use with NVIDIA vPC. For additional
information, refer to the chapter about NVIDIA GPUs recommended for virtualization in the Virtual GPU
Positioning Guide.

The following features of the NVIDIA A16 GPU offer an exceptional VDI user experience:
 64 GB of memory (on four GPUs, each with 16GB of memory)

• A16-1B (1 GB of frame buffer)

• A16-2B (2 GB of frame buffer)
 Superior quality video through support for the latest codecs:
• H.265 encoding and decoding
• VP9
• AVI
 Higher resolution monitors for streaming video & multi-media
• Four encoders (7th generation) – NVENC
• Eight decoders (5th generation) – NVDEC

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 2

 Graphics Recommendations

To deploy NVIDIA Ampere GPUs, ensure that your environment meets the following requirements:
 SR-IOV is supported.
 The following software is running in your environment:
• NVIDIA virtual GPU software version 13 or higher
• VMware vSphere ESXi 7.0 Update 2 (7.0.2) or higher
Support for NVIDIA vGPU software requires the Enterprise Plus Edition of VMware vSphere
Hypervisor (ESXi). For details, refer to VMware vSphere Edition Comparison (PDF).

vGPU Profiles for NVIDIA vPC Deployments

vGPU profiles determine the amount of frame buffer that can be allocated to a vGPU. Understanding
user workloads and their related frame buffer requirements can help you choose the most suitable
vGPU profile, which determines overall density of your environment.
The following table describes the vGPU profiles for NVIDIA vPC and the types of users and
applications.

Table 1 NVIDIA vPC User Profiles

1B Profile 2B Profile
Workload Type Knowledge Workers Power Workers
Definition Runs a variety of applications, Users with specific requirements such as
web browsing, using email, multiple, high-resolution monitors to handle
creating complex documents, larger model files and higher resolution
presentations, and media.
spreadsheets. Highly mobile,
including road warriors.
Applications Video conferencing apps, Workloads running multiple vGPU
Windows 10, office productivity accelerated applications and technologies
apps, design apps, streaming simultaneously, such as Windows 10 and
video, multi-media using the office 365 productivity applications,
latest web standards like streaming 4K video and other multimedia;
WebGL, and industry-specific and industry-specific applications that use
apps like EMR applications. graphics-intensive browsers that support:
WebGL, communications platforms, secure
access capabilities, and live real-time
streaming data.
Examples Accountants, sales managers, Financial analysts, traders, medical specialists
marketing analysts, lawyers, 2D
software developers
Monitor Resolution Single HD (1920×1080) Single 4K (4096×2160)
Single QHD (2560×1440) Single 5K (5120×2880)

It is important to note that workloads will vary for each user depending on many factors, including the
number of applications, the types of applications, file sizes, and the number of monitors and their

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 3

 Graphics Recommendations

resolution. Frame buffer should be continuously monitored because user behavior can change
throughout a project, and individual roles change within the organization. A user that was once a light
user of graphics might become a heavy user of graphics after a change of teams or assignment to a
different project. Management and monitoring tools enable administrators and IT staff to ensure that
their deployment is optimized for each user.
A commonly used monitoring tool that helps organizations track their needs is the GPU Profiler. The
profiler quickly captures real-time resource utilization while workloads are being executed within the
virtual machine. Refer to the NVIDIA vPC Sizing Guide for additional information.

vGPU Configuration on the NVIDIA A16 GPU

Optimized for high user density with NVIDIA vPC, the NVIDIA A16 GPU card contains 4 physical GPUs,
each having 16 GB of frame buffer. Each physical GPU can host one of two vPC-specific vGPU profiles
that carry a predetermined amount of GPU frame buffer. The profile types can vary between different
physical GPUs on the same card but must be the same on each physical GPU. The following graphic
illustrates this point and highlights the following configurations:
 A valid configuration with A16-1B vGPUs on GPU 0 and 2
 A valid configuration with A16-2B vGPUs on GPU 1
 An invalid configuration with mixed vGPU types on GPU 3

Note: The maximum number of vGPUs that can be created simultaneously on a physical GPU is
defined by the vGPU type and the GPU frame buffer size. Refer to the NVIDIA vGPU documentation
for maximizing density on GPUs.

Recommended ECC Memory Settings

Many NVIDIA GPUs that support vGPU software support error-correcting code (ECC) memory. ECC
memory improves data integrity by detecting and correcting the most common memory data
corruption. However, ECC is not supported on B-series vGPU profiles. Although B-series vGPUs can be
created on physical GPUs on which ECC is enabled by default, keeping ECC enabled when it is not
supported will incur some costs. This incurred cost reduces the amount of usable frame buffer on the

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 4

 Graphics Recommendations

GPU, and it is recommended to disable ECC. For more information on disabling ECC memory refer to
the NVIDIA Virtual GPU Software Documentation.

Support for Multiple and High-Resolution

Monitors
An increasing need for multiple and high-resolution monitors has impacted vGPU frame buffer
requirements, exposing a need for hardware-accelerated encoding and decoding. Each additional
monitor requires an average of 15% more frame buffer. Refer to the NVIDIA vPC Sizing Guide for
additional information on the impact of multiple monitors with higher screen resolutions.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 5

Server, VM, and Display Protocol
Recommendations

It is essential to evaluate user experience properly as you design your VDI infrastructure and consider
the effects of broad-scale deployment. For more information about quantifying user experience, refer
to chapter 4 of the NVIDIA vPC Sizing Guide for deployment best practices.

Server Recommendations
Before deploying NVIDIA vPC, it is essential to confirm that your server is an NVIDIA vGPU certified
OEM server. It is recommended to cross-check this list with the VMware Compatibility Matrix to
ensure compatibility for your deployment.
The following server specifications are recommended for hosting NVIDIA vPC:
 3.0 GHz CPU
 High-speed RAM
 Fast networking
 High-speed storage

VM Configuration Recommendations
The following VM hardware specifications are recommended for NVIDIA vPC:
 vCPU: 4 vCPU
 Memory: 4 GB to 6 GB
 vGPU profile: 1B or 2B
 Network adapter: on VMware vSphere, VMXNET3
For optimum VDI performance and ease of maintenance, follow these recommendations:
 Use a flash storage array for the OS HDD with thin provisioning when possible.
 Use a separate vCenter Server instance for the VDI infrastructure.
 Use VMware Instant Clone Technology when deploying large instances.
 Remove all unused virtual hardware devices, such as CD or DVD drives, SATA controllers, COM
ports, and floppy disk drives.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 6

 Server, VM, and Display Protocol Recommendations

 Use separate VLANs for the VDI environment.

Several best-practice designs for configuring a secure and robust network for VDI environments
are available in the Horizon Architecture Planning guide.

Golden Image Recommendations

Creating a single VM that serves as the golden image or base template VM for your VDI desktops
provides consistency in your deployment.
Follow these recommendations when creating your golden image VM:
 Install the following components:
• Hypervisor management tools, such as VMware Tools
• A Windows operating system that is compatible with VMware Horizon
To ensure that your chosen Window OS is supported by vGPU software, refer to the
Supported Products page in the NVIDIA Virtual GPU Software Documentation.
• Standard Microsoft runtimes, features, and software applications
• Microsoft updates
 Generalize Windows with the Sysprep or other tools.
 Configure Windows for optimum graphics performance as explained in "Windows 10 Settings for
Optimum Graphics Performance,” on page 9.
 (Optional) Install the NVIDIA vGPU graphics driver for Windows.

Note: NVIDIA vGPU drivers are backwards compatible. Cross-check compatibility in the NVIDIA Virtual
GPU Software Release Documentation for more information.

 Create a template of the image.

Some enterprises have the flexibility to deploy Windows 10 optimization tools for their VDI
environments Golden Image VM. Here are two commonly used optimization tools:
 VMware VDI environments – VMware OS Optimization Tool
 Citrix VDI environments – Citrix Optimizer

Note: Ensure that you review and test any tool before use in a production environment.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 7

 Server, VM, and Display Protocol Recommendations

Display Protocol Recommendations

NVIDIA GPUs contain a hardware-based encoder and decoder, which provide acceleration for several
popular codecs. Complete encoding (which can be computationally complex) is offloaded from the
CPU to the GPU by using NVENC. Hardware-based decoders (referred to as NVDEC) provide faster
real-time decoding for video playback applications. When NVIDIA hardware-based encoders and
decoders are used, the graphics engine and the CPU are free for other operations, and there is no
impact on graphics performance or quality, resulting in performance that is faster than real-time
video processing. For more information, refer to NVIDIA video codec documentation.
The following display protocols leverage NVIDIA's NVENC and NVDEC hardware-based accelerators for
VDI deployments:
 For Horizon VDI deployments, NVIDIA vPC best practice is to use the VMware Blast 3D Protocol
(H.264/H.265). The VMware Blast 3D display protocol supports virtual shared pass-through
graphics acceleration technology or vGPU software. This technology allows the VM to pass all
graphics commands directly to the NVIDIA GPU, bypassing the hypervisor. The VDI desktop takes
full advantage of the NVIDIA vGPU features, accelerating application responsiveness, enhancing
the user experience, and delivering consistent performance for your NVIDIA vPC VDI desktops.
 For Citrix deployments, refer to the Remoting Protocols For GPU-Accelerated, VDI environments
whitepaper, which provides an overview of different VDI policy settings and their impact on the
user experience. This white paper illustrates how the recommended setting YUV444 is the optimal
remoting protocol for image quality and leveraging bandwidth consumption when using an NVIDIA
vPC vGPU in your Citrix VDI environment.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 8

Windows 10 Settings for Optimum Graphics
Performance

With NVIDIA vPC, Windows 10 delivers a great user experience by leveraging NVIDIA vGPU hardware
acceleration. Hardware acceleration (or offloading) allows the GPU to do what it does best, graphics
processing. Moving rendering processes from the CPU to the GPU results in a richer user experience,
with better graphics performance than emulated software acceleration. When vGPU resources are
assigned to a VM, Windows 10 is considered fully accelerated.

Hardware Accelerated GPU Scheduling

Windows has enhanced the GPU scheduling for hardware acceleration within the Windows Display
Driver Model (WDDM). The May 2020 update for Windows 10 introduced a new GPU scheduler as a
user opt-in, but it is off by default. This new scheduler reduces the overhead of GPU scheduling. For
NVIDIA vPC, it is recommended to enable the Hardware-Accelerated GPU Scheduling feature, which
allows the NVIDIA vGPU to manage its video memory instead of the operating system. Refer to
Hardware Accelerated GPU Scheduling on the Microsoft DirectX Developer Blog for additional
information on advanced graphic settings and Windows 10 hardware acceleration configuration.

GPU Preferences for Software Applications

User workloads are the most important factor in determining which vGPU profile will deliver optimal
performance. A typical Windows 10 active workload includes the Microsoft Office 365 suite of
applications. Most of these applications have hardware acceleration turned on by default (through
application settings), but additional OS configuration settings can further optimize these applications
to improve the user experience and graphics performance.
In Windows 10, you can customize the graphics performance preferences of any application. You can
force graphic-intensive applications to use the dedicated NVIDIA vGPU, improving the applications’
performance. To change the graphics processor preference for an application, go to the graphics
settings in the System display section and set the Graphics Performance Preference to high
performance for a quality user experience.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 9

 Windows 10 Settings for Optimum Graphics Performance

Table 2 Windows 10 GPU Preferences

GPU Preference Description
System default Allow Windows to decide the best GPU for your application.
Power saving Run the application on the most power-saving GPU available.
High performance Run the application on the highest-performing GPU available.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 10

Monitoring

Typical VDI deployments have two goals: achieving the best possible user experience and maximizing
user density on the hypervisor host server. Problems can arise as density is scaled up because it can
negatively impact user experience after a certain point.
It is recommended to regularly monitor hypervisor host server and VM metrics to ensure optimal
performance and resource utilization of your VDI environment. These measurements can be used to
assess the trade-offs between your end-user experience and resource utilization. Capturing
performance metrics and logging the metrics for later analysis can show trends and patterns over
time, which helps optimize your vPC VDI configuration to meet your end users’ performance
requirements and helps determine scalability.
NVIDIA vPC provides extensive monitoring features to better understand the usage of the various
engines of the NVIDIA GPU. The utilization of the compute engine, the frame buffer, the encoder, and
the decoder can all be monitored and logged on the hypervisor host server and the VM. It is
important to continually monitor resource usage to ensure system health, stability, and scalability, as
your deployment needs may change over time.
When monitoring resource usage by the hypervisor host server and guest VMs, pay close attention to
the following metrics:
 GPU engine usage on the hypervisor host server. To ensure adequate server resources and
prevent the server from doing more than it has capacity for, GPU engine usage rates should not
exceed 90%.
These metrics are essential for understanding the number of vPC VM’s that can be hosted on the
server. If metrics are high, conduct additional testing by analyzing vGPU and vCPU metrics on
individual VM’s and reduce the number of VM’s per server.

Note: In general, 80% CPU usage is a reasonable ceiling on the Host Server, and 90% usage should
be a warning that the CPUs are approaching an overloaded condition.

 Frame buffer usage in VMs. Frame buffer usage should not exceed 90% for a short time nor
average over 70% on the 1 GB (1B) profile. If high frame buffer usage is observed, the vPC virtual
machine should be assigned a 2 GB (2B) profile.
Refer to the NVIDIA vPC Sizing Guide for additional information about and tools for monitoring
hypervisor host server and VM resource usage.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 11

Conclusion

The NVIDIA vGPU solution is the industry’s most advanced technology for virtualizing GPU hardware
acceleration. This best-practices guide provides considerations for optimally configuring your NVIDIA
vPC deployment and choosing the correct virtual GPU profile.
Density-optimized GPUs are typically recommended for knowledge worker VDI desktops to run office
productivity applications, streaming video, and Windows 10. They are designed to maximize the
number of VDI users supported in a server. The NVIDIA A16 GPU is density optimized for vPC and can
increase user density compared to the previous generation.
Windows 10 features, such as hardware acceleration, allow the GPU to do what it does best, namely,
graphics processing. Customizing the graphics performance preferences of applications within the
Windows 10 operating system can further enhance the user experience.
Workloads may vary for each user and depend on many factors, including the number of applications,
the types of applications, file sizes, number of monitors, and their resolution. It is recommended that
hypervisor host server and VM metrics be continuously monitored over a product lifecycle to ensure
optimal user experience and performance while meeting organizational needs.

NVIDIA Virtual PC Best Practices TB-10862-001_v01 | 12

Notice
This document is provided for information purposes only and shall not be regarded as a warranty of a certain functionality, condition, or quality of a product. NVIDIA
Corporation (“NVIDIA”) makes no representations or warranties, expressed or implied, as to the accuracy or completeness of the information contained in this document and
assumes no responsibility for any errors contained herein. NVIDIA shall have no liability for the consequences or use of such information or for any infringement of patents
or other rights of third parties that may result from its use. This document is not a commitment to develop, release, or deliver any Material (defined below), code, or
functionality.
NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and any other changes to this document, at any time without notice.
Customer should obtain the latest relevant information before placing orders and should verify that such information is current and complete.
NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the time of order acknowledgement, unless otherwise agreed in an individual
sales agreement signed by authorized representatives of NVIDIA and customer (“Terms of Sale”). NVIDIA hereby expressly objects to applying any customer general terms
and conditions with regards to the purchase of the NVIDIA product referenced in this document. No contractual obligations are formed either directly or indirectly by this
document.
NVIDIA makes no representation or warranty that products based on this document will be suitable for any specified use. Testing of all parameters of each product is not
necessarily performed by NVIDIA. It is customer’s sole responsibility to evaluate and determine the applicability of any information contained in this document, ensure the
product is suitable and fit for the application planned by customer, and perform the necessary testing for the application in order to avoid a default of the application or the
product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or
requirements beyond those contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or problem which may be based on or attributable
to: (i) the use of the NVIDIA product in any manner that is contrary to this document or (ii) customer product designs.
No license, either expressed or implied, is granted under any NVIDIA patent right, copyright, or other NVIDIA intellectual property right under this document. Information
published by NVIDIA regarding third-party products or services does not constitute a license from NVIDIA to use such products or services or a warranty or endorsement
thereof. Use of such information may require a license from a third party under the patents or other intellectual property rights of the third party, or a license from NVIDIA
under the patents or other intellectual property rights of NVIDIA.
Reproduction of information in this document is permissible only if approved in advance by NVIDIA in writing, reproduced without alteration and in full compliance with all
applicable export laws and regulations, and accompanied by all associated conditions, limitations, and notices.
THIS DOCUMENT AND ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND
SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE
MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. TO THE EXTENT
NOT PROHIBITED BY LAW, IN NO EVENT WILL NVIDIA BE LIABLE FOR ANY DAMAGES, INCLUDING WITHOUT LIMITATION ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL,
PUNITIVE, OR CONSEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF
NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate
and cumulative liability towards customer for the products described herein shall be limited in accordance with the Terms of Sale for the product.

Trademarks
NVIDIA and the NVIDIA logo are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be
trademarks of the respective companies with which they are associated.

VESA DisplayPort
DisplayPort and DisplayPort Compliance Logo, DisplayPort Compliance Logo for Dual-mode Sources, and DisplayPort Compliance Logo for Active Cables are trademarks owned
by the Video Electronics Standards Association in the United States and other countries.

HDMI
HDMI, the HDMI logo, and High-Definition Multimedia Interface are trademarks or registered trademarks of HDMI Licensing LLC.

OpenCL
OpenCL is a trademark of Apple Inc. used under license to the Khronos Group Inc.

Copyright
© 2022 NVIDIA Corporation. All rights reserved.

NVIDIA Corporation | 2788 San Tomas Expressway, Santa Clara, CA 95051

http://www.nvidia.com