Technical ReportPDF Available

A Plug-Loads Game Changer: Computer Gaming Energy Efficiency without Performance Compromise

Authors:
  • Lawrence Berkeley National Laboratory (Retiree Affiliate)
April 2019 | CEC-500-2019-042
Energy Research and Development Division
FINAL PROJECT REPORT
California Energy Commission
Gavin Newsom, Governor
A Plug-Loads Game
Changer: Computer
Gaming Energy Efficiency
without Performance
Compromise
PREPARED BY:
Primary Authors:
Evan Mills
Claire Curtin
Norman Bourassa Arman Shehabi
Leo Rainer Louis-Benoit Desroches
Jimmy Mai
Nathaniel Mills
Ian Vaino
Lawrence Berkeley National Laboratory
1 Cyclotron Road
Berkeley, CA 94720
Phone: 510-486-4000 | Fax: 510-486-5454
http://www.lbl.gov
Contract Number: EPC-15-023
PREPARED FOR:
California Energy Commission
Felix Villanueva
Project Manager
Virginia Lew
Office Manager
ENERGY EFFICIENCY RESEARCH OFFICE
Laurie ten Hope
Deputy Director
ENERGY RESEARCH AND DEVELOPMENT DIVISION
Drew Bohan
Executive Director
DISCLAIMER
This report was prepared as the result of work sponsored by the California Energy Commission. It does
not necessarily represent the views of the Energy Commission, its employees or the State of California.
The Energy Commission, the State of California, its employees, contractors and subcontractors make no
warranty, express or implied, and assume no legal liability for the information in this report; nor does any
party represent that the uses of this information will not infringe upon privately owned rights. This report
has not been approved or disapproved by the California Energy Commission nor has the California Energy
Commission passed upon the accuracy or adequacy of the information in this report.
i
ACKNOWLEDGEMENTS
This project benefitted enormously from engagement with experts from the gaming industry,
other energy researchers, and real gamers. A number of individuals made themselves available
to serve on the Technical Advisory Committee and to provide interviews for the project
newsletter, Green Gaming News. AMD (Donna Sadowy,* Claudio Capobianco,* Scott Wasson,*
and Justin Murrill) and Nvidia (Tom Peterson,* Phil Eisler, Sean Pelletier, Anjul Patney, Nick
Stam, John Spitzer, Luc Bisson, and Sean Cleveland) provided valuable technical input along the
way. Representatives of the console industry, including the Entertainment Software Association
(Michael Warnecke*) and representatives from Sony Interactive Entertainment America,
Nintendo of America, and Microsoft Corp. participated in a project workshop or other
information exchanges. Game developers Nicole Lazzaro* and Bob King, shared insights into
how the coding of games may impact energy use and Tom Bui* (Steam) also provided advice
from a game-distribution vantage point. Consumer-oriented product review experts from Tom’s
Hardware (Fritz Nelson, Joe Pishgar, and Chris Angelini), PC Perspective (Ryan Shrout*), and
eXtreme Outer Vision (Slava Maksymyuk) provided invaluable discussions about energy-per-
performance assessment and consumer decision-making more broadly. The underlying market
research performed by Jon Peddie Research (Ted Pollak) laid important groundwork for the
characterization of the gaming marketplace. Other valuable market information was provided
by Iowa State University (Douglas Gentile), Fraunhofer USA (Kurt Roth), and Statistica (Liisa
Jaaskelainen). Research colleagues at other institutions provided in-depth exchanges about
benchmarking and other technical and market issues, including Jonathan Koomey (Stanford
University), Pierre Delforge* (NRDC), Peter May-Ostendorp (Xergy), Douglas Alexander
(Component Engineering), and Vojin Zivojnovik and Davorin Mista (Aggios). The authors
appreciate interactions with the United States Environmental Protection Agency’s ENERGY
STAR® program early in the project (Verena Radulovic and IFC contractors Matt Malinowski,
Ben Hill, and John Clinger).* Two dozen Lawrence Berkeley National Laboratory employees
volunteered their time to intensively test an array of gaming rigs under various operating
conditions to enable the researchers to measure energy use, performance, and user experience
under real-world conditions. Ian Vaino of Lawrence Berkeley National Laboratory’s Workstation
Support Group generously provided space and support for the green-gaming lab, system
procurement and assembly, and the extensive testing process. Sarah Morgan served as Program
Manager for the project at Lawrence Berkeley National Laboratory. Pierre Delforge, Jonathan
Koomey, Donna Sadowy, Iain Walker, and Michael Warnecke reviewed a draft of this report. The
authors extend special appreciation to Felix Villanueva, the contract manager at the California
Energy Commission, who has been highly supportive of the research process.
*Engaged Technical Advisory Committee members
ii
PREFACE
The California Energy Commission’s Energy Research and Development Division supports
energy research and development programs to spur innovation in energy efficiency, renewable
energy and advanced clean generation, energy-related environmental protection, energy
transmission and distribution, and transportation.
In 2012, the Electric Program Investment Charge (EPIC) was established by the California Public
Utilities Commission to fund public investments in research to create and advance new energy
solution, foster regional innovation and bring ideas from the lab to the marketplace. The
California Energy Commission and the state’s three largest investor-owned utilities Pacific Gas
and Electric Company, San Diego Gas & Electric Company and Southern California Edison
Company were selected to administer the EPIC funds and advance novel technologies, tools,
and strategies that provide benefits to their electric ratepayers.
The Energy Commission is committed to ensuring public participation in its research and
development programs that promote greater reliability, lower costs, and increase safety for the
California electric ratepayer and include:
Providing societal benefits.
Reducing greenhouse gas emission in the electricity sector at the lowest possible cost.
Supporting California’s loading order to meet energy needs first with energy efficiency
and demand response, next with renewable energy (distributed generation and utility
scale), and finally with clean, conventional electricity supply.
Supporting low-emission vehicles and transportation.
Providing economic development.
Using ratepayer funds efficiently.
A Plug-Loads Game Changer: Computer Gaming System Energy Efficiency without Performance
Compromise is the final report for the project by the same name (Contract Number EPC-15-023)
conducted by the Lawrence Berkeley National Laboratory. The information from this project
contributes to the Energy Research and Development Division’s EPIC Program.
For more information about the Energy Research and Development Division, please visit the
Energy Commission’s website at www.energy.ca.gov/research/ or contact the Energy
Commission at 916-327-1551.
ABSTRACT
Two-thirds of Americans play computer games. Although among the most complex and energy-
intensive plug loads, gaming has been largely overlooked in energy research and development
and policy. Systems used for computer gaming in California consumed 4.1 terawatt-hours/year
in 2016 or $700 million in energy bills, with emissions of 1.5 million tons carbon dioxide-
equivalent allocated 66 percent to consoles, 31 percent to desktop personal computers, 3
percent to laptops, and less than 1 percent to emerging media streaming devices. Key findings
include:
Aggregate energy demand places gaming among the top plug loads in California, with
gaming representing one-fifth of the state’s total miscellaneous residential energy use.
Market structure changes could substantially affect statewide energy use; energy
demand could rise by 114 percent by 2021 under intensified desktop gaming, or fall by
24 percent given a major shift towards consoles coupled with energy efficiency gains.
Unit energy consumption is remarkably varied across gaming platform types: across 26
systems tested, client-side electricity use ranged from 5 to more than 1,200 kWh per
year, reflecting equipment choice and usage patterns.
Some emerging technologies and activities are driving energy demand higher, including
processor overclocking, cloud-based gaming, higher-resolution connected displays, and
virtual reality gaming.
User behavior influences gaming energy use more than technology choice; duty cycle
and game choice are particularly strong drivers of demand.
Energy efficiency opportunities are substantial, about 50 percent on a per-system basis
for personal computers and 40 percent for consoles if past rates of improvement
continue.
While simultaneously quantifying efficiency and gaming performance is problematic, evidence
suggests that efficiency can be improved while maintaining or improving user experience.
Familiar energy policy strategies can help manage gaming energy demand, although mandatory
system-level standards are not promising (component-level measures may be).
Keywords: energy efficiency, residential, computer gaming, data centers, virtual reality
Please use the following citation for this report:
Mills, Evan, Norman Bourassa, Leo Rainer, Jimmy Mai, Claire Curtin, Ian Vaino, Arman Shehabi,
Louis-Benoit Desroches, and Nathaniel Mills. University of California, Lawrence Berkeley
National Laboratory. 2019. A Plug-Loads Game Changer: Computer Gaming System
Energy Efficiency without Performance Compromise. California Energy Commission.
Publication Number: CEC-500-2019-042.
iv
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ................................................................................................................................... i
PREFACE ............................................................................................................................................................ ii
ABSTRACT ....................................................................................................................................................... iii
TABLE OF CONTENTS .................................................................................................................................. iv
LIST OF FIGURES ............................................................................................................................................ vi
LIST OF TABLES............................................................................................................................................. vii
EXECUTIVE SUMMARY ...................................................................................................................................1
CHAPTER 1: Why This Report Is Important .......................................................................................... 10
Computer Gaming: A Largely Overlooked Use of Energy .................................................................. 10
The Most Complicated Plug Load ........................................................................................................... 12
A Highly Energy-intensive Plug Load ..................................................................................................... 12
CHAPTER 2: Energy Dimensions of the Gaming Marketplace ......................................................... 13
Market Segmentation and Installed Base .............................................................................................. 13
User Behavior and Duty Cycle ................................................................................................................. 15
Online and Cloud-based Gaming ............................................................................................................ 16
Consumers’ Information Environment .................................................................................................. 17
CHAPTER 3: The Challenges of Measuring and Benchmarking Gaming Energy Use ................ 18
Measurement ............................................................................................................................................... 18
Assessing Energy Use in Light of User Experience ............................................................................. 20
CHAPTER 4: Energy Use Across the California Installed Base of Gaming Devices ................... 23
Power Requirements at the Individual System Level ......................................................................... 23
The Energy-vs-Frame-Rate Nexus ........................................................................................................... 31
Unit Energy Consumption ........................................................................................................................ 32
CHAPTER 5: Opportunities for Gaming Energy Savings ................................................................... 40
Hardware Efficiency Measures for Desktop PCs ................................................................................. 40
Software, Operational Choices, and Other User Behaviors ............................................................... 41
Efficiency Opportunities for Consoles .................................................................................................. 46
Real-time Energy Feedback to Gamers .................................................................................................. 47
v
Non-energy factors .................................................................................................................................... 47
CHAPTER 6: Statewide Energy Demand and Projections .................................................................. 50
Past and Present Structure of Demand ................................................................................................. 52
Baseline Scenario: 2021 ............................................................................................................................ 54
Energy Efficiency Opportunities: 2021 .................................................................................................. 54
Alternate Baseline Scenario 1 Surge in High-fidelity Desktop Gaming and Virtual Reality:
2021 .............................................................................................................................................................. 54
Alternate Baseline Scenario 2 Strong Shift Towards Cloud-based Gaming: 2021 .................... 54
Alternate Baseline Scenario 3 Some PC Gamers Switch to Consoles: 2021 ................................ 55
Gaming Energy Futures for California ................................................................................................... 55
Cost of Ownership, Statewide Energy Expenditures, and Greenhouse-gas Emissions ............... 60
CHAPTER 7: Policy and Planning Pathways for Achieving Greener Gaming .............................. 62
Market Tracking and Demand Forecasting ........................................................................................... 64
Consumer Information and Tools .......................................................................................................... 64
Engagement with The Game-development Industry .......................................................................... 65
Voluntary Game Ratings ........................................................................................................................... 66
Voluntary System Ratings ........................................................................................................................ 66
Voluntary Component Ratings ................................................................................................................ 67
Mandatory Standards and Ratings ......................................................................................................... 67
Cloud-based Gaming ................................................................................................................................. 68
Broader Applications of Gaming-grade Computers and Componentry ......................................... 69
CHAPTER 8: Technology Transfer: From the Lab to the Marketplace........................................... 70
CHAPTER 9: Emerging Research Questions .......................................................................................... 74
Market Issues .............................................................................................................................................. 74
Technology Issues ...................................................................................................................................... 75
CHAPTER 10: Conclusions ......................................................................................................................... 76
GLOSSARY ......................................................................................................................................................
78
REFERENCES ................................................................................................................................................... 80
APPENDIX A: Gaming Systems Evaluated in this Study .................................................................. A-1
vi
LIST OF FIGURES
Figure ES-1: Evolution of Gaming Equipment, User Experience, and Power Requirements.............. 1
Figure ES-2: Boundary Conditions for Technology Included in This Study ......................................... 3
Figure ES-3: Baseline Systems: Desktops, Laptops, Consoles, and Media Streaming Devices .......... 4
Figure 1: Computer Gaming Consumes More Electricity in California Than Many Familiar
Residential Uses ...................................................................................................................................... 11
Figure 2: Installed Base by User Type for All Platform Types (2016) ................................................. 14
Figure 3: Duty Cycle by User Type: Personal Computers, Consoles, Media Streaming Devices,
Displays: 2016 ......................................................................................................................................... 15
Figure 4: Rapidly Escalating Global Online Gaming Throughput ........................................................ 16
Figure 5: Green Gaming Laboratory and Test Equipment ..................................................................... 18
Figure 6: Factors Affecting Gaming Performance and User Experience ............................................ 22
Figure 7: Average System Power During Gaming and Non-gaming Modes: 2016 ............................ 24
Figure 8: Measured Taming Desktop Component Loads: The Role of Components Varies
Significantly Depending on Duty Cycle and Product Tier ............................................................. 25
Figure 9: Personal Computer Power in Gameplay Does Not Vary by Genre: 19 Popular Personal
Computer Games .................................................................................................................................... 27
Figure 10: Console and Media Streaming Device Power in Gameplay Does not Vary by Game
Genre: 21 Popular Console Games ..................................................................................................... 29
Figure 11: Gaming power for Skyrim TES Varies 21-fold (from 11-221 watts) ................................ 29
Figure 12: Frame rate Does Not Correlate with PC Power: Laptop and Desktops ........................... 32
Figure 13: Baseline Unit Energy Consumption for Desktops by User Type and Duty Cycle ......... 33
Figure 14: Baseline Unit Energy Consumption for Laptops by User Type and Duty Cycle ........... 34
Figure 15: Baseline Unit Energy Consumption for Consoles by User Type and Duty Cycle.......... 35
Figure 16: Gaming is One of the Highest Energy-using Plug Loads .................................................... 36
Figure 17: Network and Cloud-gaming Energy is Often More Than Half of Total Electricity Use:
2016 Conditions ..................................................................................................................................... 39
Figure 18: Dual-Graphics Processing Unit System Draws Substantially More Gaming Power and
with Lower Frame Rates than Single-Graphics Processing Unit System, and More Still in 4k
.................................................................................................................................................................... 41
Figure 19: Virtual Reality Foveated Rendering Gradient Lowers Gaming Power >30 percent ...... 42
Figure 20: Test Results for Specific Energy Efficiency Measures ........................................................ 43
vii
Figure 21: Efficiency Improvements for Three Tiers of Desktop Systems ........................................ 45
Figure 22: Enormous Potential Variations in California Computer Gaming Energy Demand
Driven by Market Structure, User Behavior, and Energy Efficiency: 2011-2021 ....................... 50
Figure 23: Structure of California Gaming Energy Use: 2016 .............................................................. 53
Figure 24: Consoles or PCs Dominate Energy Demand, Depending on Scenario............................. 56
Figure 25: The “Intensive” User Type is the Dominant in Most Cases and Scenarios .................... 57
Figure 26: California 2021 Scenarios: Systems (left) and Categories (right) ..................................... 59
Figure 27: Green Gaming Website for Technical Audiences ................................................................. 70
Figure 28: Greening the Beast Website for Consumers ......................................................................... 71
LIST OF TABLES
Table 1: Annual California Energy Consumption, Expenditures, and Emissions for Computer
Gaming...................................................................................................................................................... 61
1
EXECUTIVE SUMMARY
Introduction
California has a long history of commitment to a wide variety of energy mandates, policies,
programs, and actions to make new and existing buildings more energy efficient. Increased
energy efficiency benefits the state’s citizens by reducing energy use and costs, lowering
greenhouse gas emissions, and avoiding the need for new power plants to meet California’s
energy demand.
Plug loads items plugged into electrical outlets by the user are one of the fastest growing
sources of energy demand in residential and commercial buildings. Depending on how plug
loads are defined, they can represent almost a third of household energy use in California
today, making them a key element in the state’s actions to increase building energy efficiency.
A relatively new contributor to plug loads is computer gaming, defined in this report as gaming
on computers, video game consoles, or media streaming devices. While Pong and other simple
games in the 1970s ran on machines drawing about 10 watts of electricity, today’s high-
performance gaming computers are among the most energy-intensive residential plug loads in
use and can draw many hundreds of watts (Figure ES-1)
Figure ES-1: Evolution of Gaming Equipment, User Experience, and Power Requirements
Source: Lawrence Berkeley National Laboratory
California is arguably the epicenter of computer gaming, with deep roots in gaming technology,
software innovations that enable the development of increasingly powerful games, and
networks that carry vast amounts of data used for cloud-based games. The state is on the
cutting edge as the home of leading component manufacturers in central processing units (Intel
and AMD), graphics processing units (NVIDIA and AMD), power supplies (Corsair), virtual reality
2
headsets (Oculus/Facebook), and gaming personal computer assemblers and system integrators
(Digital Storm). The two top game development studios (Activision Blizzard and Electronic Arts)
are also located in California. Californians are particularly avid gamers, and have somewhat
higher rates of gaming system ownership than most other parts of the country.
The rise in gaming energy use drives demand for electricity throughout the state, which in turn
boosts consumer energy bills as well as the indirect costs of energy embodied in local air
quality and climate change impacts. While the efficiency of gaming components is improving,
overall energy use remains constant or increases as the result of growing numbers of gamers,
time spent in gameplay, and demand for an increasingly (and energy intensive) vivid and
immersive user experience. However, despite its significant energy use and the potential for
energy efficiency improvements, gaming has been almost entirely overlooked in energy
research and development, policy, and planning.
The full extent of energy use by computer gaming has been largely a mystery, reinforced by its
being statistically rolled in with undefined “other” uses of energy. Private industry has made
strides in raising energy efficiency for particular products, but has not provided a
comprehensive view of the scale of energy demand from gaming products or how it might
evolve in the future. In addition, the existing literature on gaming energy use focuses almost
exclusively on game consoles. Only one formal study (now dated) has looked in depth at
gaming on desktop computers, and no work had been published regarding gaming on laptops
or with emerging television-linked media-streaming devices such as Apple TV or Android TV,
which are also used for gaming. Neither has the energy used in associated networks and data
centers for cloud-based gaming been quantified. There is also no analysis of the energy use of
many specific supplementary components, such as virtual reality equipment, high-end displays,
and external graphics processing unit docks. The effect of another key driver on energy use
game choicehas only been examined for one brand of consoles. The duty cycle (the
proportion of time during which a device is operated) unique to gamers has also not been well-
characterized, and the open literature does not describe the sensitivity of gaming energy use to
user behavior (for example, hours spent gaming).
Additional data is needed to enable the California Energy Commission and others to better
understand gaming as a driver of energy demand and to improve energy efficiency in computer
gaming as part of well-established broader strategies for managing that demand.
Project Purpose
This project meets the need for additional data by characterizing the California gaming
marketplace (technology and user behavior), defining baseline energy use and savings
opportunities in light of emerging technologies, and identifying policy strategies and
recommended actions for energy planners.
Using existing data and new measurements and drawing together the lines of data, the
researchers developed a comprehensive set of energy use estimates at the individual system
level and in the aggregate for California. These estimates provide insight into the drivers of
demand and will be useful for industry, policymakers, utilities, and consumers.
3
By filling the voids in the existing knowledge base, this project provides a novel energy-relevant
assessment for California. A key overarching premise is to identify energy efficiency
opportunities that further the state’s energy and environment goals without compromising the
gaming experience in ways that would impede adoption of improved equipment and practices.
Project Approach
The researchers’ focus in this project was on a complex energy-using activity rather than a
single energy-using device. Gaming systems are multi-function devices that perform gaming as
well as other tasks for their owners. The researchers considered all grid-connected devices used
for gaming and their displays, but did not address gaming on primarily mobile devices (Figure
ES-2). The project team also considered energy use within data centers hosting gaming
workload (cloud-based gaming) together with the networks connecting them to gamers.
Figure ES-2: Boundary Conditions for Technology Included in This Study
Source: Lawrence Berkeley National Laboratory
The researchers drew from Lawrence Berkeley National Laboratory staff expertise across several
groups, departments, divisions, and major research areas. The project team also retained
leading gaming market researchers (Jon Peddie Research), and assembled a Technical Advisory
Committee representing industry (AMD, NVIDIA, the Entertainment Software Association),
national policymakers (United States Environmental Protection Agency/ENERGY STAR®), and
other stakeholders.
The researchers consulted with industry actors such as game developers and consumer product
evaluators (PC Perspective, Hardware Canucks, Tom’s Hardware, eXtreme Outer Vision, and Bob
4
King), other researchers and institutes (Fraunhofer USA, Stanford University, Xergy), and non-
governmental organizations (Natural Resources Defense Council). During project start-up the
team held a workshop with leaders in the console industry (Microsoft, Nintendo, and Sony) to
introduce the research plan and solicit feedback. Researchers consulted the Technical Advisory
Committee, offered review drafts of key documents, and considered feedback in preparing
work products.
Early in the project, the researchers developed a detailed description of the California gaming
marketplace including hardware, software, types of users, and other drivers of the duty cycle.
The researchers developed test procedures and established a Green Gaming Systems Test Lab
at LBNL for analyzing the representative gaming devices and associated settings and software
variables, and created a data-acquisition system to aggregate and analyze the large volumes of
information collected.
The research team evaluated 26 gaming systems (10 personal computers, 5 laptops, 9 consoles,
and 2 media streaming devices) representing the range of systems found in the installed base
and on the market circa 2016 (Figure ES-3). The research incorporated componentry
representing a cross-section of major manufacturers. Bench testing included various
combinations of systems, and 37 popular games.
Figure ES-3: Baseline Systems: Desktops, Laptops, Consoles, and Media Streaming Devices
System ID codes (C1, L1, and so on) can be cross-referenced to more technical information in Appendix A.
Source: Lawrence Berkeley National Laboratory
The researchers extensively reviewed emerging technologies that may shape energy demand in
the future, as well as commercially available technologies and techniques for potentially
improving energy efficiency, including high-resolution 2D displays, virtual reality headsets,
external graphics card “docks” fitted to laptops, and a range of software. Promising strategies
currently available in the market were implemented on selected base systems and retested to
determine savings. The team did not estimate the potential of future technologies yet to be
commercialized.
5
The key barriers encountered during the project included the lack of an existing testing
protocol and the enormous variety of equipment, software, and user types that comprise the
market. Drawing on the team’s expertise, researchers captured a “snapshot in time” to
characterize the market landscape. While the tests conducted represent only a sampling of the
large combination of variables that influence gaming energy use, they do bracket the many
factors that shape energy use in this complex and rapidly changing marketplace.
Project Results
At a high level, the researchers found an enormous range in energy use among various
platforms driven as much by technology “family” (consoles versus desktops) as by gaming
behavior (hours in gameplay). While there are far fewer desktop and laptop gaming systems
than consoles in the installed base, their higher per-unit consumption makes them a significant
portion of overall statewide energy consumption, particularly under certain future market
scenarios. These variations are amplified by the role played by game choice.
Researchers were surprised by some project findings, including the dominance of consoles in
overall energy use, the impact of user behavior on outcomes, the large energy requirements of
cloud-based gaming, the significant energy efficiency gains made through the industry’s own
initiative (many through software rather than hardware), and the problematic obstacles to
applying standards as an energy savings policy strategy.
Other notable findings from the research include:
Gaming is among the top plug loads in California. In 2016, California computer gaming
used 4.1 terawatt-hours per year of electricity, representing $700 million of annual
energy costs and 1.5 million tons carbon dioxide-equivalent emissions, or one-fifth of all
residential “miscellaneous” electricity use. Of the total energy consumed in 2016 by
computer gaming equipment, 66 percent was for consoles, 31 percent for desktop
computers, 3 percent for laptops, and less than 1 percent for emerging media streaming
devices. Electricity use equated to 5 percent of overall statewide residential
consumption among the investor-owned utilities, or the equivalent of about 10 million
new refrigerators that use 400 kilowatt-hours per year. Gaming mode is responsible for
41 percent of statewide client-side energy use for consoles, 32 percent for desktop PCs,
29 percent for laptops, and 7 percent for media streaming devices based on time used.
Changes in market structure can have huge impacts on statewide energy use. Despite
the increase in gaming devices, customer energy demand was roughly constant between
2011 and 2016 as customers shifted from desktops to less energy-intensive consoles.
Alternate scenarios of market share and gamer activity projected to 2021 suggest
baseline energy consumption could increase 114 percent or decrease by 24 percent
compared to 2016 demand based on certain drivers and consumer choices. In future
scenarios, as much as 27 percent of total gaming energy used shifts to the Internet and
data centers.
Energy consumption per unit varies widely across gaming platform types and by
game choice. Across individual systems and game titles, average power during
6
gameplay varied from 34 watts to 410 watts for desktop computers, 21 watts to 212
watts for laptops, and 11 watts to 158 watts for consoles. The two media streaming
devices used similar amounts of power, under 2016 conditions
Non-gaming power requirements for PCs and consoles are within the same order of
magnitude, with a good degree of overlap although consoles use less power in this
mode on average than desktop PCs, but more in most cases than laptop PCs.
Energy use while gaming on a given gaming platform varies considerably depending on
game choice: by up to 3.5-fold among various games on PCs and by up to 1.6-fold on
consoles (with no apparent correlation between game genre and energy use).
Energy use while gaming for a given game varies by 8-fold and 21-fold of the two games
playable on the widest range of platforms in the sample.
Unexpected spikes in PC power during idle mode1 corresponded to an average of 9
percent of total energy use above that of the expected idle state across all systems (up
to 55 percent on one system). This suggests a need for more realistic test procedures.
The research team did not observe similar patterns for consoles.
Energy used by the GPU ranges from 45 to 77 percent of the total in gaming mode, and
is surprisingly significant in idle mode as well (12 to 33 percent of the total).
Some emerging technologies and activities are driving energy demand higher.
Cloud-based gaming (with graphics processing in data centers) has more energy
“overhead” than local gaming, adding about 300 watts atop local power requirements
for console-gaming and 520 watts for PC and media-streaming-device gaming.
Cloud gaming adds approximately 40 to 60 percent to the otherwise total local annual
electricity use for desktops, 120 to 300 percent for laptops, 30 to 200 percent for
consoles, and 130 to 260 percent for media streaming devices.
Virtual reality can be a very energy-intensive emerging technology, with 38 percent
higher system energy use in some cases and 15 percent less in others. When left on
continuously, the peripheral virtual reality sensors appreciably contribute to overall
energy use.
4k displays result in significant increases in energy (25 to 64 percent) used by PCs while
gaming, with reductions in frame rate, resulting in reduced energy efficiency. Consoles
have also shown to exhibit significant power increases across the duty cycle.
User behavior has a stronger influence on gaming energy use than technology choice.
Duty cycle and game choice are strong drivers of demand.
While the lightest gamers considered in the study game only about 10 minutes per day,
others game 7 hours per day.
1 All idle measurements made in using the “short-idle” test procedure.
7
Energy efficiency opportunities are substantial.
Packages of commercially available efficiency improvements (hardware, BIOS, and
software) offer a ~50 percent energy savings in PCs (in both gaming and non-gaming
modes of operation). For example, improved graphics cards reduce the amount of power
required to render games and improved power management reduces loads when not
gaming). If maintained, the observed historic rate of improvement in consoles would
reduce per-system consumption by about 40 percent between 2016 and 2021.
Strategies for improving virtual reality efficiency can lower energy use by ~30 percent.
Power management is quite poor on most PC gaming systems, including idle mode, and
the componentry is not yet supportive of energy reporting as a means of user feedback.
While simultaneously evaluating efficiency and performance is a highly problematic
undertaking, the evidence suggests that energy efficiency can be improved without
apparent reduction in user experience, although user experience is highly subjective and
resistant to quantification.
User behavior (for example hours in gameplay, overclocking, game choice, in-game
settings) has a stronger influence on gaming energy use than technology choice.
Frame rates don’t correlate with power; high performance doesn’t require high power.
Significant non-energy benefits accrue from many energy efficiency strategies.
The gaming marketplace is in constant flux, including powerful technology developments
shaping the installed base of equipment and the preferences and behaviors of gamers in a state
of perpetual change. These factors directly influence the energy intensity of individual systems
as well as in the aggregate.
Between 2011 and 2016, a shift to a less energy-intensive mix of gaming products in the
marketplace and improvements in display efficiency offset the growth in electricity demand
that would have occurred due to increasing numbers of systems in the installed base. However,
actual gaming electricity demand fell considerably as a result of significant reductions in the
electricity intensity of internet infrastructure which lowered energy use for video streaming.
Energy savings opportunities can be captured through a combination of initiative from within
the industry, consumer choices, and the energy policy and R&D community.
Technology/Knowledge Transfer/Market Adoption
Technology transfer was integral to the project approach. Key audiences included utilities,
researchers, policymakers, gaming industry representatives, and consumers. Defined in terms
of current energy use by 15 million gaming systems in California, two thirds of the current
market is comprised by console users with the remaining being PC users.
8
The approaches included formal publications, newsletters, websites, convening stakeholders,
engagement in industry activities, and media outreach (Mills 2018). All project activities have
been described on one public-facing website for technical audiences
(http://greengaming.lbl.gov) and another for consumers (http://greeningthebeast.org).
LBNL’s work in this area received considerable mainstream and trade media coverage prior to
and during the EPIC project period.2 Coverage included Forbes, Grist, Newsweek, R&D Magazine,
Science Daily, Slate, and Wired. LBNL produced news releases at the project outset and
conclusion, and a 15-minute interview was broadcast on BBC radio near project completion.
Results were disseminated through the TAC and one-on-one contacts with industry and other
stakeholders at trade meetings and other venues. One consumer information provider included
the research team’s analyses in their web-based decision tool aimed at consumers. The
Consumer Electronics Association, through reports prepared for them by Fraunhofer USA,
expanded their market survey work to incorporate PC gaming.
The technology transfer process was used to disseminate the testing protocol development. As
part of this strategy, researchers engaged with energy policy agents such as ENERGYSTAR® for
whom standardized energy-efficiency measurement techniques are essential. That said, because
most aspects of the user experience and “energy services” provided by gaming systems are not
directly measurable, and thus simple quantitative energy-per-performance metrics cannot be
articulated beyond frame rates per unit power, which is too crude for standards-setting.
Broad-based uptake of “energy thinking” in the gaming marketplace, and among gamers
themselves, is a long-term challenge that cannot be addressed by a single project or report. The
team made concerted attempts to convene sessions at two of the industry’s annual Game
Developers Conferences (GDC) without success. Attempts to collaborate with third-party
information providers that help gamers specify and build do-it-yourself systems were also
largely unsuccessful. Energy efficiency is not a high priority in the minds of most gamers and
there is no unified initiative within the industry (although there are many individual efforts).
Early experiences with disseminating energy information directly to gamers were met with a
degree of skepticism and disinterest. If gaming energy issues become more widely known and
appreciated, the marketplace should grow more receptive to the information.
The researchers explored whether the providers of the underlying software used to develop
games would be receptive to various forms of collaboration, such as integrating consideration
of energy-oriented metrics into the game-design process. Interest in this group is nascent.
Benefits to California
The introduction of a gaming device into a home can significantly increase energy costs. For the
most avid gamers, the associated energy bills can amount to hundreds of dollars each year,
particularly at a household level where multiple users and systems are in use. At the marginal
electricity prices actually paid by households, the high-end tier of desktop PCs cost light
gamers about $550 and extreme gamers $1,700/year to operate over the product’s life. This is
2 http://greengaming.lbl.gov/media.
9
in some cases more than the initial purchase cost of the gaming system. Best practices can
reduce these values by half. Conversely, significant traditional efforts to reduce a home’s
energy use (for example, improved appliances) can easily be offset or otherwise thwarted by
unaddressed computer-gaming energy.
The study identifies many readily available technologies and practices that can be adopted by
consumers, and the implications of user choices among gaming platform families as well as
discretionary in-game and system-level settings. Many of the results point the way to promising
longer-term avenues for future R&D (in partnership with industry) and more accurate
approaches to energy demand forecasting.
Computer gaming in California consumed 4.1 TWh/year in 2016 at an energy cost of $700
million, with emissions of 1.5 million tons CO2-equivalent. These amounts could more than
double under a near-term evolution of market structure. Conversely, the energy savings
opportunity for the measures considered is on the order of 50 percent for the desktop systems
and 40 percent for consoles. Overall savings will also depend heavily on efficiency
improvements in displays, networks, data centers, and energy efficient design principals in the
development of games themselves, and well as user behavioral choices.
Although many methods of achieving energy savings are accessible in today’s marketplace,
realizing the energy savings opportunities identified in this report is an enormous challenge,
particularly given the complexity of the computer-gaming energy end use. While remarkable
technological progress is being made within the gaming industry, the continual rise in
consumer expectations regarding user experience tends to offset these gains, particularly for
desktop and laptop PCs. The shift of energy to networks and data centers promises to further
obscure the energy cost of gaming. Consumer awareness of energy considerations is minimal,
and there is often resistance to the subject, based in part on misperceptions that high efficiency
and high performance are mutually exclusive. Thus, new efforts to improve awareness and
provide decision-support tools to gamers are an essential complement to R&D.
10
CHAPTER 1:
Why This Report Is Important
Household electric plug loads are loosely defined as the residual segment of energy use that
remains aside from core uses such as space conditioning, water heating, cooking, laundry, and
lighting. Depending on the definition, miscellaneous plug loads3 represent almost a third of
household energy use in California today, and a far larger proportion of energy use in otherwise
highly energy efficient homes.
Computer gaming,4 a little-discussed plug load, is a major social and technological
phenomenon, engaged in by a third of humanity. California is a major global hub for the
computer gaming industry. The associated energy use is among the most significant of all plug
loads. The issue has been understudied, and it has been passed over in most energy R&D,
policy, and planning initiatives.
Computer Gaming: A Largely Overlooked Use of Energy
Energy researchers have long recognized the importance of miscellaneous uses of electricity,
often referred to as “plug loads” (Meier et al., 1992). Consumer electronics have emerged as a
particularly important type of plug load (Rosen and Meier 2000). Lacking good accounting,
energy used by plug loads can remain uncounted for altogether or incorrectly attributed to
other end-uses. Quantifying the energy use of plug loads is an elusive challenge, by simple
virtue of their number, dynamism in the markets that drive them, and the particularly heavy
role of user behavior in determining the associated energy use.
Computer gaming, is perhaps the most extraordinary instance of this challenge, as it comprises
a myriad of platforms and use cases, in turn tempered by the consumer’s time spent gaming,
choice of software, as well as settings within the application during gameplay. Game consoles
have received some attention, but desktop and laptop computers used for gaming have only
recently come into focus (Mills and Mills 2015). The implications of a new wave of media
streaming devices that deliver gaming content based on workloads shifted to networks and
data centers have not been quantified at all. The misperception that computer gaming is
conducted only at the “fringe” of society has dampened curiosity about their role in energy use.
In this study, the researchers estimate that the entire category of computer gaming (all devices,
displays, associated network, energy, and so on) represent about a fifth of miscellaneous
3 Statewide residential electricity energy use among investor-owned utilities was 77.4 TWh in 2015 see
http://www.energy.ca.gov/contracts/GFO-15-310/12-Attachment-12-Energy-Efficiency-Data_2015-11-10.xlsx.
4 The researchers adopted the term “computer gaming” to describe gaming on computers, video game consoles, or
media streaming devices used for gaming. The terminology is inconsistently used in this industry. In some documents,
“computer” gaming refers only to PCs, while “video” gaming refers only to gaming on consoles, but in many cases the
terms are used interchangeably. The team adds references to specific platform types where a distinction is being made
in the data or discussion. Note that our analysis does not include mobile gaming on predominantly or exclusively
battery-powered devices such as tablets and smartphones.
11
electricity use in California households (Figure 1). The “Computer Gaming” category includes
multiple device types including desktop and laptop computers, consoles, and media streaming
devices and associated displays, local network equipment, and speakers, as well as associated
network and data-center energy. Values shown for Color TV are net of the estimates for their
use while operating the gaming devices, and the Miscellaneous total is net of Computer Gaming.
Gaming estimate for 2016; other end uses are estimates for 2015. See
http://www.energy.ca.gov/contracts/GFO-15-310/12-Attachment-12-Energy-Efficiency-
Data_2015-11-10.xlsx.
Figure 1: Computer Gaming Consumes More Electricity in California Than Many Familiar
Residential Uses
Source: Lawrence Berkeley National Lab
As is the case for plug loads more broadly, little attention has been paid to developing policies
and programs to achieve more energy efficient computer gaming. The two exceptions in the
United States are the highly successful 80 Plus program for voluntarily labeling power supply
unit energy efficiency and the ENERGY STAR® voluntary labeling program for computer
displays. Neither of these are particularly targeted at gaming or address the most energy
intensive components within video-gaming systems or the systems as a whole, or the enormous
vacuum in useful consumer information. Meanwhile, energy planners have largely overlooked
this particular plug load in energy forecasting.
12
The Most Complicated Plug Load
Given that proper characterization of an energy end use requires a coordinated characterization
of technology, market shares, and user behavior, computer gaming could prove to be the most
complicated plug load. A supreme challenge is that the gaming marketplace is changing faster
than data can readily be gathered and policy developed.
This report answers a wide array of critical questions not addressed in the existing public-
domain literature. These include quantifying the characteristics and relative energy use of
different families of gaming devices (desktops, laptops, consoles, and media-streaming
devices), the role of duty cycle, energy use of emerging technologies such as virtual reality
headsets, the effect of game choice and in-game settings on energy use, and energy-savings
opportunities for modifiable desktop systems through hardware as well as BIOS and software
settings. The research team isolated the influences of behavior and technology, shedding light
on the roles of each independently and in combination.
A Highly Energy-intensive Plug Load
Per-unit energy use in desktop and laptop gaming equipment has been generally rising, while
the installed base has expanded both in absolute terms and towards more energy-intensive,
higher-end platforms. While one example of gaming computer performance) has improved in
many casessuggesting improved efficienciesthis can occur even as power requirements rise.
Consoles have exhibited fundamentally different behavior, with energy use declining even as
user experience is improved. Media streaming devices are among the newer gaming
technologies and have comparatively low energy use at the device level, but high energy
intensity in their connected networks and data centers. More recently, even ordinary PCs as well
as consoles can be used for cloud-based gaming. As described below, gaming systems are
among the most energy-intensive miscellaneous plug loads in California homes.
13
CHAPTER 2:
Energy Dimensions of the Gaming
Marketplace
Computer gaming traces its roots to an exhibit created for the World’s Fair in 1940.5 Today,
three-quarters of a century later, a third of humanity engages in the pastime (NewZoo 2016),
through a myriad of types of electronic devices, including even smart watches.6 In the United
States, 66 percent of people over the age of 13 engaged in gaming in 2018, up from 58 percent
just five years earlier (Nielsen 2018). The average gamer is 35 years old, and 41 percent of
gamers are women (ESA 2016).
Surveys indicate steadily increasing numbers of gamers, amounts of time spent in gameplay,
and consumer demand for progressively more vivid and immersive user experiences seems to
have no bounds. Without offsetting efficiency gains, these driving forces stand to push energy
demand for gaming far higher.
Researchers gathered and reviewed available energy-relevant information on the computer
gaming market, including associated technology trends and gaps in the consumer information
environment. The team developed a profile of the California marketplace for the purposes of
performing energy analysis at the equipment level as well as the macro level.
The resulting analytical platform is based on best-available data and industry expert
assessments. Constituent data include an array of specific gaming systems, operated by four
user types across multi-step duty cycles, and running a representative assortment of popular
game titles. This market segmentation spans the spectrum of gaming experience, system
performance, and power requirements, leveraged to develop a characterization of the installed
base of gaming equipment and its use in California.
Market Segmentation and Installed Base
Based on an extensive review of existing market research and on original analyses developed
for this project by Jon Peddie Research (Mills et al., 2017) together with subsequent survey data
from Urban et al. (2017), the researchers developed a profile of the California marketplace for
the purposes of performing energy analysis and specified a range of 26 pre-built and custom-
built gaming systems that encompasses the range of price, functionality, and user requirements
sought in marketplace circa 2016, the base year for the projections to the future. These include
PCs, consoles, and media streaming devices. The desktop and laptop computers include those
with discrete graphical processor cards (GPUs) purpose-built for gaming as well as
“mainstream” systems with integrated graphics used for gaming. The researchers further group
gaming computers into Entry-level, Mid-range, and High-end categories, based on price and
5 See https://en.wikipedia.org/wiki/Video_game_industry.
6 See http://time.com/4617407/pokemon-go-apple-watch-release-date-2016/.
14
computing power. The team did not address popular mobile gaming devices such as
smartphones used little if at all when connected to AC power.
The researchers found that there are currently more than 15 million video-gaming devices in
use in California (the geographic focus of this study). Each user type is associated with a
segment of the installed base for each system type (Figure 2). “Android TV” represented by the
Nvidia Shield. The Nintendo Switch not yet introduced as of 2016, but is considered in the
forward-looking scenarios presented later in this report. To place the population of computers
used for gaming into a broader California context, approximately 15.4 million desktop
computers exist in homes in the state (downscaled from national estimates from Urban et al.,
(2017)), of which 15 percent are used for gaming. Of the 10 million laptops, 8 percent are used
for gaming. The decision rule for inclusion in the analysis excludes systems used fewer than
one hour per week for gaming, thus eliminating incidental use and out-of-service equipment.
While the number of desktop systems in use declined in recent years in response to the
increasing popularity of mobile gaming, it is likely to increase by about 10 percent by the year
2021, with the mix of platforms and their applications shifting towards increasingly energy-
intensive configurations while time spent gaming is gradually increasing.
Figure 2: Installed Base by User Type for All Platform Types (2016)
Source: Lawrence Berkeley National Lab
15
User Behavior and Duty Cycle
User behavior (software choices, settings, and gaming activity preferences) stands to strongly
influence gaming energy use. Thus, to properly characterize energy demand, the researchers
also developed profiles for the gaming duty cycle, disaggregated across each of the baseline
systems and user types (Figure 3). The team divided utilization into a series of modes ranging
from “off” to “gaming”.
Figure 3: Duty Cycle by User Type: Personal Computers, Consoles, Media Streaming Devices,
Displays: 2016
Source: Lawrence Berkeley National Lab
There are four types of usersLight, Moderate, Intensive, and Extreme (reflecting hours per day
in gameplay mode)each with its own duty cycles that include gaming and non-gaming
activities performed on the equipment. As seen in Figure 3, Light users dominate among entry-
level PCs, while Mid-range and High-end systems are used more heavily for gaming. Consoles
and media-streaming devices have heavier gaming-use regimes than media streaming devices.
As displays are integral to the gaming activity, the researchers incorporated them in the
analysis as well.
The most impactful part of the duty cycle is time in gameplay. Across the literature, there were
found estimates ranging from just a few minutes daily to more than seven hours, with most
reports focusing on specific platforms and/or demographics, for example, children or other age
groups (Mills et al., 2018). In the characterization of PC user types, time in gameplay ranges
from approximately 30 minutes per day for Light users to 7 hours per day for Extreme users of
desktops and 6 hours for Extreme uses of laptops. For consoles and media streaming devices,
the time in gameplay varies from 15 minutes to about 6 hours per day, respectively. For the
intensive gamers, time in sleep/standby/off modes is proportionately lower.
16
Online and Cloud-based Gaming
Gaming is rapidly expanding into the Internet, creating far-ranging implications for the
extension of associated energy use into computer networks and data center infrastructure.
According to Entertainment Software Association surveys, 51 percent of the most frequent
gamers play online games at least once weekly (ESA 2016), for an average of 0.9 hours per day
for an average of 0.9 hours per day.7 As far back as 2012, PC gamers reported spending 34
percent of their total gaming time in online mode (PWC 2012). Nielsen data suggest8 that the
popularity of online gaming is rising, with 21 percent of 7th-generation console hours spent in
that mode in 2010, increasing to 28 percent for 8th-generation consoles in 2014. Console players
now spend more time playing online games than offline games.
Online gaming is projected become the fastest-growing segment of residential Internet service
globally, with the 1.1 billion users worldwide in 2015 growing to 1.4 billion by 2020 (Figure 4)
(Cisco 2016a). These values do not include appreciable cloud-based gaming, which is still in a
nascent stage of development. Source: Cisco (2014, and other years) VNI Forecast and
Methodology reports. Notably, “online gaming” is one of only four segments of “consumer
Internet traffic” data that Cisco disaggregates, the others being Internet video, web/email/data,
and file sharing, and is the fastest-growing at 47 percent/year. Gaming devices are also used for
other Internet-based activities such as web-browsing and video streaming, which of course also
create Internet traffic.
Figure 4: Rapidly Escalating Global Online Gaming Throughput
Source: Lawrence Berkeley National Lab
The original form of online gameplay retains heavy workloads on the local client, but exchanges
meta-data among one or more gamers. Another use of the Internet in conjunction with gaming
is where games are downloaded prior to play. An emerging trend with more significant energy
7 Gameplay time via personal communication, Michael Warnecke, ESA, February 24, 2017.
8 See http://www.nielsen.com/us/en/insights/news/2016/gaming-gone-global-keeping-tabs-on-worldwide-trends.html.
17
ramifications is the actual hosting of gaming servers (including graphics processing) in data
centers, referred to here as “cloud-based gaming”. No analysis has previously been published
on the relative allocation of energy use between the local gaming client and the network of
supporting core and edge data centers (referred to here as cloud-based gaming). Cisco notes
that “if cloud gaming becomes popular, gaming could quickly become one of the largest
Internet traffic categories” (Cisco 2016b).
Consumers’ Information Environment
While many gamers are highly literate regarding their technology options—some building their
systems from scratchthe energy information available to them is incomplete and highly non-
standardized.
Most relevant information for PCs is based on rough proxies (Thermal Design Point, or TDP, in
thermal watts) of power requirements for individual components within the gaming system,
with virtually nothing available to them (other than for displays) on actual power or ultimate
energy use (combination of power and duty cycle assumptions). There are no actual power
ratings for CPUs, GPUs, or motherboards, which also makes it impossible to right-size the
associated power supplies. Of particular importance, the consumer is ill-equipped to assess the
systems integration of disparate components and their aggregate power requirements. These
systems can be “bottlenecked” in a number of ways and thus in effect oversized such that
excessively power-intensive components cannot be fully used.
There are at present no game-specific standardized energy test procedures or ratings. Thus,
consumers cannot know with confidence how their choices among different titles or genres will
affect their energy use and costs. Official test procedures for computers tend to ignore active
mode, which is problematic in the case of gaming since much of gaming system energy use
occurs in during gameplay.
Technically oriented gamers can find many product reviews in the trade literature, some of
which compare power measurements together with crude measures of performance.
Unfortunately, scores of disparate games or simulated frame-rate benchmarks are used and
there are no standardized measurement protocols. The net effect is that gamers cannot readily
compare among these various information sources, and it remains difficult or impossible to
find energy data on particular systems they may be interested in purchasing.
18
CHAPTER 3:
The Challenges of Measuring and
Benchmarking Gaming Energy Use
Dozens of factors must be considered when seeking to measure gaming system energy use and
normalize it in a fashion that reflects the widely varying possible user experiences. Among
these are the system, its connected display (2D or virtual reality), the game or benchmark run
during the test, and the metric(s) of perception deemed representative of ultimate user
experience and enjoyment of their gaming session.
Measurement
The researchers established a Gaming Systems Test Lab at LBNL for the purposes of analyzing
specific gaming devices and software variables (Figure 5). The lab allowed researchers to log
power use and frame-rate/quality for each gaming system in both gaming and non-gaming
modes (2D displays as well as virtual reality). A data acquisition platform was also developed to
aggregate and analyze the large volumes of information collected.
Figure 5: Green Gaming Laboratory and Test Equipment
Source: Lawrence Berkeley National Lab
19
Key measurements made possible in the lab were including system-level high-accuracy power
readings at one-second time intervals, power readings for individual components, large-volume
video image output storage for later analysis, component temperatures, and the durations of
individual frames produced during the gaming session
The research team evaluated 26 gaming systems (10 desktop PCs, 5 laptop PCs, 9 consoles, and
2 media streaming devices) representing the range of performance found on the market (Figure
6, Appendix A). 9 Desktop systems E1, E2, M1, and H2 were pre-built commercially available
systems. The researchers custom-built the remaining six PC systems to fill in performance gaps
along the spectrum and to represent the not insignificant do-it-yourself portion of the PC
consumer market. CPU and GPU components used in the computer systems represent multiple
generations of technology in accordance with an installed base that has developed over time.
As noted above, assessing the energy use of a gaming system requires that it be run in an
automated fashion using a simulated game (commonly referred to as a frame-rate
“benchmark”) or by a person using a real game. The team experimented with 11 frame-rate
benchmarks and 37 actual games drawn from 8 broad genres, together representing 209 game-
system combinations (Bourassa et al., 2018a-b).
All gaming-mode tests were conducted with external high-definition (HD) 1080p Dell 23in
1080p display (desktops, laptops, and consoles). In the case of C7 (Wii) the team used a
Samsung 60 1080 TV monitor because the device only has a composite out. Desktop PCs were
subsequently modified to achieve energy savings and retested, with complete packages of
measures applied representative systems in the Entry, Mid-range, and High-end market
segments.
The energy use measurements were made primarily at the system level (PC, console, and so on)
because system integration determines ultimate energy use and the focus is on the effect of
packages of measures rather than piece-wise analysis. Moreover, a given component’s energy
use will vary depending on which other components it is associated with. For example, the
energy use of a given CPU will be influenced by the motherboard on which it is mounted and
which GPU it is driving, and, in turn, the energy use of that GPU will vary depending on which
display it is running. The overall system’s energy use is further tempered by the choice of
power supply. That said, selected component-level measurements were made to inform specific
research questions.
After accounting for tests redone to resolve bad or missing data and system configuration
changes, the final total of 876 unique parametric tests spanned a variety of variables and
sensitivity studies covering a multi-step duty cycle (ranging from “off” to “gaming”). Detailed
results are presented by Bourassa et al., (2018b). PC energy use in non-gaming mode is
reasonably well defined by ENERGY STAR® and other test methodologies. PC energy use and
efficiency in gaming mode, however, is poorly defined, and not considered in ENERGY STAR® or
other existing rating systems. Power requirements can vary considerably during gameplay, as a
9 For specifications, see http://greengaming.lbl.gov/technology-assessment/representative-gaming-systems.
20
function of underlying workload created by the application and the gamer’s choices as they
move through a game’s storyline.
In developing the aggregate energy demand estimates for California, the researchers considered
the entire technology and behavioral “ecosystem” influencing energy use for gaming, treating
gaming as an activity rather than a discrete device or piece of software. These ensembles of
factors comprise the core gaming platform together with a variety of peripherals including
external audio, local networking equipment, external graphics card docks, displays, televisions,
local networking equipment, virtual reality headsets and sensors, together with a wide range of
user-driven behavioral choices that also influence energy use. Gaming systems are multi-
function devices that perform gaming as well as other tasks for their owners.
Assessing Energy Use in Light of User Experience
Ideally, the energy performance of all gaming systems could be readily compared. However,
comparisons based simply on absolute energy use for a standardized game do not suffice for
most purposes, as the ability to play different games varies among devices, as does the quality
of the gaming experience. In addition, users implement a variety of unique in-game settings, or
game modifications (“mods”), each of which will simultaneously influence the system power
draw and user experience. The CPUs and GPUs of some systems can be under- or over-clocked
to change frame rates. The choice of display can also influence system energy useand of
course user experience as wellparticularly in the case of virtual reality. Defining a “typical” or
“standard” gaming setup, reference gamer behavior and game or frame-rate benchmark is thus
an elusive goal at best.
Moreover, as found in this study, the choice of game (or simulated frame-rate benchmark)
strongly influences energy use. While identifying and applying performance metrics as proxies
for the energy services being delivered is essential to gauging technical energy efficiency,
absolute energy use must also be kept in focus as the factor ultimately driving energy cost,
pollution, and other consequences of energy use.
Two kinds of “energy services” are in play: computing services and entertainment services. The
distinction is somewhat arbitrary, but they can be delineated as component-level metrics inside
the system versus visual characteristics of the delivery of the gaming experience to the user.
Core computing services at the component level include abstract diagnostic factors such as
clock-speed or numbers of threads in a CPU or teraflops of graphics power in the GPU. Rated
metrics of this sort can be readily found for virtually any component, yet there is no explicit
translation to user experience or the degree of fit to any particular game the user may seek to
play. Moreover, there are system-integration factors that may or may not make full use of
component-level functionality, or may manifest in some but not all modes of the duty cycle. An
example of the latter is the power management capabilities of processors and the motherboard,
translating into varying levels of power reduction in non-gaming modes.
The most elementary and common example of entertainment services in gaming is the frame
rate (frames delivered per second, or fps) which can also be reported as its reciprocal, the frame
time (the duration of each frame, in milliseconds). The first of many caveats regarding these
21
metrics is that the quality and delivery of frames can vary, resulting in undesirable attributes
such as stutter (changes in the frame rate), partially-rendered or “runt” frames, and frames that
are entirely dropped (rendered by the GPU but never delivered to the display). In an important
distinction, consoles modify the quality of the frames to maintain a prescribed frame rate of 60
fps, while PCs attempt to fix quality while allowing frame rate to vary. Moreover, high frame
rates are often immaterial (for example, during game segments with relatively little visual
activity). Indeed, algorithms are now being introduced by the industry to vary frame rates
during gameplay depending on the need. The research team used specialized monitoring
systems to evaluate each and every frame in each PC test session (the technology is not
available for consoles), yielding extensive information on frame quality.
However, frame rate is just one of at least eleven gameplay entertainment services defined by
the industry (Figure 6), few if any of which can be readily measured or otherwise quantified in a
consistent manner, although users can vary some of them with in-game settings.10 There are
human limits to perception, and infinitely increasing frame rates do not necessarily translate to
a better user experience. Moreover, the relative values that end users place on these diverse
metrics are entirely subjective and vary widely across the user population. Lastly, there is
interplay and potential tradeoffs among these services and they manifest uniquely for each
game title that might be played on a given gaming device, an example being the significant
reduction in frame rate when high-definition (1080p) displays are replaced with 4k (2160p)
displays. There is no methodology for measuring the value of these tradeoffs to users. The
“integrated” service level is the user experience, which varies in a highly subjective way from
user to user, and is not rigorously measurable. As Koomey et al., (2017) point out, it is
commonly known as unquantifiable “fun”.
There is interest in comparing the performance of PCs to consoles, but there are no frame-rate
or other benchmarks that run on both technology families or accepted methodologies for
objectively, repeatably, and fairly comparing user experience. Moreover, games and gaming
technology are constantly changing, further confounding efforts to establish energy-per-
performance metrics that can be used over time. The researchers found that even the
(automated) updates pushed to the local gaming system for a given game result in significant
changes to test results. The complications and limitations of analyzing gaming energy use are
discussed further Mills et al., (2018).
10 For example, dynamic reflections vary with weather conditions in the game, as well as the amount of glass in the
scene or level of detail in the reflection. Similarly, scene complexity is dependent on the number and complexity of
artistic objects/elements in the game. This is further complicated as these elements can be adjusted dynamically and
interact with one another in complex ways.
22
Figure 6: Factors Affecting Gaming Performance and User Experience
Term
Definition
Frame rate
Frame rate, also known as frame frequency, is the frequency (rate) at which an imaging
device displays consecutive images called frames. The term applies equally to film and
video cameras, computer graphics, and motion capture systems. Frame rate is usually
expressed in frames per second (FPS).
Resolution
The display resolution or display modes of a digital television, computer monitor or
display device is the number of distinct pixels in each 2D-screen dimension that can be
displayed. It is usually quoted as width × height, with the units in pixels: for example,
"1024 × 768" means width is 1024 pixels and height is 768 pixels.
Anti-aliasing
In digital signal processing, spatial anti-aliasing is the technique of minimizing the
distortion artifacts known as aliasing when representing a high-resolution image at a
lower resolution. Anti-aliasing is used in digital photography, computer graphics, digital
audio, and many other applications.
Tone mapping
Tone mapping is a technique used in image processing and computer graphics to map one
set of colors to another to approximate the appearance of high-dynamic-range images in a
medium that has a more limited dynamic range
Rendering
Rendering is the process of generating an image from a 2D or 3D model (or models in
what collectively could be called a scene file) by means of computer programs. Also, the
results of such a model can be called a rendering.
Special effects
Special effects created for games by visual effects artists with the aid of a visual editor.
Procedural
texturing
A procedural texture is a computer-generated image created using an algorithm intended
to create a realistic surface or volumetric representation of natural elements such as
wood, marble, granite, metal, stone, and others, for use in texture mapping. In-game
setting names are highly diverse, employing terms such as “texture”, “surface”, and “map”
to identify the feature.
Scene
complexity
Scene Complexity controls the in-game representation of how detailed objects are. A
higher setting here results in more complex geometry in things like particle movement,
foliage, rocks, as well as making objects remain highly detailed at farther distances from
the player. This is due to level of detail, which is used to swap lower-resolution objects in
as the player moves farther away from them and higher resolution objects in as the player
moves closer to them. Lower settings result in a less detailed world and objects lose their
detail at closer distances to the player. Depth of field is also a component of scene
complexity.
Graphical
fidelity
Graphical fidelity can be defined as the combination of any amount of the three things that
make up beautiful games (or virtual beauty in general): detail, resolution, and frame rate
Dynamic
reflections
Realistic reflections and shadowing that move in relation to the position of objects in the
game. Also referred to as ray tracing.
Visual density
The perceived "visual density" of a screenand thus the amount of anti-aliasing possibly
needed to make computer graphics look convincing and smoothdepends on screen pixel
density ("ppi") and distance from the user's eyes.
Source: Reproduced from Koomey et al., (2017) with enhancements.
23
CHAPTER 4:
Energy Use Across the California Installed
Base of Gaming Devices
The research team made extensive power measurements across the duty cycle (Bourassa et al.,
2018b), and combined them to estimate annual energy consumption per system (Mills et al.,
2018). The focus is on client-side energy (no network or data center consumption unless
otherwise noted), and exclude peripheral uses such as displays, local networking equipment,
and external audio.
Power Requirements at the Individual System Level
Defining the power requirements of a gaming device is no easy task, particularly under
gameplay for which there is no readily established test procedure.
The research team sought to determine the variation in test results that might be encountered
for gaming mode depending on testing approach. The team ran 11 frame-rate benchmarks on
desktop computers selected from each of the three product tiers and one mid-range laptop and
compared the results to those for 10 actual game titles. Running games under simulated frame-
rate benchmarks is appealing because they are automated and highly replicable. However, the
exercise made it evident that power requirements vary depending which frame-rate benchmark
or game title is chosen.
Human gameplay is ostensibly more realistic than simulated frame-rate benchmarks, but
potentially less repeatable. Researchers evaluated both approaches. To minimize “noise” caused
by variations in human gameplay, the team developed a detailed test script for each game
(Bourassa et al., 2018a). The metrics reported here are the average power measured over a
standardized test period. For example, in the case of Skyrim, this period involved an
approximate 6-minute test of a highly scripted and repeatable section of the game.
Contrary to a popular perception that simulated frame-rate benchmarks don’t approximate
real-world gameplay, exploratory testing found that all but two of the common benchmarks
bracketed a range of power requirements very similar to those of the range of real-world games
that were tested.
Given that actual games are as or more intrinsically representative of real-world utilization and
energy use, the focus was on those results. As discussed below, the research team found that
disciplined human testing of actual games to be highly reproducible. The researchers selected
Fire Strike as the representative simulated benchmark (for PCs only) for all subsequent tests for
cases where simulated benchmarks were preferred over human gameplay.
The team subsequently tested the full range of desktop and laptop computers as well as
consoles and media-streaming devices (Figure 7). Across the systems and game titles, average
power during gameplay varied 12-fold (34 to 410 watts) for the desktops, 10-fold (21 to 212
24
watts for the laptops, and 15-fold (11 to 158 watts) for the consoles. Two media streaming
devices used similar amounts of power at approximately 4 and 8 watts.
Figure 7: Average System Power During Gaming and Non-gaming Modes: 2016
Source: Lawrence Berkeley National Lab
Conversely, for individual systems, gameplay power varied depending on the game chosen by
18-fold (15 to 270 watts) for the desktops, 41-fold (3 to 127 watts) for the laptops, 9-fold (7 to
61) watts for the consoles, and 2-fold (2 to 4 watts) for the media streaming devices. The non-
gaming power use of these systems can be significant as well, and, interestingly, follows a
different relative pattern across systems than during gameplay.
Manufacturers have brought to market external graphics-card docks for boosting laptop gaming
power. Tests of such products resulted in a three-fold increase (by 60 watts) in gaming power in
one case and two-fold (by 90 watts) in another.
The GPU plays an important role even in idle mode, and is dominant in Mid-range and High-end
systems during gameplay. That said, the CPU-motherboard assembly is responsible for half or
more of the total power in idle mode across all system tiers, and even in the entry-level system
during gameplay. The role of GPU ranges from 45 percent (System E3) to 77 percent (System
H1) in gaming mode, and is surprisingly significant in idle mode as well (12 to 33 percent)
(Figure 8). Average power over gameplay. Entry-level system is E3, Mid-range is M4, and High-
end is H1. Other is calculated as the residual of total system power minus GPU and
Motherboard power.
25
Figure 8: Measured Taming Desktop Component Loads: The Role of Components Varies
Significantly Depending on Duty Cycle and Product Tier
Source: Lawrence Berkeley National Lab
Component nameplate ratings are important insofar as DIY gamers use them to size power
supply units, and energy analysts may use them to estimate energy use in lieu of measured
values. The research team performed direct measurements of GPU and CPU/motherboard
component power for a cross-section of the base systems. Measured maximum values did not
agree well with nameplate, varying from 63 to 113 percent of actual for GPUs and 45 to 76
percent of actual values for CPUs for the units measured.
Power management
Gaming systems handle widely varying workloads, ranging from no gaming or other workloads
in idle mode to full-on gaming. Ideally, power management is implemented in system design
and system integration to scale power up and down in keeping with these varying workloads.
The concept of “energy proportionality” has been used to signify the degree to which energy
use scales with the workload in computing equipment. The degree to which this factor has been
considered in the design of the test systems clearly varies. Some of the systems performed
barely better than a 1:1 ratio (no difference between gaming and idle power for PCs and
navigation power for consoles and media streaming devices), with the best desktop PCs
operating in the range of 4:1, laptops 4:1, consoles 2.5:1, and media streaming devices 1.5:1. An
important caveat to this metric is that inefficiency in gaming mode can contribute to a greater
differential and thus the appearance of “better” energy-proportionality.
26
Displays: 2D and virtual reality
Display choice strongly affects gaming power within the gaming system. While frame rate
decline when switching from high-definition (1080p) to ultra-high definition (4k) resolution, PC
system power requirements typically rise (in systems that can handle the added processing
load). These power increases are sometimes very significant (up to 60 percent in the testing),
while frame rates decline, resulting in a significant reduction in the fps/watt metric. The
research team did not evaluate the effect of display choice on console power, but others have
observed results analogous to ours for PCs (Microsoft, Nintendo, and Sony Interactive
Entertainment 2017).
Console gaming is most commonly conducted using a television for the display, and
increasingly so as these devices become the broader “entertainment hub” for streaming video
and other services in the home. TV energy use varies widely. On-mode power requirements of
4k displays (2160p) range as high as 400 watts, and according to one report none meet the
ENERGY STAR® 7.0 qualifying levels (NRDC 2015). The leading recommended television for
console gaming from one consumer site was a 65” 4k unit, rated at 212 Watts of power when in
use.11 This is substantially more than the device-specific gaming-mode power use of most of the
consoles tested. Among 55” 4k displays, measured on-mode power use varies from 60 to
almost 170 watts, and, among the simpler measures, automatic brightness control can reduce
on-mode power requirements from 10 to 50 percent (NRDC 2015).
The average computer display power used in the assessments was 25 watts in 2016, while that
for average television was 82 watts. Treatment of displays is described more fully in Mills et al.,
(2018).
Virtual reality (VR) is gaining considerable interest among gamers, with several manufacturers
bringing products to the market for gaming computers and consoles. Initial consideration
suggests an intrinsic potential energy savings, due to the smaller active display area which is
rendered to the full display emitter resolution. However, VR requires much higher frame rates
than two-dimensional displays, thus placing greater computing demands on the gaming system
and in some cases independently powered sensors and headsets. Moreover, 2D displays are
routinely used in conjunction with VR for orientation and to enable others in the room to
follow the gaming session.
The researchers produced the first publicly available measured data on gaming computer and
console energy use under VR. The variations in PC energy use between viewing gameplay on 2D
displays versus VR headsets are notable. The direction of change varies, ranging from an
increase of 38 percent (93 watts) to a reduction of about 15 percent (52 watts run in a more
energy-efficient rendering mode).
These results include energy used by the VR headset and sensors. The Oculus Rift headset is
powered by a USB connection to the system, while the HTC Vive has a constant 16.2-watt
11 See https://www.sony.com/electronics/televisions/xbr-x900e-series/specifications.
27
accessory load provided by an external power supply that was added to the system power. Left
on continuously, the HTC sensors would consume more than 140 kWh/year.
Virtual reality is also available for PlayStation consoles. Energy use for the Batman Arkham title
under VR for the PlayStation 4 Slim and PlayStation 4 Pro resulted in power in gaming mode of
74 watts and 127 watts, respectively (excluding external display). Unfortunately, the other
Batman Arkham series games available for conventional console displays bears little
resemblance to the VR version, and so it was not possible to make the absolute comparison to
2D gameplay. System power for this game under VR was 22 to 32 percent higher than that of a
variety of 6 other popular 2D games on the PlayStation. Foveated rendering appears to be
embedded in the Playstation VR system, but with no user control or settings.
The Role of Game Choice
Variations in image quality and complexity among games suggest a wide range of rendering
workload, yet the actual correlation and corresponding variations in energy use have been
largely unquantified. The team measured gaming power requirements while running 37 games
on selected systems (none can be run on all platforms) and 11 frame-rate benchmarks.
The researchers conducted tests of 19 popular game titles across the 16 base PC systems. Even
within many of the individual systems, the range in gameplay energy was on the order of a
factor of three or ~150 watts, depending on which game title was played. For the game most
widely playable. Somewhat surprisingly, among the PC game titles tested, energy use did not
correlate with game genre (Figure 9). For example, power requirements for Candy Crush and
Sims4 did not trend lower than that of more intricate and high-fidelity games, and indeed drew
even more power in some cases (for example, compared to Skyrim TES and League of Legends).
Figure 9: Personal Computer Power in Gameplay Does Not Vary by Genre: 19 Popular Personal
Computer Games
28
The team evaluated 9 consoles and 2 media streaming devices across 21 popular games and
found qualitatively similar results (Figure 10), although with only one exception, variation
within a given platform was much less than for PCs. Measured energy use for the Nvidia Shield
was relatively low, but this is because most of the workload is shifted to upstream networks
and data centers, an issue treated later in the report. Apple TV only supports local client
gaming.
To provide a more in-depth view of how much power for given game varies across PCs and
consoles, the team evaluated power use for Skyrim across the 22 of the 26 systems with which
it is compatible (Figure 11). Skyrim is one of the least energy-intensive games evaluated in the
testing, but is available over the broadest variety of systems and hence appropriate for the
analysis depicted here. Skyrim is generally capped at 60 fps, but laptops L1 and L2 and desktop
E2 experienced bottlenecks that resulted in lower frame rates. Gameplay power levels are the
average power measured across all games. Average power during gameplay ranged from 32 to
85 watts across 5 laptops, 50 to 221 watts on 10 desktops, and 11 to 143 watts across 9
consoles. In all, gaming power while gaming varied by 21-fold across the systems. Interestingly,
frame rates are fixed at 60 fps in this game, so there are no performance differences by that
metric (except for three systems that were not capable of running at 60 fps). Researchers ran
the other widely applicable game, Sims 4, on 12 PC systems. Sims is much more
computationally-intensive than Skyrim. Average power during gameplay ranged 8.3-fold, from
32 to 269 watts.
Not all systems are able to play all games. The Fire Strike frame-rate benchmark is included for reference.
Source:
29
Figure 10: Console and Media Streaming Device Power in Gameplay Does not Vary by Game
Genre: 21 Popular Console Games
Source: Lawrence Berkeley National Lab
Figure 11: Gaming power for Skyrim TES Varies 21-fold (from 11-221 watts)
Source: Lawrence Berkeley National Lab
30
Notably, in comparing across PC and console product categories, there is clear overlap in
gaming power for the more energy-intensive consoles and, all levels of gaming laptops, and the
entry-level gaming desktops (as well as one of the mid-level desktops). Also, of interest, system
H2 (the Digital Storm - Velox) is the highest-performing system, yet under Skyrim TES uses less
energy than many of the lesser desktop systems and less than the PlayStation PS4 Pro.
Hardware and software settings and in-game “mods”
In-game settings are user-adjustable attributes of a game’s look and feel, influencing the level
of detail and realism of the scene. These effects are highly subjective and not all are necessarily
detectable by gamers. The effects measured for ten different adjustments varied between 1
percent and 6 percent. Another adjustment, VSync, achieved a far larger impact, discussed later
in this report. The ranges of effects applied individually; energy impacts would likely be greater
when applied in combination.
Games often support unique “mods” that can be installed by the user to enhance the gaming
experience. The energy effects of these settings have not previously been described. The
research team examined a series of mods for Minecraft and discovered significant impacts on
gaming power requirements.12 In particular, the Optifine mod (which increases framerate but
has no other visual impact) increased base power from 187 to 218 watts (a 16 percent increase).
Adding shaders to this setting (which dramatically enhance illumination quality, shadows, and
other details) increased power to 252 watts (a 35 percent increase from the base settings).
The team conducted exploratory tests to determine the effect of the popular behavior of “over-
clocking” the GPU and underclocking the CPU. GPU underclocking had a greater effect on
system power than overclocking (range -25 percent to +6 percent), while CPUs responded
strongly in both directions (-26 percent to +37 percent). Power use typically changed more
rapidly than performance, resulting in declining efficiency metrics (fps/W). Underclocking can
serve as a legitimate energy-savings measure, particularly in cases where changes in framerate
are not particularly noticeable to the gamer.
Variations in gameplay power requirements among actual gamers
During structured testing, the researchers captured the gaming power requirements of gaming
systems during highly scripted gameplay sessions conducted by two research staff members
and, for PCs, from simulated frame-rate benchmarks with near-perfect reproducibility.
The team also recruited 22 experienced gamers to play 87 individual game sessions in their
own way on the some of the gaming systems. The testers played a variety of game titles: fifteen
games on desktop gaming systems, eight games on consoles and two virtual reality titles also
used in the standardized bench tests.
Average power during gameplay was highly similar to the Fire Strike frame-rate benchmark and
human-gameplay measurements during the scripted lab-bench trials. Average results for given
12 These were implemented on the energy-efficient high-end PC described in Mills and Mills (2015).
31
system-game combinations for PCs were on a par at 2.5 percent lower (4 watts) with the average
bench tests and 1.7 percent lower (1 watt on average) for consoles. These discrepancies are
within the measurement error of the testing process. These testers also scored their user
experience based on five criteria. No particular pattern emerged suggesting that the “higher-
end” systems yielded a superior user experience.
These results provide high confidence in the realism and representativeness of the energy
measurements taken using lab-bench test methods, while reinforcing the aforementioned
concern that there are many elements of user experience that simplified framerate
measurements do not capture.
The Energy-versus-Frame-Rate Nexus
Popular mythology holds that boosting performance requires more energy input. While it is
true that frame may increase with rated power, the correlation is overwhelmed by many other
factors. As discussed at length above, most aspects of performance are highly subjective and
difficult or impossible to measure. The research team has been able to measure the most
accessible user-experience metric, frames per second (fps), in great detail and compare it with
measured power during gameplay. While frame rates are the predominant metric used in the
marketing of games and in product reviews, they fail to capture many aspects of user
experience.
As seen in Figure 12, high frame rates can be achieved at almost any power level. Measured
average fps and power over the frame-rate benchmark test cycle: all games and configurations.
Not all games are played or playable on all systems. Only windows systems are shown, as it was
not possible to measure frame rate for Mac OS or consoles. Conversely, at a given power level,
the frame rate achieved varies widely. Variations are similarly large when outcomes are viewed
in terms of efficiencies (frames per second per watt). A high level of efficiency does not
correlate to lower absolute power requirements.
32
Figure 12: Frame rate Does Not Correlate with PC Power: Laptop and Desktops
Source: Lawrence Berkeley National Lab
The caveats about framerate notwithstanding, these results underscore the notion that
improved efficiency needn’t require a performance compromise in the range of frame rates
generally deemed respectable.
Unit Energy Consumption
The researchers integrated the preceding assessments of power requirements by mode with the
duty cycles and other behavioral factors to estimate annual energy use for gaming. The results
represent an enormous envelope of unit energy consumption, driven by many technological as
well as behavioral variables. As with the power consumption data shown in preceding sections,
here the focus continues to be on client-side electricity consumption (no network or data center
consumption), and exclude peripheral uses such as displays, networking equipment, and
external audio.
In many cases, energy use during gameplay is on the order of one-quarter to one half the total
annual energy use across all parts of the duty cycle for the weighted-average case of all user
types. For “Extreme” users the value can rise to nearly 75 percent. An additional overarching
observation is that user type (intensity of gameplay) has as much or more impact on total
annual energy use as does the selection of gaming platform.
33
Client-side gaming
For desktops, absolute and relative energy use across the duty cycle varied substantially (Figure
13a), with particularly low relative gaming energy among the Entry-level systems. Upper panel
(a) is absolute energy; lower panel (b) apportionment by system. Total annual energy use varied
by 3-fold (248 to 648 kWh/year) across the three broad tiers of systems and their stock-
weighted average duty cycle (and much more across individual systems comprising these tiers).
Behavioral factors also strongly influence outcomes (which vary approximately five-fold), as
indicated in Figure 13b. Variations are even high within a product tier. For example, the High-
end systems’ energy use varies from 337 kWh/year for “Light” users to 1,124 kWh/year for
“Extreme” users (excluding displays and network energy), depending on user type. Viewed
differently, an Extreme user on an Entry-level system uses significantly more energy than a
Light user on a High-end system.
Figure 13: Baseline Unit Energy Consumption for Desktops by User Type and Duty Cycle
Source: Lawrence Berkeley National Lab
For laptops, absolute and relative energy across the duty cycle also varied substantially (Figure
14a), with particularly low relative gaming energy among the Entry-level systems. Upper panel
34
(a) is absolute energy; lower panel (b) shows apportionment by system. All laptop testing
conducted with batteries removed or fully charged; thus energy losses associated with charging
are not included. Total annual energy use varied by 6-fold (45 to 249) kWh/year) across the
three broad categories of systems (and much more across individual systems comprising these
tiers). Behavioral factors also strongly influence outcomes (which vary approximately 12-fold),
as indicated in Figure 14b. For example, the High-end systems’ energy use varies from 139
kWh/year for “Light” users to 515 kWh/year for “Extreme” users. Viewed differently, an
Extreme user on an Entry-level system uses only slightly less energy than a Light user on a
High-end system.
Figure 14: Baseline Unit Energy Consumption for Laptops by User Type and Duty Cycle
Source: Lawrence Berkeley National Lab
For consoles, absolute and relative energy use across the duty cycle varied substantially (Figure
15a). Upper panel (a) is absolute energy; lower panel (b) shows apportionment by system.
Annual energy consumption varies 18-fold (10 kWh/year for the Switch to 182 kWh/year for the
PS4 Pro) and 7-fold (8 to 51 kWh/year) for the media streaming devices. As discussed below,
additional unavoidable energy use not shown here is required in the upstream network and
35
data centers by the Nvidia Shield. Behavioral factors also strongly influence outcomes (which
vary approximately 75-fold), as indicated in Figure 15b. For example, the Xbox 360 varies from
34 kWh/year for “Light” users to 319 kWh/year for “Extreme” users. Viewed differently, an
Extreme user on the relatively low-energy Switch uses as much energy as a Light user on the
Xbox 360 or the PlayStation 3 Super Slim. Unlike the preceding analyses for PCs, here the
research team evaluated two generations of consoles since both are heavily represented in the
installed base. The “learning-curve” effect of improving efficiency over time is reflected the
comparison of the Nintendo Wii to the Wii U to the Switch. Current-generation systems (for
example PS4 Pro and Xbox One) will likely exhibit further improvements as their market
lifecycle progresses.
Figure 15: Baseline Unit Energy Consumption for Consoles by User Type and Duty Cycle
Source: Lawrence Berkeley National Lab
The research team assessed a hypothetical “worst-case” setup, involving the average of the two
High-end PC systems, overclocking, three displays at 4k resolution, cloud-based gaming (see
below), and the “Extreme” user profile. This configuration would result in annual electricity use
of 2,560 kWh/year (at 2016 Internet network electricity intensity), which is more than double
the Baseline unit energy consumption for that equipment tier.
36
This information can be put into context by comparisons with other residential plug loads
(Figure 16). Envelopes shown for the various platforms reflect the range of equipment selection
and time in active use (gaming, streaming, browsing) across the four user types defined in this
study. The upper bound reflects the Extreme user on the High-end equipment product tier.
Worst-case examples shown for cloud-based gaming on each device, including associated
network and data center energy. Some users will game an even greater number of hours than
indicated here. Non-gaming device values per Urban et al., (2017) and the Home Energy Saver
database. See http://hes-documentation.lbl.gov/calculation-methodology/calculation-of-energy-
consumption/major-appliances/miscellaneous-equipment-energy-consumption/default-energy-
consumption-of-mels. Gaming desktop computers are among the very most energy-intensive
plug load activities in homes. Consoles also rank quite high, especially for more intensive use
cases. Media streaming devices rank much lower, although their gameplay energy is deferred to
networks and data centers. When counting this “upstream” energy use, the media streaming
device is as or more intensive as the desktop PC. Also, of note, gaming energy use is more
sensitive to behavior than most other plug loads.
Figure 16: Gaming is One of the Highest Energy-using Plug Loads
Source: Lawrence Berkeley National Lab
Cloud-based gaming
The emergence of cloud-based gaming shifts an increasing amount of gaming activity, along
with its associated energy use, into the Internet. This type of gaming requires energy-intensive
37
server equipment located in off-site data centers to execute the game logic and render game
images, as well as the use of data networks to receive and send data from these servers to the
client-side user devices. The assessment described here is documented in substantially greater
detail in Mills et al., (2018).
The estimates of cloud-based gaming energy requirements are based on the beta version of
Nvidia GeForce NOW for the Mac, the Nvidia Shield TV system13 and analogous systems for
consoles and related published literature on data center and data network energy use in the
United States
The client-side in cloud-based gaming mode typically requires minimal power (typically 10 to
15 watts) since the majority of computer processing is occurring remotely, however the amount
of data streaming to and from the client device is significant. The Shield, for example, streams
at average rate of 15 Mbps, or 6.75 GB transmitted hourly, which translates to approximately
180 watts at the Internet electricity intensity levels prevailing in 2016.
GeForce NOW currently uses rack servers enhanced with eight Tesla P40 Nvidia GPUs. Average
server electricity use, excluding GPUs, is estimated to be 257 watts per user,14 based on typical
hardware and operation characteristics found in large data centers. When accounting for data
center server and auxiliary power, as well as the data center power when gaming services are
not being used, 340 watts is required per user at the data center while in cloud-based game.
Together with Internet requirements, total power is about 520 watts while in gameplay. This
excludes the gaming device in the home that receives the information.
The team conducted a similar analysis for consoles, which are also beginning to have access to
cloud-based gaming services such as Playstation NOW. In the absence of publicly available data,
the team developed a generic configuration. Network energy is identical to that in the Shield
example, at 180 watts on the network, plus 120 watts in the data center during gameplay. Total
cloud-based gaming power is thus about 300 watts.
Where systems elect cloud-based GPUs, the base energy on the client-side declines, although the
net effect will tend to be an increase in overall energy use unless the associated network losses
are offset by extremely significant efficiency gains within the servers in relation to the client-
side systems. Figure 17 shows that for these systems 23 to 82 percent of total system energy
use falls in networks and data centers. Values are shown for video streaming as well as gaming.
Cloud gaming values include network energy and energy used in the data center. Lower values
for Entry-level systems reflect the relatively high proportion of “Light gaming” user types. There
is currently no cloud-based gaming option for PS3, Xbox 360, Nintendo devices, or Apple TV.
Display energy not included. Accounting for network and cloud-computing energy
requirements reduces the relative energy-use differential between desktops, laptops, consoles,
and media streaming devices. For conditions prevailing in 2016, cloud gaming adds
approximately 40 to 60 percent to the otherwise total local annual electricity use for desktops,
13 See https://www.nvidia.com/en-us/shield/shield-tv/.
14 Assumes a dual-processor volume server with an average processor utilization of 50 percent.
38
120 to 300 percent for laptops, 30 to 200 percent for consoles, and 130 to 260 percent for
media streaming devices. Cloud-based gaming is by far the most energy-intensive form of
gaming via the Internet (compared to traditional online gaming or downloading games), and
while the electricity intensity of networks is declining quickly, that of data centers is not.
39
Figure 17: Network and Cloud-gaming Energy is Often More Than Half of Total Electricity Use:
2016 Conditions
Source: Lawrence Berkeley National Lab
These estimates are based on representative equipment and published data, but cloud-based
gaming is an increasingly diverse and rapidly evolving gaming medium. The centralization of
servers handling the graphics workload provides unique opportunities for efficiency
improvement (technological and operational) as well as introducing carbon-free power sources.
The energy efficiency of data networks has drastically improved while the amount of data being
transferred is constantly increasing. That said, the energy use attributed to cloud-based gaming
is reliant on the amount of time the equipment remains unused and idle while still consuming
electricity. Providing more cloud-based gaming capacity than needed will ultimately increase
the energy intensity of these services. Much uncertainty remains in the specific energy use
values of current and future cloud-based gaming, but the estimates provide a framework for
future analysis and outline the energy-consuming components associated with cloud-based
gaming that require attention to better understand the energy impact of this emerging form of
popular entertainment.
It is important to note that other use modes available for client-side gaming devices also
consume network energy. During video streaming, the client-side user device is used to view
video content stored in the cloud through services such as Netflix, Hulu, or YouTube. While the
data center energy use during video streaming has been shown to be negligible on a per-viewer
basis (Shehabi et al., 2014), the data streaming to and from the client device can be significant
depending on the quality and resolution of the video, corresponding to about 100 watts during
streaming.
40
CHAPTER 5:
Opportunities for Gaming Energy Savings
The gaming industry (and makers of components used in gaming systems) has made material
efforts at improved energy efficiencies, in some cases in tandem with policy efforts and in
other cases on their own.
Component efficiencies have improved steadily, along with efforts to achieve power
management through software and BIOS avenues. Console manufacturers have made the
greatest strides. What can be observed is that each in-generation release of consoles has
historically achieved energy savings compared to the prior version, and that the cross-
generation trend is generally downwards. The researchers estimate that the historical rate of
improvement in console power per unit has been 11 percent per year, representing a blend of
potential current-generation and next-generation improvements. The leading manufacturers
have publicly identified thirteen specific strategies that have been applied to various usage
modes of the 8th-generation Microsoft and Sony consoles, and project these to reduce energy
use in compared to the baseline by 65 percent by the year 2020 although claim “little further
opportunity for reduction” beyond that, although recognizes that greater reductions are
“conceivable” (Microsoft, Nintendo, and Sony Interactive Entertainment 2017).
As an indication of further potential, there remain large variations in energy use during
gameplay across the representative systems evaluated while providing similar measurable user
experiences, as well as variations in the ability of systems to use less power in non-gaming
modes. Moreover, for PCs, a steady stream of software innovations is entering the market that
depends on users to implement.
The researchers tested a wide range of commercially available hardware and operational
changes and measured their savings, individually and in packages (Mills et al., 2018), but did
not estimate the potential of future technologies yet to be commercialized.
Hardware Efficiency Measures for Desktop Personal
Computers
Historical progress notwithstanding, virtually every component in gaming systems can be more
efficient. This includes central processing units, graphical processing units, motherboards,
power supplies, cooling, as well as displays and other peripherals devices. For networked
gaming, the opportunities extend into networks and upstream data centers.
GPUs offer by far the greatest fractional savings opportunities. As an illustration of these
improved GPU opportunities, the upgrade of a High-end DIY system (H1) achieved substantial
energy savings by changing from two AMD R9 Fury X GPUs (the base system) to one RX Vega 64
liquid-cooled GPU. Notably, savings varied considerably depending on which game and display
was in use (Figure 18). That said, power consumption for each game increased when the high-
41
definition (1080p) display was replaced with the ultra-high-definition (4k) display. The metric
fps/W improved for all cases.
Figure 18: Dual-Graphics Processing Unit System Draws Substantially More Gaming Power and
with Lower Frame Rates than Single-Graphics Processing Unit System, and More Still in 4k
Source: Lawrence Berkeley National Lab
Of note, the researchers observed significantly improved power management within the
efficient GPUs evaluated. In these cases, the ratio of gaming-to-idle power increased
considerably.
Power supply units also offer material savings opportunities. Potential improvements over the
units shipped with the tested commercial systems averaged 13 percent. Moreover the systems
were virtually all significantly oversized by a factor of three on average for the desktops and by
25 percent for the laptops--suggesting further savings opportunities.
Software, Operational Choices, and Other User Behaviors
User choices regarding system BIOS and software settings, duty cycle, and in-game settings can
have as much influence on energy use as their hardware choices. Parametric analyses were used
to isolate the effect of these variables, and to understand their influence in combination.
Gamers have historically been irked by visual anomalies in 2D displays such as image “tearing”
and “stuttering”. Tearing occurs when a frame is outputted by the GPU while the monitor is in
the middle of a refresh. One solution to this issue involves enabling VSync (Vertical Sync),
forcing the GPU to wait to release frames until the monitor is ready to refresh itself. Energy
savings can result if the system would otherwise operate at higher frame rate, essentially
reflecting a system that is undersized for the game it is trying to run. However, this can cause
42
unacceptable delays in screen refreshes which users must trade off against lower-quality
frames. In the testing, VSync achieved 14 percent and 39 percent power reductions for the M2
and H2 systems, respectively. The researchers did not observe material savings for other 2D-
display strategies such as G-Sync and FreeSync.
Among the particularly impactful operational measures is dynamic voltage frequency scaling
(DVFS), which automatically slows frame rates when the rendering requirements are not critical.
Other studies have found large savings from this strategy,15 although savings were far lower on
game types where activity levels are particularly constant.
Another dramatic savings opportunity observed was foveated rendering for virtual reality, in
which image quality is gradually attenuated towards the periphery of the field of vision.
Researchers measured 30 to 36 percent savings for this strategy, depending on which VR
headset was in use (Figure 19). Results are for system H2. Excludes power of secondary 2D
display commonly used in conjunction with VR. An important caveat for the future is that
gaming equipment manufacturers or software developers could “take back” these savings in the
form of increased performance and associated computing power.
Figure 19: Virtual Reality Foveated Rendering Gradient Lowers Gaming Power >30 percent
Source: Lawrence Berkeley National Lab
When properly implemented, neither of these strategies compromise user experience, and can
in fact enhance it by reducing congestion in the graphics pipeline.
The preceding discussion provides a sense of the array of efficiency measures available to
gaming system designers and owners. A cross-section of results from the testing is provided in
Figure 20. These results are on diverse systems (noted in the axis labels) are measured
independently of other measures applied to the given system. Thus, these values cannot be
combined in an additive or multiplicative fashion. The Antialiasing and Qualities cases reflect
15 See https://www.tomshardware.com/reviews/amd-radeon-chill-ocat-relive,4846.html.
43
the change in power use over the full range of settings. VSync tests did not give reliable FPS
results, which are omitted here. The PSU impact is calculated across a range of system types.
Chill is likely to have significantly greater savings on games with less constant activity levels.
Frame rate could not be measured while in VR.
Figure 20: Test Results for Specific Energy Efficiency Measures
Systems Integration
Gaming involves complex assemblies of components. The primary device (PC, laptop, console,
or media-streaming device with associated data centers) contains many interacting components
and subsystems, and are in turn connected to peripheral devices (displays, VR headsets, audio
equipment, and so on) that create further interactions. And, of course, the gamer is part of the
system as well, making key operational choices and decisions and ultimately perceiving the
output, which is the ultimate service produced by the system. As with virtually every energy-
using system, proper systems integration offers pathways to reduced energy use and improved
performance beyond what can be achieved by piecemeal measures.
Even with today’s much-improved componentry, gains can be made with right-sizing. The most
familiar sub-optimization in this regard is the oversizing of power supplies, which operate less
efficiently on either side of 50 percent load. Other subtler interactions occur when systems are
Source: Lawrence Berkeley National Lab
44
“over-spec’d”, meaning components are more powerful than needed to run the games desired
by the user, and/or one component’s capacity causes bottlenecks with another. The typical case
of the latter is a CPU more powerful than necessary to run the selected GPU, creating
“bottlenecking” that results in higher than necessary energy use with no performance benefit.
The measurements determined that display choice has a strong effect on energy use within the
PC, particularly the graphics card. As shown above, the choice of virtual reality can have an
even more profound influence on energy use and when the gamer is recognized as part of the
“system” advantage can be taken of diminished perception in the periphery of the field of
vision to throttle back rendering (and thus computing workload and associated energy use) in
that region.
Finally, the energy use patterns of gaming equipment cannot be defined without understanding
key user preferences: game choice, in-game settings, and the duty cycle.
Savings packages
The research team assembled packages of hardware, BIOS, and system-settings measures for
each of the three PC systems (one from each market tier), which were tested under efficiency
measure retrofit scenarios. Detailed breakdowns of the measure packages are provided by Mills
et al., (2018). Given the large number of potential component combinations, and limitations on
the number of tests conducted suggest that deeper energy savings could well be identified. This
applies particularly in the case of CPUs and motherboards, as well as to software and in-game
settings. The team also did not include VSync, which can clearly achieve large savings in higher-
performance systems. Based on lab-bench testing of the PCs, researchers identified sets
(“packages”) of applicable measures and evaluated their impact on the desktop systems from
each of the performance tiers. Among the hardware measures, the primary focus was on the
GPUs, as they are the key driver of energy use. The efficiency packages for PCs varied by
system.
For the desktop systems, the team found overall average measured savings of 52 percent in
gaming mode and 48 percent in non-gaming mode. The resulting savings in gaming mode
ranged from 29 to 54 percent and those in the non-gaming mode ranged from 35 to 62 percent.
A further breakdown for hardware and operational measures is outlined in Figure 21.
Additional energy saving factors and strategies have not been included in this analysis. Among
these are:
Deep savings are possible through VSync, but the measure is only applicable to systems
that are sufficiently powerful to not experience unacceptable reductions in frame rate.
Benefits of “right-sizing” componentry, particularly GPUs to match actual gaming need
and displays set at a resolution matching the need.
Certain minor component-level measures. These include more efficient fans or fan-less
cooling of PCs.
Innovations in game design and code management to reduce energy use without
compromising performance or user experience.
45
Behavioral choices that could be made by gamers involving duty cycle changes.
Consumer product choices (beyond those captured in the scenarios) made with the
intent of reducing energy use. Among these would be a shift towards less energy-
intensive laptop computers for gaming or a shift to less energy-intensive consoles or
media streaming devices.
In many climates, waste heat from gaming contributes to household air-conditioning
costs, which will decline as gaming systems become more efficient.
Figure 21: Efficiency Improvements for Three Tiers of Desktop Systems
s
46
Source: Lawrence Berkeley National Lab
A key observation from Figure 21 is that the energy use of the improved high-end system was
in range of that of the entry-level system. Performance and temperature metrics are averages
measured during gameplay. Non-energy factors occurring in parallel with the energy efficiency
improvements include cases of improved frame rates, improved frame quality, reduced system
stress, and significantly reduced CPU and GPU temperatures.
Laptops and media streaming devices are sealed systems, and represent a very small segment
of gaming energy use. The researchers did not attempt to estimate efficiency opportunities for
these devices.
Efficiency Opportunities for Consoles
As video game consoles are also sealed proprietary systems, the research team did not attempt
any direct efficiency improvements to these devices. Rather, the team estimated the historical
year-on-year reductions power levels for 7th-generation consoles (Xbox 360, PS3) and fitted
these to obtain an average improvement rate of 41 percent by the year 2021. The researchers
applied this improvement rate to the latest-generation Xbox One and PS4 systems’ power levels,
as these are the models that replace older consoles in the stock projections and which will
likely be undergoing further improvements, or succeeded by more efficient models. The
consoles industry does not readily discuss or publish energy efficiency opportunities going
forward.
47
Real-time Energy Feedback to Gamers
The vast majority of gaming consumers do not particularly understand or prioritize addressing
the energy consumption of their gaming systems. Moreover, the wider computer game
developer industry does not have a significant built-in market incentive to produce game titles
that consider and visualize the game system energy consumption among its primary concerns.
Looking at this from the game developer and user’s point-of-view, Task 6 of this project began
with the hypothesis that easily-assimilated real-time energy reporting readouts, integrated
within the gaming user interface, could alter the gamer’s consideration of their game system
energy consumption.
To test this assumption, the project team documented the current state of the hardware,
software and game developer industry ability to deploy real-time energy reporting software
implementations that can be implemented in server-client gaming systems (Vaino et al., 2018).
The primary research vehicle for this exploration was a proof-of-concept energy reporting
software system on high performance computer (HPC) workstations currently in use by the
LBNL Engineering Division,16 running applications such as Computer Assisted Design (CAD),
modeling software or scientific data analysis tools.
The software that was developed collected real-time power draw data using APIs provided by
the GPU manufacturer. The power draw data was stored on a central database, with metadata
about the hardware, operating system, CPU and GPU loads, and running applications.
In addition, within the multiple gaming industry outreach activities and recruitment of
Technical Advisory Council members for the project, the team met with three established game
developers for ideas that can help integrate game system energy efficiency into the active
incentive structures of multiplayer games.
The exploration of technical and gaming industry development factors found that the real-time
reporting concept is plausible despite some soluble technology barriers. However, more
favorable market conditions are needed before the concept can attract the right gaming
industry champions to take on the challenge.
Non-energy Factors
Non-energy factors are often key drivers of consumer interest in improved energy efficiency
(Mills and Rosenfeld 1996)or, conversely, can become reasons that consumers reject the
efficiency recommendations. For most gamers, these benefits (or perceived downsides) are
decidedly more important than energy use per-se.
A current example is the strong desire to achieve wireless VR headsets. First-generation
headsets are physically tethered to the PC, creating discomfort and restricted range of motion
for the gamer, as well as safety hazards. Energy efficiency may offer a pathway for solving this
problem.
16 http://engineering.lbl.gov.
48
In a more generalized example, waste heat production is a side effect of high energy intensity
that irks most gamers. All electricity entering the system ultimately becomes heat, and thus a
500-watt gaming system is like a 500-watt space heater. The problem is significant enough that
it is some gamers place a portable fan or AC unit next to their gaming area. Conversely, energy
savings translate directly into less heat production.
Cooling systems in desktop systems, usually involving multiple fans, are also a source of
unwanted noise that many gamers find distracting. More efficient devices can enable the
elimination of fans, or algorithms that run the fans only when needed.
Gaming laptops offer an interesting “existence proof” of how non-energy factors drive
efficiency improvements. Key design constraints are heat removal and duration of gameplay on
a given battery charge, both of which are served by maximizing efficiency so as to reduce waste
heat and obtain the greatest number of hours of operation on a given battery charge.
Systematically lower energy use is attained by gaming laptops. The advent of the Nintendo
Switch is another example of this process, that is, miniaturization and efficiency pursued to
achieve portability and long battery life.
There are indications that certain energy efficiency strategies may improve game performance.
Following are examples encountered in the testing and market research:
AMD states that its “Chill” software, which varies frame rate depending on the required
rendering loads, can achieve up to 30 percent energy savings (battery life
improvements) and reduced GPU temperature, while decongesting the graphics pipeline
with unneeded frames thereby improving user experience (37 percent decrease in frame
time).17 This benefit is highly game-specific and negligible for games where activity
levels are consistently high.
Systems are often “over-spec’d”, meaning that they are overpowered for the games
desired. This results in energy use that does not contribute to performance or user
experience. Better system integration will save energy and reduce system cost. In some
of the test trials (perhaps due to bottlenecks arising from poor systems integration),
under-clocking the GPU reduces power requirements while increasing graphics
performance.
Mismatches in component sizes can create bottlenecks. For example, a CPU more
powerful (and energy-using) than needed to drive the GPU will not add value. Again,
system integration is the solution to first-cost savings.
By varying refresh rates to meet the need, G-sync and FreeSync displays can provide
imagery that many gamers believe matches the smoothness and quality of that
otherwise generated by higher-power GPUs, although the team did not test this
hypothesis in the research.
17 See https://gaming.radeon.com/en/radeonsoftware/adrenalin/chill/.
49
The application of foveated rendering in VR headsets saves energy while enabling better
user experience and opening up new opportunities for in-game functionality. According
to Nvidia,18 resolution can be boosted well above normal in the central area of vision
even while saving energy overall by relaxing resolution in the periphery, where it won’t
be noticed. Meanwhile, knowing where the eye is focused will allow game developers to
key storylines to where the user is looking.
In defining the efficiency packages, the researchers looked closely at a set of non-energy
indicators. The metrics included frame rate, dropped frames, proxies for stutter and system
stress, and maximum temperatures in the GPU and CPU. In virtually every case the indicators
moved in the direction of improved user experience as efficiency was improved (Figure 21).
18 See interview of Nvidia’s Anjul Patney in Issue #5 of Green Gaming Newssee
http://greengaming.lbl.gov/newsletter/issue-5.
50
CHAPTER 6: Statewide Energy Demand and
Projections
By applying the baseline unit energy consumption values for each system (weighted by the
associated mix of user types and duty cycles) to the current installed base of equipment
represented by that system, the researchers have estimated statewide gaming energy use and
projections for the future (Figure 22). Solid lines are baseline projections, while dotted lines of
the same color represent near-term efficiency improvements for the indicated scenario (same
proportionate savings assumptions as Baseline scenario described in the text). The “Frozen
efficiency and market shares” case (dotted black line) reflects constant unit energy
consumption and unchanging proportionate mix of the various gaming products, while the
overall installed base increases. Includes energy associated with displays, local network
equipment, and external speakers, as well as networks and data centers involved in cloud-based
gaming and video streaming. In the short timeframe of this scenario, savings do not fully
reflect stock turnover of core systems and displays. None of these trajectories are intended or
presented as predictive forecasts, but, rather, exercises to help outline the bounds of how
energy demand could develop under different market and technological circumstances.
Figure 22: Enormous Potential Variations in California Computer Gaming Energy Demand Driven
by Market Structure, User Behavior, and Energy Efficiency: 2011-2021
Source: Lawrence Berkeley National Lab
51
An initial Baseline scenario looks at the effect of structural change at present efficiencies.
Providing context for alternate broader structural market developments, the team created three
alternative Baseline scenarios reflecting structural and market trends that could drive energy
use either upward or downward. The research team cast each scenario in the context of existing
and improved efficiencies. The resulting combinatory array of 8 scenarios illustrates an
envelope of possible energy futures. The savings estimates thus defined are to be regarded as
reflecting full-saturation technical opportunities considered for the particular set of measures
considered, as distinct from what might actually be achieved in practice. These improvements
could be achieved by any combination of advances emanating unilaterally from industry,
choices made independently by consumers, and/or as the result of policy initiatives interjected
by third parties. Not all potential savings measures have been assessed. Furthermore, it is
equally possible that improved efficiencies will be offset by increased workloads (for example,
for streamed VR gaming).
In this stage of the analysis, additional second-order energy use is also estimated. This includes
that of displays, household networking equipment and audio peripherals as well as upstream
network energy associated with streaming video and games. For cloud-based gaming, the
researchers also included energy used in data centers hosting gaming servers, per the method
described above.
Cloud-based gaming takes on varying importance in the scenarios. While on the one hand
power requirements of GPUs and other componentry in cloud-gaming servers may decline over
time, the base systems represent best-available current technology and this project has not
scoped possible “technology roadmaps” that could lead to new technology introduction in that
segment of the market. Nor has the team modeled the prospective rate of absorption of new
technology into widespread use. Furthermore it is equally possible that improved efficiencies
will be offset by increased workloads (for example, for streamed VR gaming). Meanwhile, the
broader Power Utilization Efficiencies assumed for the base-year data center facilities overall
are lower (more efficient) than typical practice and even projected improvements in the tier of
facilities currently hosting cloud-servers (Shehabi et al., 2016). They are also not much higher
than projected stock-averaged “Best Practices” for the short timeframes of the scenarios. Thus,
the researchers did not alter server characteristics or the PUE-1.5 assumption across scenarios.
This is an area that merits future investigation.
Games can also be downloaded from the Internet, which results in incremental energy use
associated with Internet data transmission. Insufficient data are available on the prevalence of
this activity, particularly by system type and model, to make rigorous estimates. Other
secondary sources of energy use associated with gaming include that in manufacturing and
distributing games and gaming equipment, but insufficient data were available to incorporate
that in the analysis.
A key assumption for both cloud-based gaming and game downloads is the rapidly evolving
network electricity intensity (kWh/GB of data transmitted).19 This strongly attenuates the
19 Per Aslan et al., (2018) the researchers assume a rate of 0.1449 kWh/GB in 2011, 0.0266 kWh/GB in 2016, and 0.0049
in 2021.
52
energy that would otherwise be used in association with streaming and cloud-based gaming.
The researchers assume that the streaming rate stays the same, although it could well go up
given trends and the need to transmit increasingly large amounts of data.
Past and Present Structure of Demand
Innovations in gaming technology (hardware as well as software) are progressing at a rapid
pace, consumer preferences are evolving, and the Internet is becoming inc