Intel Server Hardware Roadmap And Technology Directions Jim

Intel Server Hardware Roadmap And Technology Directions Jim Fister Architecture and Planning Manager Intel Corporation james. d. fister @ intel. com

2 YEARS Intel’s Design And Process Cadence To Maintain Product Leadership Shrink/Derivative XPF 5000 Series 65 nm New Microarchitecture Core 2 Duo XPF 5100/5300 Series Shrink/Derivative Wolfdale Harpertown 45 nm New Microarchitecture NEHALEM Shrink/Derivative WESTMERE 32 nm New Microarchitecture SANDY BRIDGE Over 112 Mu 65 nm CPU shipments to date

45 nm Manufacturing Fabs D 1 D Oregon Fab 32 Arizona Fab 28 Israel Ramp in 2 H ‘ 07 Ramp in 1 H ‘ 08 Fab 32 New Mexico Ramp in 2 H ‘ 08 Three 300 mm factories are planned to be manufacturing 45 nm products by 1 H ’ 08. Four in 2 H ’ 08

® Intel Core™ Microarchitecture Wider Faster Intel Wide Dynamic Execution 33% Wider (4 -vs-3 wide) 66% bursts with Op-fusion Macrocode & Microcode Instructions per clock focus 2 x SSE Throughput via Intel Advanced Digital Media Boost (128 bit SSE) Smarter Intel Smart Memory Access Intelligent Pre-fetch Memory Disambiguation Dynamic Cache allocation Larger Energy Efficient Intelligent Power Capability Ultra-fine Clock Gating Cache power optimized Intel Advanced Smart Cache 2 x Bandwidth 2 x Size, up to 4 MB with dynamic sharing EM 64 T 64 -bit ALUs & Instruct.

Multi-Core Power Efficiency System Power Performance (measured system Watts) Lower is better (SPECjbb 2005) Same Power Envelope Outstanding Performance per Watt Higher is better 1. 5 X Socket and Platform Compatible dual-core Over 4 X single core Single Dual 5160 Quad E 5345 3. 6 GHZ 3. 0 GHZ 2. 33 GHZ Mainstream parts at 80 W, Low Voltage at 50 W Intel Core Microarchitecture

Intel Server Platform Roadmap 2007 Intel® 2 9000 Sequence Richford Platform Intel® Itanium® 2 Platform Dual-Core Itanium® 2 Processor 9000 Series Montvale Tukwila Intel® E 8870 & OEM Chipsets ` Intel® Xeon® MP 7000 Sequence Intel® Xeon® MP Platform *Dual-Core Intel® Xeon® Processor 7100 series Tigerton Dunnington Clarksboro / OEM Chipset Quad-Core Intel® Xeon® processor 5300 series Dual-Core Intel® Xeon® processor 5100 series Bensley Platform 45 n. M Quad Core Processor 45 n. M Dual Core Processor Intel® 5000 P/V Chipsets Kaylo Platform Intel® UP Server Poulson Future Chipset Caneland Platform 8501 & OEM Chipset Intel® Xeon® DP 5000 Sequence Future 2008 Itanium® Garlow Platform Quad-Core Intel® Xeon® processor 3200 series Dual-Core Intel® Xeon® processor 3000 series 45 n. M Quad Core Processor Intel® 3000/3010 Chipset Bigby P/V Chipset 45 n. M Dual Core Processor * = Intel Net. Burst® Microarchitecture Copyright © Intel Corporation, 2006. All rights reserved. Third-party marks and brands are the property of their respective owners. All products, dates, and figures are preliminary and subject to change without notice. Future Platform Future Processor Future Chipset Future DP Platform Future Processors Future Chipset Future UP Platform Future Processors Future Chipset 4+ Cores

Intel Xeon DP Platform, 2006 -2008 Platform Performance: It’s all about Bandwidth & Latency FSB Scaling 800 MHz 1067 MHz 1333 MHz Point to Point Interconnect Large Shared Caches 64 GB AM B A M B Easy Capacity Expansion 17 GB/s A M B AM B A M B A M B Energy AM B Perf A M B Local and Remote Memory Latencies Consistent AM B Central Coherency Resolution Sustained & Balanced Throughput CONSTANTLY ANALYZING THE REQUIREMENTS, THE TECHNOLOGIES, AND THE TRADEOFFS *Graphics not representative of actual die photo or relative size

Intel MP Server Platform, 2007+ Quad-Core Next Generation micro-architecture Caneland Platform Advanced power capabilities and improved cache architecture 2 X Cores Dedicated high-speed interconnects AM B AM B Increased Memory Capacity & Efficiency AM B Increases b/w, capacity & efficiency Multiple improvements contribute to platform performance AM B I/O Acceleration AM B Next generation memory technology (FB-DIMM) AM B Reduces address bus traffic AM B Chipset AM B Snoop Filter AM B Chipset snoop filter Reduced Traffic AM B Increases frequency and data bandwidth >2 X Bandwidth

Intel Quick. Data Technology Faster, scalable, more reliable I/O Open Intel Xeon Processor Opening Data Movement Acceleration Engine Intel Xeon Processor Fast 1 Intel Quick. Data Technology 2 Memory Intel Chipset Up to >2 X better throughput 1 Up to 40% lower relative CPU utilization Saleable and Secure Scales to 8+ Gb. E ports for networking Uses trusted OS TCP stack Intel or 3 rd Party Devices Source Intel, Feb ’ 06; 1. Single-Core Xeon w/4 GB RAM vs. Dual. Core Xeon 5100 w/4 GB RAM 2: Xeon 5100 with IOAT vs. Xeon 5100 without IOAT Intel I/O Acceleration Technology (IOAT) deliversbest implementation of Quick. Data Technology *Other names and brands may be claimed as the property of others

Intel Virtualization Technology Evolution VMM software evolving to use integrated hardware support Intel Virtualization Technology for I/O +ext Standard IO-device sharing for multi-context I/O devices - under definition in the PCI-SIG* IOV 2008/9 for Xeon DP/MP Intel Virtualization Tech. for Directed I/O 2007/8 for Xeon UP/DP Intel Virtualization Technology Device virtualization allows DMA remapping – the direct assignment of I/O devices to VMs. Improved security and performance Intel Hardware Virtualization support delivers a more robust and simplified virtualization capability 2005/6 Virtualization Headroom Best performance across range of workloads Proven Intel Reliability Most deployed platform, Most reliability features Broadest Software Ecosystem Broadest choice of software environments Integrated Virtualization Support *Other names and brands may be claimed as the property of others

Server Virtualization Uses R&D Server Consolidation App OS OS OS HW HW App … VMM HW Enables rapid deployment VMM HW 10: 1 in many cases Dynamic Load Balancing Disaster Recovery App … App OS OS VMM HW Upholding high-levels of business continuity Production App 1 App 2 App 3 App 4 OS OS VMM HW CPU Usage 90% CPU Usage 30% Balancing utilization with head room

2 YEARS Penryn Family Design Goals Shrink/Derivative Presler · Yonah · Dempsey 65 nm New Microarchitecture Increase performance per given clock cycle Increase processor frequencies Intel Core Microarchitecture 2 YEARS Extend energy efficiency Shrink/Derivative Penryn Family 45 nm New Microarchitecture Nehalem Shrink/Derivative Westmere 32 nm New Microarchitecture Sandy Bridge Deliver lead product for 45 nm High k + metal gate process technology Deliver optimized processors across each product segment and power envelope

45 nm Next Generation Intel Core 2 And Xeon Family Processors (Penryn) Built Upon Enhanced Intel Core Microarchitecture Greater Performance at Given Frequency AND Higher Frequencies Introduces New SSE 4 Instructions For Media/Gaming/Graphics New Levels of Energy Efficiency Larger Caches, Faster Buses Growing Performance and Energy Efficiency

Enhanced Intel Core Microarchitecture Intel Wide Dynamic Execution Fast Radix-16 Divider Enhanced Intel Virtualization Technology Intel Advanced Smart Cache Larger Caches: up to 6 MB, 12 MB Intel Smart Memory Access Split Load Cache Enhancement Higher bus speeds Intel Advanced Digital Media Boost Intelligent Power Capability Intel SSE 4 instructions Super Shuffle Engine Deep Power Down Technology Enhanced Intel Dynamic Acceleration Tech

45 nm Hi-k Intel Xeon Processors UP/DP Intel Xeon Processor >3 GHz Socket Compatible Up to 1600 MHz Bus 2 nd Generation Quad-core L 2 4 Cores − 12 MB L 2 Cache − 50/80 W/120 W TDP Dual-core L 2 − 6 MB Shared L 2 Cache − 40 W/65 W/80 W TDP MP Intel Xeon Processor Socket Compatible All products and dates are preliminary and subject to change without notice. 2 Cores

Nehalem 2 YEARS The next brand new microarchitecture Shrink/Derivative Presler · Yonah · Dempsey 65 nm New Microarchitecture Intel Core Microarchitecture Shrink/Derivative Penryn Family 45 nm New Microarchitecture Nehalem Shrink/Derivative Westmere 32 nm New Microarchitecture Sandy Bridge

Dynamic Scalability For Efficient Performance On Demand Leverages 4 issue Intel Core micro-architecturetechnology Simultaneous multi-threading Multi-level shared cache arch Performance enhanced dynamic power management Fully unlocks Intel 45 nm Hi-K silicon process benefits Nehalem 45 nm Hi-k

Design Scalability Optimizes For Each Marketing Segment Next generation platform architecture Scalable performance: 1 to 16+ threads utilizing 1 to 8+ cores Scalable and Configurable: Cache, Interconnect & memory controllers Optional High Performance Integrated graphics for client Broad family of products begin ramping production 08‘ Nehalem 45 nm Hi-k

Intel Itanium Processor Momentum Intel Itanium System Revenue vs. SPARC and Power Intel Itanium processor volume * * 2 H’ 06 shipments 2 X over 2 H’ 05 * Source: IDC Q 3’ 06 Server Tracker Application Growth Building critical mass > 6, 000 > 10, 000* +$6 B cumulative system revenue ~80% G 100 customer wins 56% 86% positive press Fujitsu/Intel marketing MOU > 2, 400 > 700 Enabled SAS, IBM Tivoli, Symantec AV, Oracle AS, Biz Objects DI

Itanium 2 Processor 9000 Series Up to 2 X 20% Lower Power 1 Better Performance 1 Up to 2. 5 x Better Performance per Watt 1 Hitachi* Cold Fusion-3 e/4 S-4 U Server (Itanium® 2 Processor 9050, 16 GB) and Intel® SR 870 BN 4 Server System (with Itanium® 2 Processor w/9 M L 3 Cache, 16 GB). Measured power. Star. CD, Workload Version: V 3. 22 (64 bit). Other names and brands may be claimed as the property of others

Dual-core Intel Itanium 2 Processor Delivers Outstanding Scalable Performance For Business-critical Workloads World record ERP performance – most SAP users supported by Server SAP* SD 2 -tier benchmark World record TPC Benchmark H* result – make smarter decisions with complex data at a better value Higher is better 1. 59 x 1. 27 x 3. 49 x 3. 5 x 64 P/ 64 C/ 128 T 64 P/ 64 C/ 64 T 64 P/ 128 C/256 T Data Source: Published results as of Dec 18, 2006. See backup for details. P = processors, C = cores, and T = threads. SAP* SD 2 -tier benchmark (Number of users) 53. 80 $/Qph. H@ 10, 000 GB 118 $/Qph. H@ 10, 000 GB 72 P/ 144 C/ 144 T 64 P/ 64 C/ 64 T 38. 98 $/Qph. H@ 10, 000 GB 64 P/ 128 C/ 128 T Data Source: Published results as of Dec 18, 2006. See backup for details. P = processors, C = cores, and T = threads. TPC Benchmark H* - 10, 000 GB Best-in class OLAP price / performance Dual-Core Itanium 9050 – Dual-Core Intel® Itanium® 2 Processor 9050 (“Montecito 1. 60 GHz”); Itanium 2 Processor 9 MB – Intel® Itanium® 2 Processor 1. 60 GHz with 9 MB L 2 cache; (“Madison 2. 66 GHz”); Power 5 – IBM System P 5* Model 595 using IBM Power* 5+ Processor 2. 30 GHz); Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate performance of Intel products as measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, visit http: //www. intel. com/performance/resources/limits. htm or call (U. S. ) 1 -800 -628 -8686 or 1 -916 -356 -3104. Copyright © 2006, Intel Corporation. * Other names and brands may be claimed as the property of others.

® ® Intel Itanium 2 Processor Momentum Continues Great Momentum World’s Top TPC-C* Result 1 2 System Revenue Exceeds $1 B >12, 000 Applications 3 Poulson New Platforms Tukwila Next Generation New, Faster High-speed Interconnect Technology Better Chipset Design Synergies Montvale Enhanced RAS and Virtualization Future Development 1 - Source: HP Integrity Superdome Itanium 2/1. 6 GHz/24 MB i. L 3 running Oracle 10 g achieved 4. 092, 799 TPC-C, $2. 93 $/tpmc http: //www. tpc. org/tpcc/results/tpcc_result_detail. asp? id=107022701, as of Feb 27, 2007 2 - Source: IDC Q 4’ 06 Server Tracker 3 - Source: As of February 2007 from Itanium Solutions Alliance http: //www. itaniumsolutionsalliance. org *Other names and brands may be claimed as the property of others 3 All dates, product features and plans are subject to change without notice.

Growing Interest In Accelerators Math-Intensive Transactions Financial, scientific, economic models Attribute: Expansive number-crunching Visualization and Media Processing Graphics, video, speech Attribute: Real-time response Embedded Content Processing Data mining, encryption, compression, XML Attribute: Fixed algorithmic functions

Accelerators Needs And Geneseo Approach Volume Graphics Mass Attach/Virtualized IO Software & Platform latencies are 100 X the physical IO latency for mostaccelerators Geneseo Approach: Extend PCIe, Fix the platform bottlenecks with improved protocols and SW architecture PCIe 2 Intelligent IO PCIe Extensions Math Acceleration 3 D Graphics/Physics Content Processing Fixed algorithm/fixed function optimized data flow, high margin FP Acceleration MP Server General Co-Processor Tightly coupled Processor Interconnect (As Announced at IDF: Xilinx/Altera FPGA on FSB) Geneseo

Broad Tools Support Compilers: C++ and Fortran Libraries: Math Kernel and Performance Primitives Tuning: VTune Threading: Checker, Profiler, and Building Blocks Clustering: Toolkit

Call To Action Understand how your hardware and software needs to support Intel’s server platforms Ensure your drivers deliver support for Intel Quickdata Technology where applicable Explore opportunities to support virtualization uses that will be deployed by IT Deliver 64 -bit support for Xeon and Itanium platform in your drivers Use Intel tools for best performance in your software development