+–– The Impact of Graphics Cards on Desktop Computer Energy Consumption September 2012 BY Pierre Delforge | NRDC Christopher Wold | CLASP Acknowledgements: Project Co-Managed by Pierre Delforge, NRDC Christopher Wold, CLASP Report Written by Pierre Delforge, NRDC Executive Summary Prepared by Christopher Wold, CLASP Methodology, Data Collection, and Analysis Conducted by Eric Wanless, Brendan Trimboli, Jeffrey Swofford, Craig Billingsley, Ecova William Westwater Report Reviewed by Jonathan Wood, Tenvic Jim McMahon, Better Climate Steven Pantano, Kathleen Callaghy, CLASP 1 Contents Executive Summary ................................................................................................................................... 4 1. Introduction .......................................................................................................................................... 6 2. Background .......................................................................................................................................... 8 2.1 Scope ................................................................................................................................................. 8 2.2 Computer Energy Use ..................................................................................................................... 8 2.3 Discrete Graphics Energy Use ....................................................................................................... 9 3. Methodology ...................................................................................................................................... 10 3.1 Overview .......................................................................................................................................... 10 3.2 System Level Testing .................................................................................................................... 10 3.3 Idle Mode Testing .......................................................................................................................... 11 3.4 Test Protocol ................................................................................................................................... 12 3.5 Discrete Graphics Card Selection ............................................................................................... 12 Selection Criteria and Rationale ..................................................................................................... 12 Selected Graphics Card Models ..................................................................................................... 13 3.6 Test Computer Configurations ..................................................................................................... 13 3.7 Power Supply Configurations ....................................................................................................... 13 4. Data Analysis & Key Findings ......................................................................................................... 15 4.1 Single Card Test Results .............................................................................................................. 15 4.2 Average by Card ............................................................................................................................ 17 4.3 Average by ECMA Category ........................................................................................................ 19 4.4 Dual Card Test Results ................................................................................................................. 19 5. Recommendations ............................................................................................................................ 22 5.1 Setting Adders for Discrete Desktop Graphics Cards .............................................................. 22 Considerations for Setting Adders .................................................................................................. 22 Use a linear regression based on frame buffer bandwidth to set adders ................................. 23 Managing the Impact of Breakthrough Innovation in Low-Power Graphics Technology ....... 26 5.2 Setting Adders for Additional Graphics Cards (Beyond the First Card) ................................ 26 5.3 Setting Adders for Discrete Notebook Graphics Cards ............................................................ 27 6. Conclusion ......................................................................................................................................... 28 2 7. Appendices ........................................................................................................................................ 29 Appendix I – Test Methodology .......................................................................................................... 29 Test Equipment ................................................................................................................................. 29 Energy Star Idle Mode Test Procedure ......................................................................................... 29 Multi-Card Test Methodology .......................................................................................................... 30 Other Test Conditions and Documentation ................................................................................... 30 Appendix II – Graphics Card Models ................................................................................................. 32 Appendix III – Test Computer Configurations .................................................................................. 34 Overview ................................................................................................................................................ 34 Goal ........................................................................................................................................................ 34 Requirements for Meeting the Goal ................................................................................................... 34 General Hardware Design ................................................................................................................... 35 Computer Construction .................................................................................................................... 35 CPU ..................................................................................................................................................... 35 Integrated Graphics .......................................................................................................................... 36 Motherboard ....................................................................................................................................... 37 Secondary Components .................................................................................................................. 38 Detailed Test Computer Configurations ........................................................................................ 38 Appendix IV – Power Supply Configurations and Efficiencies ...................................................... 40 Power Supply Models ....................................................................................................................... 40 Computer - Power Supply Combinations ...................................................................................... 41 Influence of Power Supply Efficiency on Test Results ................................................................ 42 Appendix V – Detailed Test Data and Analysis ............................................................................... 46 3 Executive Summary In 2012, the Collaborative Labeling and Appliance Standards Program (CLASP) and the Natural Resources Defense Council (NRDC), in collaboration with Ecova, initiated an innovative study designed to provide policy makers with data on discrete graphics card energy consumption in desktop computers. This data was gathered to support the establishment of effective energy consumption allowances (or “adders”) for graphics cards in the Version 6.0 ENERGY STAR computer specification, as well as in other labeling initiatives and mandatory standards that use the ENERGY STAR computer specification as a framework. Graphics card adders impact energy saving. Overly lenient graphics card adders can provide a significant excess allowance of energy consumption for the rest of the computer system, which enables less energy efficient computers to meet efficiency requirements for standards and labeling programs. On the other hand, overly stringent graphics card adders may restrict market access for efficient computers that require cards for graphics-intensive applications (e.g. computer gaming). Setting graphics card adders at appropriate levels will ensure that standards and labeling programs support the market for energy efficient computers while excluding inefficient models. Stand-alone graphics cards are typically measured and evaluated independently from the computer systems in which they are used; system-level power demand impacts are then derived by applying a power conversion factor. This traditional approach does not account for the impact of the graphics card on other components in the computer system, which may be significant in some cases. This study employed a novel approach for measuring the power impact of discrete graphics