In mid-2006, we reached new levels of energy-efficient performance with our Intel Core2 Duo processors and Dual-Core Xeon processor 5100 series, both produced with our latest 65-nanometer (nm) silicon technology and microarchitecture . Now we’re ready to top that with the world’s first mainstream quad-core processors for both desktop and mainstream servers—Intel Core2 Quad processors, Intel Core2 Extreme quad-core processors, and others.
White PaperIntel Multi-Core ProcessorsIntel Multi-Core ProcessorsMaking the Move to Quad-Core and BeyondR.M. RamanathanIntel CorporationUntitled DocumentWhite Paper Intel Multi-Core Processors: Making the Move to Quad-Core and BeyondIntroductionOne constant in computing is that the world s hunger for faster performance isnever satisfied. Every new performance advance in processors leads to another levelof greater performance demands from businesses and consumers. Today theseperformance demands are not just for speed, but also for smaller, more powerfulmobile devices, longer battery life, quieter desktop PCs, and in the enterprise betterprice/performance per watt and lower cooling costs. People want improvements inproductivity, security, multitasking (running multiple applications simultaneously onyour computer), data protection, game performance, and many other capabilities.There s also a growing demand for more convenient form factors for the home,office, data center, and on the go.Through advances in silicon technology, microarchitecture, software, and platformtechnologies, Intel is on a fast-paced trajectory to continuously deliver new generationsof multi-core processors with the superior performance and energy-efficiency necessaryto meet these demands for years to come. A new cadence1in the microarchitecturearena (see sidebar next page), combined with Intel s ability to continue to extendMoore s Law, will enable Intel to bring new levels of performance, power savings,and computing capabilities year after year. In mid-2006, we reached new levels ofenergy-efficient performance with our Intel Core"2 Duo processors and Dual-CoreIntel Xeon processor 5100 series, both produced with our latest 65-nanometer(nm) silicon technology and microarchitecture (Intel Core" microarchitecture). Nowwe re ready to top that with the world s first mainstream quad-core processors forboth desktop and mainstream servers Intel Core"2 Quad processors, Intel Core"2Extreme quad-core processors, and others.This paper explains the advantages and challenges of multi-core processing, plusprovides a glimpse into the upcoming Intel quad-core processors and the directionin which Intel is taking multi-core processors to the future. We discuss many of thebenefits you will see as we continue to increase processor performance, energyefficiency, and capabilities.2Untitled Document3Intel Multi-Core Processors: Making the Move to Quad-Core and Beyond White PaperRedefining PerformanceFor years, Intel customers came to expect a doubling of performance every 18-24 months in accordance withMoore s Law. Most of these performance gains came fromdramatic increases in frequency (from 5 MHz to 3 GHz in theyears from 1983 to 2002) and through process technologyadvancements. Improvements also came from increases ininstructions per cycle (IPC). By 2002, however, increasingpower densities and the resultant heat began to reveal somelimitations in using predominately frequency as a way ofimproving performance. So, while Moore s Law frequencyincreases, and IPC improvements continue to play an importantrole in performance increases, new thinking is also required.The best example of this new thinking is multi-core processors.By putting multiple execution cores into a single processor(as well as continuing to increase clock frequency), Intel isable to provide even greater multiples of processing power.Using multi-core processors, Intel can dramatically increasea computer s capabilities and computing resources, providingbetter responsiveness, improving multithreaded throughput,and delivering the advantages of parallel computing toproperly threaded mainstream applications.A New Cadence forTechnological AdvancementBuilding on the foundation of Intel Core microarchitecture (introduced in 2006), Intel isestablishing a new cadence that will speed up thedelivery of products featuring superior performanceand energy-efficiency for years to come. Intel plansto deliver a new, optimized energy-efficient perfor-mance microarchitecture approximately every twoyears that supports all its process technologyadvancements. For instance, in late 2007, Intelprocess technology will transition to 45 nm andeffectively double the number of transistors in agiven die size. In 2008 Intel will follow this gain witha new microarchitecture codenamed Nehalem expected to deliver new capabilities and severalpercentage-point improvements in performance and energy-efficiency. This cycle will then move on to 32 nm and another new microarchitecturetargeted for 2010. A Fundamental Theoremof Multi-Core ProcessorsMulti-core processors take advantage of a fundamental relationship between power and frequency. By incorporatingmultiple cores, each core is able to run at a lower frequency,dividing among them the power normally given to a singlecore. The result is a big performance increase over a single-core processor. The following illustration based on our labexperiments with commonly used workloads illustrates this key advantage. Figure 1.Increasing clock frequency by 20 percent to a single core delivers a 13 percent performance gain, butrequires 73 percent greater power. Conversely, decreasingclock frequency by 20 percent reduces power usage by 49percent, but results in just a 13 percent performance loss. Under-ClockingRelative single-core frequency and Vcc1.13x0.51x1.73x0.87x1.00xOver-clocked (+20%) Max Frequency Under-clocked (-20%)PowerPerformance Figure 2.Here we add a second core on the underclockedexample in Figure 1. This results in a dual-core processorthat at 20 percent reduced clock frequency effectivelydelivers 73 percent more performance while usingapproximately the same power as a single-core processor at maximum frequency.Multi-Core Energy-Ef cient PerformanceRelative single-core frequency and Vcc1.02x1.73x1.00x1.13x1.73xOver-clocked (+20%) Max Frequency Dual-core (-20%)PowerPerformance Dual-coreUntitled DocumentWhite Paper Intel Multi-Core Processors: Making the Move to Quad-Core and Beyond4Extending the World s MostPopular Processor Architecturewith New InstructionsFrom the original Intel 8086 to the recent additionof Supplemental Streaming SIMD Extensions 3(Supplemental SSE3) found in Intel Core"2 Duoprocessors, Intel has led the charge in expandingthe capabilities of the world s most popular andbroadly used computer architecture Intel architecture.Intel will soon add new instructions enabling ourmicroprocessors across all volume market segmentsto deliver even greater performance and energyefficiency to a broad range of 32-bit and 64-bit applications.Best Energy-EfficientPerformance ProcessorTransistors in Volume" Intel Second Generation Strained SiliconTechnology increases transistor performance 10 to 15 percent without increasing leakage. " Compared to 90 nm transistor technology, Intel senhanced energy-efficient performance 65 nmtransistors provide over 20% improvement intransistor switching speed and over 30%reduction in transistor switching power " We announced and demonstrated 45 nm processtechnology first. We are on track to deliver 45 nmproducts in 2007. This fundamental relationship between power and frequencycan be effectively used to multiply the number of cores fromtwo to four, and then eightand more, to deliver continuousincreases in performance without increasing power usage.To do this though, there are many advancements that mustbe made that are only achievable by a company like Intel.These include:" Continuous advances in silicon process technology from65 nm to 45 nm and to 32 nm) to increase transistor density.In addition, Intel is committed to continuing to deliver superiorenergy-efficient performance transistors. " Enhancing the performance of each core and optimizingit for multi-core through the introduction of new advancedmicroarchitectures about every two years. " Improving the memory subsystem and optimizing dataaccess in ways that ensure data can be used as fast aspossible among all cores. This minimizes latency andimproves efficiency and speed. " Optimizing the interconnect fabric that connects the coresto improve performance between cores and memory units." Optimizing and expanding the instruction set toenhance the capabilities of Intel architecture and enablethe industry to deliver advanced applications with greaterperformance and lower power requirements. Some of theseinstructions can effectively dedicate a core to deliverspecific capabilities. " Continuing to grow Intel s commitment to developingmulti-core software tools and programs by workingclosely with developers, independent software vendors(ISVs), operating system vendors (OSVs) and academia.Through these efforts, Intel enables the industry to develop software that runs faster and better on our energy-efficient performance multi-core platforms.Untitled Document5Intel Multi-Core Processors: Making the Move to Quad-Core and Beyond White PaperIntel Achievements in Dual-Core Processing Intel first implemented multi-core processing through dual-coreprocessors across all key sectors (desktop, workstation, mobile, andmainstream server). In accordance with our new cadence for processtechnology and microarchitecture, Intel s second generation of dual-core processors, released in the third quarter of 2006, uses the newIntel Core microarchitecture. These products include Intel Core"2Duo desktop and mobile processors, and Dual-Core Intel Xeon 5100processor 5100 series for dual-processor servers. By transitioningthe majority of our volume products to Intel Core microarchitecture-based dual-core processors, Intel took the lead in performance and energy efficiency in most of these product segments. According to benchmark tests: "The Dual-Core Intel Xeon 5100 server processordeliversup to 135 percent performance improvements2and up to a40 percent reduction3in energy consumption over previousIntel server products." The Intel Core 2 Duo desktop processordelivers up to a 40percent improvement in performance and up to a 40 percentreduction in power as compared to today's high-end Intel Pentium D processor 960.4" The Intel Core 2 Duo mobile processordelivers greater than 2X CPU performance5and up to a 28 percent powerreduction6with new Intel Centrino Duo mobile technologylaptops based on the Intel Core 2 Duo processor as comparedto previous-generation Intel Centrino mobile technology-based laptops.Introducing Intel Quad-Core Technology The next milestone in multi-core processor design and performancewill be Intel s unveiling of the industry s first quad-core processors fordesktops, workstations and volume servers. Intel is the only companywith the manufacturing resources to take this next step so quickly.Intel s implementation of quad-core takes advantage of our rich historyof engineering expertise, along with our industry-leading manufacturingtechnologies and capabilities. This translates into excellent volumepricing and consistent supply. The industry will be able to make afast transition as well these quad-core processors are designed toplug into current motherboards meeting the proper thermal andelectrical specifications.7Intel Core" MicroarchitectureIntel Core" microarchitecture is the foundationfor new Intel architecture-based desktop, mobileand mainstream server multi-core processors. Thisstate-of-the-art, multi-core optimized microarchitecturedelivers a number of new and innovative featuresthat have set new standards for energy-efficientperformance. Two of these features acceleratethe execution of memory-related instructions.Intel Advanced Smart Cache Intel Advanced Smart Cache is multi-core optimizedcache that improves performance and efficiency by increasing the probability that each core of amulti-core processor can access data from a higher-performance, more-efficient cache subsystem. IntelAdvanced Smart Cache works by sharing the Level2 (L2) cache among cores so that data is stored inone place that each core can access. Sharing L2cache enables each core to dynamically use up to100 percent of available L2 cache, thus optimizingcache resources. Intel Smart Memory Access Intel Smart Memory Access improves systemperformance by optimizing available data bandwidthfrom the memory subsystem and hiding the latencyof memory accesses through two techniques: a newcapability called memory disambiguation, and aninstruction pointer-based prefetcher that fetchesmemory contents before they are requested.The Race is OnIntel plans to ship millions of multi-core processorsin 2006 and expects to exit the year shipping dual-core and quad-core processors at a rate of morethan 75 percent for its performance and mainstreamdesktop, 90 percent for its performance mobileprocessor families, and 85 percent for its servers.Intel expects to exit 2007 shipping multi-coreprocessors at a rate of 90 percent for its performancedesktop and mobile families and close to 100percent for its servers. Untitled DocumentWhite PaperIntel Multi-Core Processors: Making the Move to Quad-Core and Beyond6The Quad-Core Line UpFirst up are the Intel Core"2 Extreme quad-core processorQX6700 and the new Quad-Core Intel Xeon 5300 processor forservers. Slated for introduction in late 2006, these 65 nm quad-coreprocessors feature four complete execution cores within a singleprocessor and are based upon the revolutionary and proven Intel Core" microarchitecture. " Intel Core"2 Extreme Quad-Core Processor World s FirstQuad-Core for the Desktop.This quad-core desktop processorwill be the ultimate gaming machine and multimedia processingengine for today s growing list of threaded applications. In additionto being excellent for intensive multitasking, the Intel Core 2 Extremequad-core processor will provide impressive gaming performance,offering plenty of headroom for tomorrow s thread-intensivegames. Gamers can expect a smoother, more exciting gamingexperience through the distribution of artificial intelligence (AI),physics and rendering across four hardware threads. Ideal forprocessor-intensive, highly threaded applications, the Intel Core2 Extreme quad-core processor will be the top choice for multimediaenthusiasts, gamers, and workers in demanding multitaskingenvironments. It will feature 2.66 GHz core speed and 1066 MHzfront side bus speed. " Intel Xeon 5300 Processor Breakthrough Performancefrom the Industry s First Quad-Core Standard High-VolumeProcessor.This new quad-core processor will enable servercustomers to turbo boost their general purpose servers withbreakthrough energy-efficient performance, greater density and fewer cooling challenges. The Quad-Core Intel Xeon 5300processor provides up to 50 percent better performance8(SPECintRate) than dual-core 2-way Intel Xeon processors oncertain applications. The additional threads from quad-coretechnology and key Intel platform-level innovations deliver themost headroom for running multiple applications simultaneouslyand virtualized environments on a two-way server. The Quad-Core Intel Xeon 5300 series will feature 2.66 GHz to 1.60 GHz cores speeds, 1333 to 1066 MHz bus speeds, anda 105 watt thermal design point (TDP). A low power version(L5310) with a 50 watt TDP will be available in the first quarterof 2007. Another version will be available for single-processorservers and workstations in the same time frame. Quad-Core Intel Xeon 5300 processor platforms also includeseveral advanced capabilities:"Intel Virtualization Technology (Intel VT).This is the industry sfirst hardware-assisted technology supporting today s industry-leading virtualization software. " Fully-buffered DIMM Technology. The latest in memorytechnology, fully-buffered DIMM technology provides significantlygreater performance and capacity while improving memory reliability." Intel I/O Acceleration Technology.This unique Intel technologymoves network data more efficiently through Intel Xeon processor-based serversfor fast, scaleable, and reliable networking. Beyond Quad-Core: Tera-Scale Computing Spurred by increasing globalization, growing device intelligence,and the explosion of digital data, Intel believes the next decade sapplications will be much more computationally intensive thananything we ve seen to date. This will be the tera era an agewhen people need teraflops (a trillion floating point operations persecond) of computing power, terabits (a trillion bits per second ofcommunications bandwidth), and terabytes (1,024 gigabytes) ofdata storage to handle the information all around them.Coming in Q1 2007Intel Core"2 Quad processor Q6600" Multimedia powerhouse for demandingentertainment applications" Ideal choice for processor intensive, highly-threaded applications" 2.40 GHz core speed, 1066 MHz bus speedQuad Core Intel Xeon processor L5310 " Low Power version of 5300" 50 watt thermal design point" 1.6 GHz core speed, 1066 MHz bus speedQuad Core Intel Xeon processor 3200" For single-processor servers and workstation systemsUntitled Document7Intel Multi-Core Processors: Making the Move to Quad-Core and Beyond White PaperWith the tera era in mind, Intel researchers are today workingto shape future Intel microprocessors through the Intel Tera-scaleComputing Research Program. Intel has over 100 R&D projectsworldwide in the tera-scale area. Our researchers are addressingthe hardware and software challenges of building and programmingsystems with dozens (even hundreds) of energy-efficient coreswith sophisticated memory hierarchies to deliver the performanceand capabilities needed by these systems.Transitioning the Industry to Multi-Core Processing One immediate benefit of multi-core processors is how they improvean operating system s ability to multitask applications. For instance,say you have a virus scan running in the background while you reworking on your word-processing application. This often degradesresponsiveness so much that when you strike a key, there can be adelay before the letter actually appears on the screen. On multi-coreprocessors, the operating system can schedule the tasks in differentcores so that each task runs at full performance. Another major multi-core benefit comes from individual applicationsoptimized for multi-core processors. These applications, when properlyprogrammed, can split a task into multiple smaller tasks and run themin separate threads. For instance, a word processor can have find andreplace run as a separate thread so doing a find and replace on a bigdocument doesn t have to keep you from continuing to write or edit.In a game, a graphics algorithm needing extensive processing powercould be one thread, rendering the next scene on the fly, while another thread responds to your commands for acharacter s movements. The critical element in multi-core computing is the software. The throughput, energy efficiency, and multitasking performanceof multi-core processors will all be more fully realized whenapplication code is threaded and multi-core ready. Intel providesextensive partner programs with software developers, operatingsystem vendors, ISVs, and academia to accelerate the deliveryof dual-core and quad-core products. Intel has recently updatedthe Intel Threading Building Blocks, Intel Thread Profiler, andIntel Thread Checker tools to support our quad-core products. Intel s software products and programs include:" Intel Software Development Products These productsand tools embrace multi-core programming best practices, facilitatethreading applications, and help developers shorten the time tomarket. Products include Intel Compilers, Intel PerformanceLibraries, Intel VTune" Performance Analyzers, Intel ThreadingTools, and Intel Cluster Tools. Intel also offers software platformproducts such as Intel Platform Administration Technology,Intel Innovation Framework for EFI, and Intel Media Codecs to help speed value delivery in the platforms." Software Programs and Services Beyond softwareproducts, Intel works with leading software vendors to providetools, resources, expertise and relationships to drive threadoptimization across a wide range of applications. Through deeptechnical onsite collaborations with operating system (OS),firmware and database vendors, we enable them to fully utilizethe multi-core platforms prior to launch. Intel also works withindustry groups like ODSL and Eclipse to shape the direction ofthe open source community. In addition, Intel works with thousandsof ISVs to help them prepare and optimize their solutions fornew platform technologies and with academia in developingmulti-core curriculum. Intel software programs and servicesinclude the Intel Early Access Program, Threading ImmersionProgram, Intel Competency Centers, Intel Software College,Intel Software Network, Intel Solution Services, and Intel Capital.Progress has been fast. Microsoft Windows XP*, Windows Server*,Microsoft Vista* , some applications in Microsoft Office 2007*,various Linux* vendor offerings, and others are already threadedfor better performance with Intel multi-core processors. In addition,Intel s Software and Solutions Group (SSG) has worked with hundredsof independent software vendors (ISVs) in enabling their applicationsto achieve greater performance running on Intel multi-core processors.No one is in a better position to spearhead platform developmentthan Intel. Intel s leadership in the industry, our commitment toinvestment in research and development, our understanding of allsegments of the market, and our long history of collaboration withother industry leaders, puts us in a unique position to lead thecharge for multi-core processor-based platforms. Untitled Documentwww.intel.comEmpowering the Ecosystem Intel Capital, Intel s strategic investment program, is investing hundreds of millions of dollars in companies whose products andservices supplement Intel s own product lines and capabilities. Other ecosystem development activities include Intel s sponsorshipof the Intel Developer Forum (IDF), an annual series of worldwide conferences that provide insights into Intel's future technologydirections and enable the developer community to share their knowledge, ideas, and products. SummaryMulti-core processors are the future of computing. As the wealth and complexity of the data around us grows, multi-core processors will become increasingly important for helping run businesses, governments, our homes, and our entertainment. Multi-core processorswill empower the development of new usage models that will enable wide-ranging advances in everything from medicine to IT, as wellas revolutionize the digital office, digital home, computing on the go, and computer gaming.Think of a time a decade or so from now when the full power of high performance computing and parallel processing is available tocomputer users everywhere, and it might be possible to hold the power of a computer with hundreds of execution cores in the palmof your hand. Chances are, we ll look back and wonder how people ever managed with computers having just a single execution corein their processor. Intel quad-core processors are the next step in this process. 1. Intel CEO Paul Otellini in Spring 2006 Financial Analyst Forum2. 135 percent Performance Claim based on published results on SPECjbb2005 benchmark as of June 26, 2006. Configuration details: Dual-Core Intel Xeon Processor 2.80 GHz basedplatform details: Fujitsu Siemens Computers PRIMERGY RX300 S2 server platform: Two Dual-Core Intel Xeon processors 2.80 GHz with 2x2MB L2 cache 800 MHz system bus, 4GB DDR2, Microsoft Windows Server* 2003. Java HotSpot(TM) Server VM (build 1.5.0_06-b05). Referenced as published at 41986 bops and 41986 bops/jvm. For more informa-tion see www.spec.org/jbb2005/results/Dual-Core Intel Xeon Processor 5160 based platform details: Fujitsu Siemens BX620 S3 Server platform with two Dual-Core Intel Xeon Processor 5160, 3.00 GHz with 4M L2Cache, 1333 MHz system bus, 8GB (8x1GB) FB-DIMM memory, Windows 2003 Enterprise Edition. BEA JRockit(R) 5.0 P26.4.0. Referenced as published at 100407 bops and100407 bops/jvm. For more information see www.spec.org/jbb2005/results/3. 40 percent reduction based on Processor TDP comparison between previous generation Dual-Core Intel Xeon Processor 2.80GHz and new Dual-Core Intel Xeon Processor 5160.4. Performance based on SPECint*_rate_base2000 (2 copies) and energy efficiency based on Thermal Design Power (TDP), comparing Intel Core"2 Duo E6700 to Intel Pentium DProcessor 960. Actual performance may vary. See www.intel.com/performance for more information.5. As measured by SPEC* CPU2000* (SPECfp*_rate_base2000 and SPECint*_rate_base2000) comparing Intel Pentium M Processor 780 and 750 with Intel Core"2 Duo ProcessorT7600 and T5600. Actual performance may vary. See http://www.intel.com/performance/mobile/benchmarks.htm for important additional information. SPEC, SPECint, SPECfp,SPECrate, SPECweb, SPECjbb are trademarks of the Standard Performance Evaluation Corporation. See: http://www.spec.org for more information on the benchmarks.6. Based on power utilization measured by average power of Intel components of pre-production Intel Core"2 Duo processor-based laptop as compared to an Intel Pentium Mprocessor-based notebook. Actual performance may vary. See http://www.intel.com/technology/eep/platforms.htm for important additional information.7. Contract your motherboard vendor for more information.8. Performance may vary. Please refer www.intel.com/performance for more details.Copyright 2006 Intel Corporation. All rights reserved. Intel, Intel logo, Intel. Leap ahead., Intel. Leap ahead. logo, Intel Core 2 Duo, Pentium, Intel Core, Intel Core 2 Extreme,Centrino, VTune, and Xeon are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries.*Other names and brands may be claimed as the property of others. Printed in the United States. 0906/RMR/HBD/PDF 315451-001 USLearn MoreIntel Multi-Core www.intel.com/multi-coreNew Instructions www.intel.com/technology/architecture/new_instructions.htmIntel Architecture www.intel.com/technology/architectureTera-Scale Computing www.intel.com/technology/techresearch/terascaleSilicon Technologies www.intel.com/technology/siliconTechnology and Research at Intel www.intel.com/technology