Legal Disclaimers • INFORMATION in THIS DOCUMENT IS PROVIDED in CONNECTION with INTEL PRODUCTS

Legal Disclaimers • INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL'S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. A "Mission Critical Application" is any application in which failure of the Intel Product could result, directly or indirectly, in personal injury or death. SHOULD YOU PURCHASE OR USE INTEL'S PRODUCTS FOR ANY SUCH MISSION CRITICAL APPLICATION, YOU SHALL INDEMNIFY AND HOLD INTEL AND ITS SUBSIDIARIES, SUBCONTRACTORS AND AFFILIATES, AND THE DIRECTORS, OFFICERS, AND EMPLOYEES OF EACH, HARMLESS AGAINST ALL CLAIMS COSTS, DAMAGES, AND EXPENSES AND REASONABLE ATTORNEYS' FEES ARISING OUT OF, DIRECTLY OR INDIRECTLY, ANY CLAIM OF PRODUCT LIABILITY, PERSONAL INJURY, OR DEATH ARISING IN ANY WAY OUT OF SUCH MISSION CRITICAL APPLICATION, WHETHER OR NOT INTEL OR ITS SUBCONTRACTOR WAS NEGLIGENT IN THE DESIGN, MANUFACTURE, OR WARNING OF THE INTEL PRODUCT OR ANY OF ITS PARTS. • Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked "reserved" or "undefined". Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The information here is subject to change without notice. Do not finalize a design with this information. • The products described in this document may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request. • Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order. • Copies of documents which have an order number and are referenced in this document, or other Intel literature, may be obtained by calling 1-800-548-4725, or go to: http://www.intel.com/design/literature.htm • Intel, the Intel logo, Intel Xeon, and Xeon logos are trademarks of Intel Corporation in the U.S. and/or other countries.

1 Legal Disclaimers - Continued • Intel processor numbers are not a measure of performance. Processor numbers differentiate features within each processor family, not across different processor families: Go to: Learn About Intel® Processor Numbers http://www.intel.com/products/processor_number • Some results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. • Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. • Intel does not control or audit the design or implementation of third party benchmarks or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmarks are reported and confirm whether the referenced benchmarks are accurate and reflect performance of systems available for purchase. • Relative performance is calculated by assigning a baseline value of 1.0 to one benchmark result, and then dividing the actual benchmark result for the baseline platform into each of the specific benchmark results of each of the other platforms, and assigning them a relative performance number that correlates with the performance improvements reported. • SPEC, SPECint, SPECfp, SPECrate, SPECpower, SPECjbb, SPECompG, SPEC MPI, and SPECjEnterprise* are trademarks of the Standard Performance Evaluation Corporation. See http://www.spec.org for more information. • TPC Benchmark, TPC-C, TPC-H, and TPC-E are trademarks of the Transaction Processing Council. See http://www.tpc.org for more information. • No computer system can provide absolute reliability, availability or serviceability. Requires an Intel® Xeon® processor E7-8800/4800/2800 v2 product families or Intel® Itanium® 9500 series-based system (or follow-on generations of either.) Built-in reliability features available on select Intel® processors may require additional software, hardware, services and/or an internet connection. Results may vary depending upon configuration. Consult your system manufacturer for more details. For systems also featuring Resilient System Technologies: No computer system can provide absolute reliability, availability or serviceability. Requires an Intel® Run Sure Technology-enabled system, including an enabled Intel processor and enabled technology(ies). Built-in reliability features available on select Intel® processors may require additional software, hardware, services and/or an Internet connection. Results may vary depending upon configuration. Consult your system manufacturer for more details. For systems also featuring Resilient Memory Technologies: No computer system can provide absolute reliability, availability or serviceability. Requires an Intel® Run Sure Technology-enabled system, including an enabled Intel® processor and enabled technology(ies). built-in reliability features available on select Intel® processors may require additional software, hardware, services and/or an Internet connection. Results may vary depending upon configuration. Consult your system manufacturer for more details. 2 Risk Factors

The above statements and any others in this document that refer to plans and expectations for the third quarter, the year and the future are forward-looking statements that involve a number of risks and uncertainties. Words such as “anticipates,” “expects,” “intends,” “plans,” “believes,” “seeks,” “estimates,” “may,” “will,” “should” and their variations identify forward-looking statements. Statements that refer to or are based on projections, uncertain events or assumptions also identify forward-looking statements. Many factors could affect Intel’s actual results, and variances from Intel’s current expectations regarding such factors could cause actual results to differ materially from those expressed in these forward-looking statements. Intel presently considers the following to be the important factors that could cause actual results to differ materially from the company’s expectations. Demand could be different from Intel's expectations due to factors including changes in business and economic conditions; customer acceptance of Intel’s and competitors’ products; supply constraints and other disruptions affecting customers; changes in customer order patterns including order cancellations; and changes in the level of inventory at customers. Uncertainty in global economic and financial conditions poses a risk that consumers and businesses may defer purchases in response to negative financial events, which could negatively affect product demand and other related matters. Intel operates in intensely competitive industries that are characterized by a high percentage of costs that are fixed or difficult to reduce in the short term and product demand that is highly variable and difficult to forecast. Revenue and the gross margin percentage are affected by the timing of Intel product introductions and the demand for and market acceptance of Intel's products; actions taken by Intel's competitors, including product offerings and introductions, marketing programs and pricing pressures and Intel’s response to such actions; and Intel’s ability to respond quickly to technological developments and to incorporate new features into its products. The gross margin percentage could vary significantly from expectations based on capacity utilization; variations in inventory valuation, including variations related to the timing of qualifying products for sale; changes in revenue levels; segment product mix; the timing and execution of the manufacturing ramp and associated costs; start-up costs; excess or obsolete inventory; changes in unit costs; defects or disruptions in the supply of materials or resources; product manufacturing quality/yields; and impairments of long-lived assets, including manufacturing, assembly/test and intangible assets. Intel's results could be affected by adverse economic, social, political and physical/infrastructure conditions in countries where Intel, its customers or its suppliers operate, including military conflict and other security risks, natural disasters, infrastructure disruptions, health concerns and fluctuations in currency exchange rates. Expenses, particularly certain marketing and compensation expenses, as well as restructuring and asset impairment charges, vary depending on the level of demand for Intel's products and the level of revenue and profits. Intel’s results could be affected by the timing of closing of acquisitions and divestitures. Intel's results could be affected by adverse effects associated with product defects and errata (deviations from published specifications), and by litigation or regulatory matters involving intellectual property, stockholder, consumer, antitrust, disclosure and other issues, such as the litigation and regulatory matters described in Intel's SEC reports. An unfavorable ruling could include monetary damages or an injunction prohibiting Intel from manufacturing or selling one or more products, precluding particular business practices, impacting Intel’s ability to design its products, or requiring other remedies such as compulsory licensing of intellectual property. A detailed discussion of these and other factors that could affect Intel’s results is included in Intel’s SEC filings, including the company’s most recent reports on Form 10- Q, Form 10-K and earnings release.

Rev. 7/17/13 3 Optimization Notice

Optimization Notice Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel.

Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice.

Notice revision #20110804

4 Table of Contents

. Intel® Xeon® Processor E5-2600 v3 Product . Java, Web, Cloud Benchmark Results Family Overview & Performance Summary . Hadoop Performance . AVX & Turbo Frequency . I/O Performance . Snoop Modes & Memory Bandwidth . Intel® Xeon® Processor E5 v3 Family . SPECCPU*2006 Results Platform Solutions . Energy Efficiency Benchmark Results . Software + Hardware Upgrade – 2x2s . Virtualization Benchmark Results . Performance Tuning Recommendations . Business Processing Benchmark Results . Backup: Turbo Tables & Memory POR . High Performance Computing Benchmark Results

5 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Overview & Performance Summary

6 Intel® Xeon® Processor E5-2600 v3 Product Family

Segmented Optimized 18C (2U) 145W SKU Stack 2.3 GHz E5-2699 v3 12C 135W . 35-45 MB LLC cache - 2.5MB/core 16C (1U) 135W Advanced 2.6 GHz E5-2690 v3 . 9.6 GT/s QPI 2.3 GHz E5-2698 v3 . DDR4-2133 12C 120W . Intel® Hyper-Threading 14C (2U) 145W . 25-30 MB LLC cache -2.5MB/core 2.5 GHz . Intel® Turbo boost 2.6 GHz E5-2697 v3 . 9.6 GT/s QPI E5-2680 v3 . DDR4-2133 14C (1U) 120W 12C 120W 2.3 GHz . Intel® Hyper-Threading 2.3 GHz E5-2670 v3 E5-2695 v3 . Intel® Turbo boost 14C (1U) 120W 10C 105W 2.0 GHz E5-2683 v3 2.6 GHz E5-2660 v3 Frequency Optimized 10C 105W 2.3 GHz . ≥2.5MB/core LLC cache – See SKU 8C (20MB) 135W E5-2650 v3 . 9.6 GT/s QPI 3.2 GHz (2U) . DDR4-2133 E5-2667 v3 . Intel® Hyper-Threading 6C (20MB) 135W . Intel® Turbo boost 3.4 GHz (2U) E5-2643 v3 Standard . 2U cooling 8C 90W 4C (15MB) 135W . 15-20 MB LLC cache - 2.5MB/core 2.6 GHz E5-2640 v3 3.5GHz (2U) E5-2637 v3 Legend . 8.0 GT/s QPI . 10MB LLC cache . DDR4-1866 8C 85W . 8.0GT/s QPI High Core . Intel® Hyper-Threading 2.4 GHz E5-2630 v3 . DDR4-1866 4C (10MB) 105W Count (HCC) . Intel® Hyper-Threading 3.0GHz (1U) . Intel® Turbo boost 6C 85W . Intel® Turbo boost E5-2623 v3 Medium Core 2.4 GHz E5-2620 v3 . 1U cooling Count (MCC) Workstation Only Low Core . 25MB LLC cache Count (LCC) . 9.6 GT/s QPI 10C 2S 160W . DDR4-2133 . Intel® Hyper-Threading 3.1 GHz E5-2687W v3 Basic 6C 85W . Intel® Turbo boost . 15 MB LLC cache - 2.5MB/core 1.9 GHz E5-2609 v3 Low Power 12C 65W . 6.4 GT/s QPI . 2.5 MB/Core LLC Cache . DDR4-1600 6C 85W . E5-2650L v3: 9.6 GT/s QPI 1.8 GHz E5-2650L v3 1.6 GHz E5-2603 v3 . E5-2630L v3: 8.0 GT/s QPI . E5-2650L v3: DDR4-2133 8C 55W . E5-2630L v3: DDR4-1866 1.8 GHz E5-2630Lv3 . Intel® Hyper-Threading . Intel® Turbo boost 7 Key Differences Between E5-2600 v2 & E5-2600 v3

Xeon E5-2600 v2 Xeon E5-2600 v3 “Ivy Bridge-EP” “Haswell-EP” Core Count Up to 12 Cores Up to 18 Cores TDP & Turbo Freq Frequency TDP & Turbo Frequencies AVX & AVX Turbo Freq Intel® AVX Intel® AVX 2 AVX Support 8 DP Flops/Clock 16 DP Flops/Clock

4xDDR3 channels 4xDDR4 channels Memory Type RDIMM, UDIMM, LRDIMM RDIMM, LRDIMM

Memory Frequency RDIMM: 2133 (1DPC), 1866 (2DPC), 1600 1866 (1DPC), 1600, 1333, 1033 (MHz) LRDIMM: 2133 (1&2DPC), 1600

QPI Speed Up to 8.0 GT/s Up to 9.6 GT/s Up to 145W Server, 160W Workstation TDP Up to 130W Server, 150W Increase due to Integrated Voltage Workstation Regulator Per-core P-states Power Same P-states for all cores Independent uncore frequency scaling Management Same core & uncore frequency Energy Efficient Turbo

8 Intel® Xeon® Processor E5-2600 v3 Product Family Generational Performance Summary

Intel® Xeon® Processor E5-2699 v3 (18C, 2.3GHz, 145W) vs. Intel® Xeon® Processor E5-2697 v2 (12C, 2.7GHz, 130W) 3.5 3.02 3 2.45 2.5 1.91 1.97 2 1.67 1.70 rformance 1.60 1.60 1.38 1.42 1.44 1.5 1.20 1.24 1.28 1.27 1 Relative Pe Relative 0.5 Intel® Xeon® Processor E5-2699 v3

0 mweb 2-Tier TPC*-H (Triad) Linpack SAP SD* STREAM sc2013 Baseline SPECvirt*_ ssj2008+ base2007 E5-2697 v2 E5-2697 MultiJVM MultiJVM max-jOPs base2006 base2006 VMmark* 2.5 VMmark* SPECmpiM*_ base2012++ memcached* critical-jOPs SPECompG*_ SPECpower*_ SPECfp*_rate_ SPECjbb2013* SPECjbb2013* SPECint*_rate_

+Baseline is E5-2660 v2 (10C, 2.2GHz, 95W) ++New configuration is E5-2690 v3 (12C, 2.6GHz, 135W) Source as of Sept 8 2014, see next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced 9 benchmark data are accurate and reflect performance of systems available for purchase. *Other names and brands may be claimed as the property of others Benchmark configurations for Generational Performance Summary (1 of 2) 1. SGI Rackable C2112-4RP4 platform with two Intel Xeon Processor E5-2697 v2 (12 cores), 24 MPI ranks, Intel Compiler version 14.0.0.051, SPECmpiM_base2007 score: 6.51. Source. Intel Corporation Endeavor with two Intel Xeon processor E5-2697 v3 (14 cores), 28 MPI ranks, Intel Compiler version 14.0.3.174, SPECmpiM_base2007 score: 7.85. Source as of September 25, 2014.

2. Dell Inc. PowerEdge T620 platform with two Intel Xeon Processor E5-2660 v2, IBM J9 VM, 8,533 overall ssj_ops/watt. Source. Sugon I620-G20 platform with two Intel Xeon Processor E5-2699 v3, 8x16GB DDR4-2133 DR-RDIMM, Microsoft Windows Server 2008 R2, IBM J9 VM, 10,599 overall ssj_ops/watt. Source, results submitted as of Sept. 8 2014.

3. Cisco UCS C240 M3 Server platform with two Intel Xeon E5-2690 v2, TPC-H performance: 304,361 QphH@1000GB. Source. HP ProLiant DL380 Gen9 platform with two Intel Xeon E5-2699 v3, TPC-H performance: 379,643 QphH@1000GB. Source.

4. Supermicro SuperServer 6027AX-TRF platform with two Intel Xeon Processor E5-2697 v2, Intel Compiler version 14.0.0.080, SPECfp_rate_base2006 score: 714. Source. Huawei RH2288 V3 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 6.5, Intel Compiler version 14.0.0.080, SPECfp_rate_base2006 score: 911. Source as of September 25, 2014.

5. Supermicro SuperServer 6027AX-TRF platform with two Intel Xeon Processor E5-2697 v2, Intel Compiler version 14.0.0.080, SPECint_rate_base2006 score: 981. Source. HP ProLiant DL360 Gen9 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 7, Intel Compiler version 14.0.0.080, SPECint_rate_base2006 score: 1,360. Source as of September 25, 2014.

6. Cisco Systems Cisco UCS C240 M3 platform with two Intel Xeon Processor E5-2697 v2 (12 cores), 48 OpenMP threads, SPECompG_base2012 score: 6.79. Source. Cisco Systems Cisco UCS C220 M4 platform with two Intel Xeon Processor E5-2699 v3 (18 cores), 72 OpenMP threads, SPECompG_base2012 score: 9.67. Source.

7. Source as of August 2014 on STREAM (triad), stream_omp v5.4 with Intel Compiler version 14.0.3.174: Intel® Server Board S2600CP with two Intel® Xeon® Processor E5-2697 v2, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, 24x16GB DDR3-1066 DR-RDIMM, score: 58.9 GB/sec. New Configuration: Intel® Server System R2208WTTYS with two Intel® Xeon® Processor E5-2690 v3, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, Cluster on Die mode, 24x16GB DDR4- 1600 DR-RDIMM, score: 85.2 GB/sec.

8. HP ProLiant DL380p Gen8 platform with two Intel Xeon E5-2697 v2, VMWare ESXi 5.1.0 U1, VMmark v2.5.1 score: 16.54 @ 14 tiles. Source. Fujitsu PRIMERGY RX2540 M1 platform with two Intel Xeon Processor E5-2699 v3, VMware ESXi 5.5.0 U2, VMmark v2.5.2 score: 26.48 @ 22 tiles. Source. 10 Benchmark configurations for Generational Performance Summary (2 of 2) 9. Dell PowerEdge R720 platform with two Intel Xeon Processor E5-2697 v2 (12 cores, 24 threads), Red Hat Enterprise Linux 6.5, SAP ASE 16, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 10,253. Source. Dell PowerEdge R730 platform with two Intel Xeon Processor E5-2699 v3 (18 cores, 36 threads), Red Hat Enterprise Linux 7, SAP ASE 16, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 16,500. Source as of September 11, 2014.

10. Intel internal measurements using mweb on platform with two E5-2697 v2, 8x8GB DDR3-1866, HT, Turbo, Prefetchers NUMA enabled, 10GbE Niantic, OS: Fedora 18, Nginx 1.4.4, PHP 5.4.16, Memcached 1.47. Platform with two E5 v3 , HT, Turbo, Prefetchers enabled, NUMA & ES mode, 8x8GB DDR4-2133, BIOS 25R02, HT, Turbo, Prefetchers NUMA enabled, 10GbE Niantic, OS: Fedora 18, Nginx 1.4.4, PHP 5.4.16, Memcached 1.47.

11. IBM System x3650 M4 platform with two Intel Xeon Processor E5-2697 v2, SPECvirt_sc2013 947.0 @ 53 VMs. Source. Hewlett-Packard Company ProLiant DL360 Gen9 with two Intel Xeon Processor E5-2699 v3, SPECvirt_sc2013 1614 @ 95 VMs. Source.

12. Source as of August 2014 on Linpack*. Baseline configuration: Intel® Server Board S2600CP with two Intel® Xeon® Processor E5-2697 v2, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, 8x8GB DDR3-1866, RHEL* 6.3, Intel® MKL 11.0.5, score: 528 GFlops. New configuration: Intel® Server System R2208WTTYS with two Intel® Xeon® Processor E5-2699 v3, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, 8x16GB DDR4-2133, RHEL* 6.4, Intel® MKL 11.1.1, score: 1,012 GFlops.

13. Source as of June 2014 TR#1427: Intel internal measurements on platform with two E5-2697 v2, 8x8GB DDR3-1866, HT, Turbo, Prefetchers NUMA enabled, RHEL 6.5, Memcached 1.6 Bag LRU, 2x10GbE Niantic, Score: 4.2M requests/sec. Platform with two E5-2699 v3, HT, Turbo, Prefetchers enabled, NUMA & ES mode, 8x8GB DDR4-2133, RHEL 6.5, BIOS 25R02, Memcached 1.6 Bag LRU, 2x10GbE Niantic, Score: 8.28M requests/sec.

14. Hewlett-Packard Company HP ProLiant ML350p Gen8 platform with two Intel Xeon Processor E5-2697 v2, Oracle Java Standard Edition 7 update 40, 57,594 SPECjbb2013- MultiJVM max-jOPS, 32,103 SPECjbb2013-MultiJVM critical-jOPS. Source. Cisco UCS C220 M4 platform with two Intel Xeon Processor E5-2699 v3, Oracle Java Standard Edition 8 update 20, 160,283 SPECjbb2013-MultiJVM max-jOPS, 58,478 SPECjbb2013-MultiJVM critical-jOPS. Source.

15. Intel® Xeon® E5 v2 Family Baseline: Cisco Systems Cisco UCS C240 M3 platform with two Intel Xeon Processor E5-2697 v2, Oracle Java Standard Edition 7 update 45, 63,079 SPECjbb2013-MultiJVM max-jOPS , 23,797 SPECjbb2013-MultiJVM critical-jOPS. Source. Hewlett-Packard Company HP ProLiant ML350 Gen9 platform with two Intel Xeon Processor E5-2699 v3, Oracle Java Standard Edition 8 update 11, 190,674 SPECjbb2013-MultiJVM max-jOPS, 47,139 SPECjbb2013-MultiJVM critical-jOPS. Source.

11 Feature Comparison across Intel® Xeon® Generations

Intel® Xeon® Processor X5600 Intel® Xeon® Processor E5-2600 Intel® Xeon® Processor E5-2600 v2 Intel® Xeon® Processor E5-2600 v3 Series (Westmere-EP) Product Family (Sandy Bridge-EP) Product Family (Ivy Bridge-EP ) Product Family (Haswell-EP) Essentials Launch Date Q1'11 Q1'12 Q3'13 Q3'14 Maximum # of Cores 6 8 12 18 Maximum # of Threads 12 16 24 36 Last Level Cache (LLC) Up to 12 MB Up to 20 MB Up to 30 MB Up to 45 MB Maximum QPI Bus Speed 6.4 GT/s 8 GT/s 8 GT/s 9.6 GT/s Instruction Set Extensions SSE4.2 Intel® AVX Intel® AVX Intel® AVX 2 Intel Process Technology 32 nm 32 nm 22 nm 22 nm Intel® Turbo Boost Technology 1.0 2.0 2.0 2.0 Same P-States for All Cores Same P-States for All Cores Same P-States for All Cores Per Core P-States Power Management Fixed Uncore Frequency Same Core and Uncore Frequency Same Core and Uncore Frequency Independent Uncore Frequency Scaling Memory Specifications Max Memory Size per Socket 288 GB 384 GB 768 GB 768 GB DDR3 DDR3 DDR4 DDR3 800/1066/1333/1600 800/1066/1333/1600/1866 1600/1866/2133 Memory Types 800/1066/1333 RDIMM/UDIMMs RDIMM/UDIMM RDIMM RDIMM/UDIMM Quad Rank LRDIMM Quad Rank LRDIMM Quad Rank LRDIMM # of Memory Channels 3 4 4 4 Max # of DIMMs/Channel 3 3 3 3 Max # of DIMMs/Socket 9 12 12 12 Theoretical Max Memory Bandwidth 32 GB/s 51.2 GB/s 59.7 GB/s 68.2 GB/s per Socket Expansion Options PCI Express Revision 2.0 (in Chipset) 3.0 3.0 3.0 Max # of PCI Express Lanes N/A 40 40 40 x4, x8, x16 x4, x8, x16 PCI Express Configurations N/A x4, x8, x16 x16 Non-Transparent Bridge x16 Non-Transparent Bridge

All comparisons based on a single socket. 12 Feature Comparison across Intel® Xeon® Generations

Advanced Technologies Thermal Monitoring Technologies N/A √ √ √ Intel® Integrated I/O N/A √ √ √ Intel® Data Direct I/O N/A √ √ √ Integrated Voltage Regulator (IVR) N/A N/A N/A √ Intel® Virtualization Technology Cache QoS Monitoring Intel® Virtualization Technology (VT-x) √ √ APIC Virtualization EPT A/D Bits Intel® Data Protection Technology

Intel® Advanced Encryption Standard √ √ √ AVX2 enhancements to AES-NI New Instructions (AES-NI)

Secure Key N/A N/A √ √ Intel® Platform Protection Technology

Greater hypervisor and OS support Greater hypervisor and OS support Intel® Trusted Execution Technology (TXT) √ SMM External Call Trap (SECT) Launch control policy enhancements Launch control policy enhancements

OS Guard N/A N/A √ √

13 Feature Glossary Feature Benefit Instruction Set Extensions SSE 4.2 8 Single Precision (SP) Flops/Clock 4 Double Precision (DP) Flops/Clock Cryptography acceleration instructions Intel® AVX 16 SP Flops/Clock 8 DP Flops/Clock Floating point vectors increase from 128 bit to 256 bit Intel® AVX2 32 SP Flops/Clock 16 DP Flops/Clock Integer Vectors Increase from 128 bit to 256 bit Floating-Point Fused Multiply Add (FMA) Gather Instructions Intel® Turbo Boost Technology Version 1.0 Turbo up to TDP Limit Version 2.0 Turbo above TDP limit for short burst (<10 sec) Algorithm enhancements Advanced Technologies Intel® Integrated I/O Integrates PCIe 3.0 interface on processor Intel® Data Direct I/O I/O Device can read and write directly to Cache instead of main memory Increases I/O Performance Integrated Voltage Regulator (IVR) IVR integrates legacy power delivery onto processor package/die IVR enables power Management benefits Simplified platform power design

14 Feature Glossary

Feature Benefit Intel® Virtualization Technology (VT-x) APIC Virtualization (APICv) Reduce unique overheads of virtualization VMCS Shadowing Efficient Nested VMMs Cache QoS Monitoring Monitoring the cache space utilization per VM EPT A/D Bits Enables efficient Live Migration and Fault Tolerance Trusted Execution Technology SMM External Call Trap (SECT) Enhanced protection from privilege escalation attacks Intel® Data Protection Technology AES New Instructions (AES-NI) Improve the speed of applications performing encryption and decryption AVX2 enhancements to AES-NI AVX2 (256-bit integer, better bit Manipulation, Permute Granularity) and 4th ALU speed up crypto algorithms. RORX accelerate hash algorithms. New instruction (MULX) improves asymmetric crypto—ease more crypto challenges. Enhancements and extensions for symmetric cryptography –Intel® AES-NI and beyond Secure Key Used for generating high-quality keys for cryptographic protocols Intel® Platform Protection Technology Intel® Trusted Execution Technology Protects against attacks toward hypervisor and BIOS, firmware, and other software components. (TXT) OS Guard Protects operating system from escalation-of-privilege attacks.

15 Intel® Xeon® Processor E5-2600 v3 Product Family 4-Year Refresh Performance Summary

Intel® Xeon® E5-2699 v3 (18C, 2.3GHz, 145W) vs. X5690 (6C, 3.46GHz, 130W) 10 8.7 9 8 7 6.3 6

Relative Performance Relative 5 4 3.1 3.2 3.2 3.3 3.3 3.3 3 minary 2 Intel® Xeon® Processor E5-2699 v3 – 2014 Launch Preli 1 0 X5690 SPECpower*_ STREAM SPECint*_rate_ SAP SD* VMmark* 2.x SPECfp*_rate_ Linpack SPECjbb2013* 2011 Launch ssj2008+ (Triad) base2006 2-Tier base2006 MultiJVM max-jOPs +Baseline is X5675 (6C, 3.06GHz, 95W) Source as of Sept 8 2014, see next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced 16 benchmark data are accurate and reflect performance of systems available for purchase. *Other names and brands may be claimed as the property of others Benchmark Configuration Details for 4-Year Refresh Performance Summary 1. HP ProLiant DL360 G7* with two Intel® Xeon® Processor X5675, Source, Score: 3,329 overall ssj_ops/watt. Sugon I620-G20 platform with two Intel Xeon Processor E5-2699 v3, 8x16GB DDR4-2133 DR-RDIMM, Microsoft Windows Server 2008 R2, IBM J9 VM, 10,599 overall ssj_ops/watt. Source, results submitted as of Sept. 8 2014.

2. Source as of August 2014 TR#3044 on STREAM (triad), stream_omp v5.4: Supermicro X8DTN+ platform with two Intel® Xeon® Processor X5680, 18x8GB DDR3-800, Intel® Compiler version 12.0.1.107, score: 26.5 GB/sec. New Configuration: Intel® Server System R2208WTTYS with two Intel® Xeon® Processor E5-2699 v3, 24x16GB DDR4-2133 @ 1600MHz DR-RDIMM, Intel Compiler version 14.0.3.174, score: 85.2 GB/sec.

3. Fujitsu platform with two Intel Xeon processor X5690 (6 cores, 12 threads), SuSE Linux Enterprise Server 11 SP1, MaxDB 7.8, SAP enhancement package 4 for SAP ERP 6.0, number of SAP SD benchmark users: 4875. Source: Certification #2011028. Dell PowerEdge R730 platform with two Intel Xeon Processor E5-2699 v3 (18 cores, 36 threads), Red Hat Enterprise Linux 7, SAP ASE 16, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 16,500. Source as of September 11, 2014.

4. Fujitsu PRIMENERGY RX300 S6 with two Intel® Xeon® Processor X5690, VMware ESXi 4.1 U1, VMmark* v2.1.1 score: 7.59 @ 7 tiles, source. Fujitsu PRIMERGY RX2540 M1 platform with two Intel Xeon Processor E5-2699 v3, VMware ESXi 5.5.0 U2, VMmark v2.5.2 score: 26.48 @ 22 tiles. Source.

5. 2-socket Intel® Xeon® processor X5690 platform, 48GB memory, RHEL 6.2. Compiler version: Intel Compiler 12.1.0. Source. Score: 425. HP ProLiant DL360 Gen9 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 7, Intel Compiler version 14.0.0.080, SPECint_rate_base2006 score: 1,360. Source as of September 25, 2014.

6. 2-socket Intel® Xeon® processor X5690 platform, 48 GB memory, SLES11* SP1, Intel Compiler 12.1.0. Source. Score: 271. Huawei RH2288 V3 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 6.5, Intel Compiler version 14.0.0.080, SPECfp_rate_base2006 score: 911. Source as of September 25, 2014.

7. Source as of August 2014 on Linpack*. Supermicro* X8DTN+ with two Intel® Xeon® processor X5690, RHEL* 6.1, 12x4GB DDR3-1333, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, SMP Linpack 10.3.5. Intel internal measurements TR#1236. Score: 159.36 GFlops. New configuration: Intel® Server System R2208WTTYS with two Intel® Xeon® Processor E5-2699 v3, Intel® HT Technology disabled, Intel® Turbo Boost Technology enabled, 8x16GB DDR4-2133, RHEL* 6.4, Intel® MKL 11.1.1. Intel internal measurements TR#3034. Score: 1,012 GFlops.

8. Supermicro* X8DTN+ with two Intel® Xeon® Processor X5690, Oracle Java* Standard Edition 7 Update 11, source. Score: 21,709 SPECjbb*2013-MultiJVM max-jOPs, 3,587 SPECjbb*2013-MultiJVM critical-jOPS. Hewlett-Packard Company HP ProLiant ML350 Gen9 platform with two Intel Xeon Processor E5-2699 v3, Oracle Java Standard Edition 8 update 11, 190,674 SPECjbb2013-MultiJVM max-jOPS, 47,139 SPECjbb2013-MultiJVM critical-jOPS. Source. 17 Intel® Xeon® Processor E5-2600 v2 Product Family Historical 2S Integer Throughput Performance Up to Higher is better 1600 22X 1400 1200 1000 800 600 400

performance 200 0 SPECint*_rate_base2006 SPECint*_rate_base2006 2006 2007 2007 2008 2009 2010 2012 2013 2014 Xeon 5160 X5272 X5365 X5492 X5570 X5680 E5-2690 E5-2697 v2 E5-2699 v3 (2C, (2C, 3.4GHz, (4C, (4C, 3.4GHz, (4C, 2.93GHz, (6C, 3.33GHz, (8C, 2.9GHz, (12C, 2.7GHz, (18C, 2.3GHz, 3GHz, 80W) 80W) 3GHz, 150W) 150W) 95W) 130W) 135W) 130W) 145W)

Up to 22x increase in integer throughput on Intel® Xeon® since 2006

Historical 2S Energy Efficient Performance Up to 27x 12,000 Higher is better 10,000

8,000 /watt) 6,000 *_ssj2008 *_ssj2008

ssj_ops 4,000

2,000 SPECpower (Overall 0 2006 2007 2007 2008 2009 2010 2012 2013 2014 Xeon 5160 X5272 L5335 L5430 X5570 X5675 E5-2660 E5-2660 v2 E5-2699 v3 (2C, (2C, 3.4GHz, (4C, (4C, 2.66GHz, (4C, 2.93GHz, (6C, 3.06GHz, (8C, 2.2GHz, (10C, 2.2GHz, (18C, 2.3GHz, 3GHz, 80W) 80W) 2GHz, 50W) 50W) 95W) 95W) 95W) 95W) 145W)

Up to 27x increase in energy efficiency on Intel® Xeon® since 2006

Intel® Xeon® Processor E5-2600 v3 Product Family AVX & Turbo Frequency

20 *Intel® AVX refers to Intel® AVX, Intel® AVX2 or Intel® AVX Intel® or AVX2 Intel® AVX, toIntel® refers AVX *Intel® . . Intel for Xeon® for will be defined frequencies turbobase & AVX dependency, separate workload Due to depends on: frequency achieved of turbo Amount . and processor temperature and processor Type ® Turbo Boost Technology 2.0Boost Technology Turbo ®

Frequency of processors starting with E5 v3 product family with E5v3product starting processors Generations Previous

AVX/Rated Base consumption, & power current estimated active cores, number of workload, AVX/Rated Turbo E5 v3& Future Generations Rated Base Rated Turbo AVX Base -

512 AVX Turbo . . . Additional Resources: Performance Performance on Intel Xeon Processors Using Intel AVX to Achieve Maximum Frequency Upside Intel® Xeon® Turbo Boost Opportunistic with Intel AVX Whitepaper and - Optimize Performance Intel® AVX Intel® * 21 How does frequency on HSW cores change with AVX workloads? . Core detects presence of AVX instructions . AVX instructions draw more current & higher voltage is needed to sustain operating conditions . Core signals to Power Control Unit (PCU) to provide additional voltage & core slows the execution of AVX instructions . Need to maintain TDP limits, so increasing voltage may cause frequency drop . Amount of frequency drop will depend on workload power & AVX frequency limits . PCU signals that the voltage has been adjusted & core returns to full execution throughput . PCU returns to regular (non-AVX) operating mode 1ms after AVX instructions are completed

22 Which workloads will run at what frequency? Frequency range of E5-2699 v3

Max All Core Turbo Frequency (Non-AVX) 2.8 2.7

Expected frequencies for 2.6 AVX Max All Core Turbo Frequency non-AVX workloads 2.5 2.4

Base Frequency (Non-AVX) 2.3 Expected frequencies for 2.2 most AVX Expected workloads frequencies for 2.1 workloads with heavy AVX usage 2.0

1.9 AVX Base Frequency

23 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Snoop Modes & Memory Bandwidth

24 Intel® Xeon® Processor E5-2600 v3 Product Family Die Configurations

14-18 Core (HCC) 10-12 Core (MCC) 4-8 Core (LCC)

Not representative of actual die-sizes, orientation and layouts – for informational use only. . HCC consists of 4 columns and 2 Memory Controllers. . MCC consists of 3 columns 4 (core + LLC slices) and 2 Memory Controllers. . LCC consists of 2 columns 4 (core + LLC slices) and single Memory Controller.

25 Intel® Xeon® Processor E5-2600 v3 Product Family Snoop Modes

Each mode is configurable through BIOS settings . Memory bandwidth & latency . Early Snoop Mode tradeoffs will vary across the . Intel’s BIOS default for HSW-EP 3 die configurations for each . Same mode available on SNB-EP snoop mode

. Home Snoop Mode . Intel recommends exposing . Same mode available on IVB-EP* all snoop modes as BIOS . Cluster on Die Mode options to the user . New mode introduced on HSW-EP

*Home Snoop mode is available on IVB-EP but is not the default setting 26 Intel® Xeon® Processor E5-2600 v3 Product Family Snoop Modes Supported in 2S Configurations

Early Snoop Home Snoop Cluster on Die (Default for HSW-EP)

New for Previously available on E5-2600 (SNB) E5-2600 v2 (IVB)* E5-2600 v3

Check Directory Cache, Snoop sent by Caching Agent Home Agent then Home Agent

NUMA workloads that Latency sensitive Highly NUMA Best used for need max local & workloads optimized workloads remote bandwidth SPECCPU*_rate2006, TPC*-E, SPECCPU*2006 SPECjEnterprise*2010, Benchmarks TPC*-C (speed) SPECpower*_ssj2008, SAP* SD

*Home Snoop mode is available on E5-2600 v2 but is not the default setting 27 Cluster on Die (COD) Mode

COD Mode for 18C HSW-EP . Supported on 2S HSW-EP SKUs with 2 Cluster0 QPI Cluster1 Home Agents (10+ cores) 0/1 IIO

Cbo Core . Targeted at NUMA workloads where Sbo LLC

latency is more important than sharing Cbo Cbo Cbo Cbo Core Core Core Core data across Caching Agents (Cbo) LLC LLC LLC LLC Cbo Cbo Cbo Cbo Core Core Core Core . Reduces average LLC hit and local memory LLC LLC LLC LLC latencies Cbo Cbo Cbo Cbo Core Core Core Core LLC . HA mostly sees requests from reduced set LLC LLC LLC Cbo Cbo Cbo Cbo Core of threads which can lead to higher Core Core Core LLC LLC LLC LLC memory bandwidth Cbo Core Sbo LLC . OS/VMM own NUMA and process affinity HA0 HA1 decisions

28 Cluster on Die (COD) Mode

COD Mode for 14C HSW-EP COD Mode for 12C HSW-EP Cluster0 Cluster1 QPI IIO 0/1 QPI IIO Sbo 0/1 Sbo Cbo Cbo Core Core Cbo Cbo Cbo LLC LLC Core Core Core LLC LLC LLC Cbo Cbo Cbo Cbo Core Core Core Core Cbo Cbo Cbo LLC LLC LLC LLC Core Core Core LLC LLC LLC Cbo Cbo Cbo Cbo Cbo Cbo Cbo Core Core Core Core Core Core Core LLC LLC LLC LLC LLC LLC LLC Cbo Cbo Cbo Cbo Cbo Cbo Cbo Core Core Core Core Core Core Core LLC LLC LLC LLC LLC LLC LLC Sbo Sbo HA0 HA1 HA0 HA1

29 Cluster on Die (COD) Mode

Snoops are handled by: . Directory Cache on die . New feature . 14KB cache in each HA that holds the directory state of migratory cache lines . Stores 8-bit vector tracking which CA may have a copy of the cache line . Directory bits in memory . Used if there is a Directory Cache miss . 2-bits stored with each cache line: Local/Invalid, SnoopAll, Shared . Reduces amount of snoops sent from HA due to tracking of shared cache lines

Best for highly NUMA optimized workloads

30 Intel® Memory Latency Checker

. Integrates functionality of internal Intel tools used to measure idle & loaded memory latencies and bandwidth . Latest release (MLC v2.1) automatically identifies system topology and eliminates the need for many of the parameters that were required in the previous versions . Publicly available at: http://www.intel.com/software/mlc MLC measures: . Idle latency per socket . Peak memory bandwidth of varying amounts of reads/writes to local memory . Memory bandwidth per socket . Latency at different bandwidth points

31 Intel® Xeon® Processor E5-2600 v3 Product Family Memory Read Latency & Bandwidth

245 NUMA HS NUMA ES NUMA COD IVT 225 Up to 14% 100% local memory read increase in 205 Sequential traffic bandwidth 185

165

145

Latency(ns) 125

105

65 0 20,000 40,000 60,000 80,000 100,000 120,000 140,000 Mem BW (MB/s) Source as of 17 Dec 2013: Intel internal measurements on platform with two E5-26xx v3 (14C, 2.7GHz, 145W), Turbo disabled, 8x16GB DDR4-2133, RHEL 6.3. Platform with two E5-2697 v2, Turbo enabled, 8x16GB DDR3-1866, RHEL 6.3. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 32 Intel® Xeon® Processor E5-2600 v3 Product Family Idle Memory Read Latency

NUMA - Local NUMA - Min Remote NUMA - Max Remote UMA - Min UMA - Max 220 Lower is better 205

) 190 175 160 145 130

Latency(ns 115 100 85 70 IVB-EP HSW-EP HSW-EP HSW-EP DDR3-1866 DDR4-2133 DDR4-2133 DDR4-2133 HS Mode ES Mode COD Mode

Source as of 17 Nov 2013: Intel internal measurements using Intel® Memory Latency Checker on platform with two E5-26xx v3 (14C, 2.7GHz/3.0GHz uncore), HT & Turbo disabled, EPB-Performance mode, 8x16GB DDR4-2133, RHEL 6.3. Platform with two E5-2697 v2 (12C, 2.7GHz), HT disabled, Turbo enabled, 8x16GB DDR3-1866, RHEL 6.3. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the 33 property of others. Snoop Mode Characteristics of HCC & MCC SKUs (10-18C)

For memory bandwidth, higher is better. For memory latency, lower is better Relative Snoop Mode Performance Configured as NUMA

Performance Metric Early Snoop COD Home Snoop LLC Hit Latency Low Lowest Low Local Memory Latency Medium+ Low High+ Remote Memory Latency Lowest Low-High* Low Local Memory Bandwidth High Highest High Remote Memory Bandwidth Medium Medium High Relative Snoop Mode Performance Configured as UMA

Performance Metric Early Snoop COD Home Snoop

Memory Latency Low Not an advised Low Memory Bandwidth High configuration High

*Depends on the directory state. Clean directory – low latency ; Dirty directory – high latency. +Local latencies are snoop bound. Source Intel Estimates based on internal testing. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. 34 For more information go to http://www.intel.com/performance Snoop Mode Characteristics of LCC SKUs (4-8 cores) For memory bandwidth, higher is better. For memory latency, lower is better Relative Snoop Mode Performance Configured as NUMA

Performance Metric Early Snoop COD Home Snoop

LLC Hit Latency Low Low

Local Memory Latency Lowest Low Not supported on Remote Memory Latency Lowest Low Core Count Low SKUs Local Memory Bandwidth High High

Remote Memory Bandwidth Low* High

Relative Snoop Mode Performance Configured as UMA

Performance Metric Early Snoop COD Home Snoop Memory Latency Lowest Not supported on Low Low Core Count Memory Bandwidth High SKUs High * Low remote bandwidth in NUMA & ES mode is expected because ES mode is designed to favor local traffic. Source Intel Estimates based on internal testing. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. 35 For more information go to http://www.intel.com/performance Memory Frequencies & Advantages of DDR4

X5600 Series E5-2600 v2 E5-2600 v3 DDR3 DDR3 DDR4 3 memory channels/socket 4 memory channels/socket 4 memory channels/socket 2500 2133 2000 1866 1866 1600 1600 1500 1333 1066 1066

requency (MHz) 1000 800

500 DIMM F

0 1 DIMM/Channel 2 DIMM/Channel 3 DIMM/Channel DDR4 enables faster memory speeds at larger memory capacities for Intel® E5 v3 DIMM frequency shown for RDIMM configurations. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 36 Intel® Xeon® Processor E5-2600 v3 Product Family More details in Grantley Platform Memory Config Latency & Bandwidth DIMM Comparison Guide Local Memory Idle Latency 100% Local Read Bandwidth Lower is better Higher is better 82 66 81.4 64.2 64 63.3 81 62.7 80.1 62 80 79.5 79.3 60 79 58 78.2 55.8 55.5 56 Latency Latency (ns) 78

Bandwidth (GB/sec) 54 77 52

76 50 SR RDIMM SR RDIMM DR RDIMM DR RDIMM QR LR-DIMM SR RDIMM SR RDIMM DR RDIMM DR RDIMM QR LR-DIMM 8GBx4 8GBx4 16GBx4 16GBx4 32GBx4 8GBx4 8GBx4 16GBx4 16GBx4 32GBx4 1DPC-2133 2DPC-1866 1DPC-2133 2DPC-1866 1DPC-2133 1DPC-2133 2DPC-1866 1DPC-2133 2DPC-1866 1DPC-2133

Source as of June 2014: Intel internal measurements on platform with one E5-26xx v3 (14C, 2.7GHz, 145W), 8GBx4 SR RDIMM, 16GBx4 DR RDIMM, 32GBx4 QR LRDIMM. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual 37 performance. . For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. Technical Computing Benchmark Performance Memory Bandwidth – STREAM (Triad)

Generational Performance Refresh Performance 3.0 3.5 3.2 2.8 2.8 2.5 3 2.5 2.0 1.44 2 1.5 1.21 1.16 1.5

Relative Relative Performance 1.0

Relative Relative Performance 1 0.5 0.5 0.0 0 1 DIMM/Channel 2 DIMM/Channel 3 DIMM/Channel 1 DIMM/Channel 2 DIMM/Channel 3 DIMM/Channel

E5-2697 v2 E5-2699 v3 X5680 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.33GHz, 130W) (18C, 2.3GHz, 145W)

Source as of July 2014 STREAM(triad): Intel internal measurements on Supermicro XDTN+ platform with two X5680, 8GB DDR3 RDIMMs; platform with two E5-2697 v2 ,16GB DDR3 RDIMMs. New Configuration: Intel® Server System R2208WTTYS with two E5-2699 v3, 16GB RDIMM DDR4-2133. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 38 Intel® Xeon® Processor E5-2600 v3 Product Family Memory Bandwidth – STREAM (triad)

140 118.7 118.7 118.3 120 113.6

100 DDR4-2133 90.4 89.6 2 memory controllers (MC) 80 COD mode DDR4 2133 DDR4 2 MC 58 GB/sec 60 2133 HS mode 1 MC DDR4 HS mode 1866 40 1 MC DDR4 HS mode 1600 20 1 MC HS mode 0 E5-2699 v3 E5-2697 v3 E5-2690 v3 E5-2660 v3 E5-2667 v3 E5-2640 v3 E5-2609 v3 (18C, 2.3GHz, 145W) (14C, 2.6GHz, 145W) (12C, 2.6GHz, 135W) (10C, 2.6GHz, 105W) (8C, 3.2GHz, 135W) (8C, 2.6GHz, 90W) (6C, 1.9GHz, 85W)

Source as of August 2014: Intel internal measurements using Intel® Server System R2208WTTYS with 8x16GB DR-RDIMMs on stream_omp v5.4 with Intel compiler 14.0.3.174 from TR# 3044 for 12-18C, TR#3028 for E5-266x v3, TR#3029 for 6-8C. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 39 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family SPECCPU*2006 Benchmark Measurements

40 Enterprise Benchmark Performance General Computing – SPECint*rate_base2006

Generational Performance Refresh Performance 3.5 3.2 3 • More cores & larger 2.5 cache 2 1.44 • DDR4 memory 1.5 • ive Performance Intel® AVX2 1 • Cluster on Die snoop

Relat 0.5 0 mode E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 29, 2014. Baseline: Supermicro SuperServer 6027AX-TRF platform with two Intel Xeon Processor E5-2697 v2, Intel Compiler version 14.0.0.080, SPECint_rate_base2006 score: 981. Source. HP ProLiant DL360 Gen9 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 7, Intel Compiler version 14.0.0.080, SPECint_rate_base2006 score: 1,360. Source as of September 25, 2014. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed 41 as the property of others. Enterprise Benchmark Performance General Computing – SPECint*rate_base2006

Generational Performance Refresh Performance 3.5 3.2 3 2.5 2 1.44 1.5

ive Performance 1

Relat 0.5 0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014 on SPECint*rate_base2006. Baseline Configuration: 2-socket Intel® Xeon® processor X5690 platform, 48GB memory, RHEL 6.2. Compiler version: Intel Compiler 12.1.0. Source. Score: 425. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 42 Technical Computing Benchmark Performance Floating Point Throughput – SPECfp*rate_base2006

Generational Performance Refresh Performance 3.5 3.3 3.0 • More Cores & Larger 2.5 Cache 2.0 • DDR4 Memory 1.5 1.33 e Performance • Intel® AVX2 1.0 • Cluster on Die Snoop Relativ 0.5 Mode 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W)

Source as of Sept 8 2014. New configuration: Huawei RH2288 V3 platform with two Intel Xeon Processor E5-2699 v3, 16x16GB DDR4-2133 DR-RDIMM, RHEL 6.5, Intel Compiler version 14.0.0.080, SPECfp_rate_base2006 score: 911. Source as of September 25, 2014. Baseline: Supermicro SuperServer 6027AX-TRF platform with two Intel Xeon Processor E5-2697 v2, Intel Compiler version 14.0.0.080, SPECfp_rate_base2006 score: 714. Source. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 43 *Other names and brands may be claimed as the property of others. Technical Computing Benchmark Performance Floating Point Throughput – SPECfp*rate_base2006

Generational Performance Refresh Performance 3.5 3.3

3.0

2.5

2.0

1.5 1.33 e Performance 1.0

Relativ 0.5

0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W)

Source as of Sept 8 2014 on SPECfp*rate_base2006. Baseline Configuration: 2-socket Intel® Xeon® processor X5690 platform, 48 GB memory, SLES11* SP1, Intel Compiler 12.1.0. Source. Score: 271. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 44 Intel® Xeon® Processor E5-2600 v3 Product Family Detailed SPECCPU*rate_base2006 Performance

E5-2699 v3 (18C, 2.3GHz, 145W) vs. E5-2697 v2 (12C, 2.7GHz, 130W) 1.90 1.80 1.70 1.60 1.50 1.44 1.40 1.33 1.30

Relative Relative Performance 1.20 1.10 1.00 Internal 3.gcc 2006 9.mcf obmk mmer .bzip2 bench 481.wrf 40 42 470.lbm 433.milc 473.astar 458.sjeng 465.tonto 447.dealII 401 444.namd 450.soplex 453.povray 445.g 456.h 454.calculix 410.bwaves 416.gamess 434.zeusmp 437.leslie3d 482.sphinx3 464.h264ref 435.gromacs 471.omnetpp SIR_base 400.perl SFR_base2006 483.xalancbmk 436.cactusADM 459.GemsFDTD 462.libquantum

Source as of Oct.13, 2014: results published on http://www.spec.org. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 45 SPECCPU*rate_base2006 SPECCPU*rate_base2006 Performance on Intel http://www.intel.com/performance benchmark data are reportedand confirm whether benchmarkthe referenced data are accurate and reflect performance of systems of design or implementation third party benchmark data or Websites in referenced this document. Intel encourages its all of other information and testsperformance youinto assist as Source: Published results on

SYSmark Basic Standard Advanced 1030 1080 266 824 884 970 306 509 666 703 and and ® MobileMark Xeon 2.5 GHz 2.5 120W 12C GHz 2.6 135W 12C 1.6 GHz 1.6 85W 6C GHz 1.9 85W 6C GHz 2.4 85W 6C GHz 2.4 85W 8C GHz 2.6 90W 8C GHz 2.3 105W 10C GHz 2.6 105W 10C GHz 2.3 120W 12C , are using measured specific computer systems, components, software, operations and functions. Any to any change of thosefa www.spec.org *Other names and brands may beclaimed as theproperty of others. ® Processor Processor E5 as of Oct 13, 2014. Software 2014. as ofOct13, E5 E5 E5 E5 E5 E5 E5 E5 E5 E5 ------2603 v3 2603 v3 2609 v3 2620 v3 2630 v3 2640 v3 2650 v3 2660 v3 2670 v3 2680 v3 2690 fully evaluating fully evaluating your contemplated purchases, including the performance 280 675 689 754 763 792 321 466 561 578 and and workloads used inperformance tests may have been optimized forperformance only on Intelmicroprocessors. - 2600 v3 Product 2600 v3Product Family

Low Frequency Segmented Power Optimized Optimized 1070 1080 1190 1230 1360 556 775 412 457 669 815 cus tomers the referenced to visit Web sites orothers where similar performance of that product product with of that when combined other products. av 2.0 GHz 2.0 14C GHz 2.3 14C GHz 2.6 14C 2.3 GHz 2.3 135W 16C 1.8 GHz 1.8 55W 8C GHz 1.8 65W 12C GHz 3.4 135W 6C (20M) GHz 3.2 135W (20M) 8C 2.3 GHz 2.3 145W 18C 3.0GHz 3.0GHz 105W 4C (10M) 3.5GHz (15M) 135W 4C ailable ailable for purchase. 120W 120W 145W E5 E5 E5 E5 E5 E5 E5 E5 E5 E5 E5 For more IC14 ------2698 v3 2698 v3 2699 cto 2683 v3 2683 v3 2695 v3 2697 2630L v3 2630L v3 2623 v3 2637 v3 2643 v3 2667 2650L v3 2650L rs may cause vary. the resultsto information go to 471 619 378 422 571 653 782 791 838 860 911 Intel does Intel does not controloraudit the Performance Performance tests You You should consult Workstation Workstation Only 3.1 GHz 3.1 160W 2S 10C , such , SPECint SPECfp _base2006 _base2006 _base2006 _base2006 954 741 score score E5 - *_rate 2687W v3 2687W *_rate Processor Transition Guidance – v2 to v3 SIR score SFR score IVB-to-HSW %gains on on IC14 E5-2600 v3 on IC14 E5-2600 v2 (IVB) SPECint*_rate SIR, SFR (HSW) _base2006 Up to 44%, 33% 1360 18C 145W 911 score SIR score SFR score 2.3 GHz E5-2699 v3 on IC 14 on IC 14 on IC 14 16C 135W Up to 30%, 26% 1230 860 SPECfp*_rate 2.3 GHz E5-2698 v3 _base2006 942 12C 130W 682 Up to 26%, 22% 1190 14C 145W 838 score 2.7GHz E5-2697 v2 2.6 GHz E5-2697 v3 on IC14 12C 115W 1070 14C 120W 791 890 2.4GHz E5-2695 v2 659 Up to 21%, 20% 2.3 GHz E5-2695 v3 Up to 20%,18% 1070 14C 120W 774 2.0 GHz E5-2683 v3 8C 130W 8C 135W 780 3.3GHz E5-2667 v2 602 Up to 4%, 8% 815 653 3.2 GHz E5-2667 v3 617 6C 130W 522 6C 135W 3.5GHz E5-2643 v2 Up to 8%, 9% 669 3.4 GHz E5-2643 v3 571 415 4C 130W 387 Up to 10%, 9% 457 4C 135W 422 3.5GHz E5-2637 v2 3.5GHz E5-2637 v3 Up to 1%, 1% 412 4C 105W 378 3.0GHz E5-2623 v3 10C 70W 546 462 Up to 41%, 33% 775 12C 65W 619 1.7GHz E5-2650L v2 1.8 GHz E5-2650L v3 446 6C 60W 556 8C 55W 2.4GHz E5-2630L v2 400 Up to 24%, 17% 1.8 GHz E5-2630L v3 471

Source: Published results on www.spec.org as of Oct. 13, 2014. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more 47 information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. Processor Transition Guidance – v2 to v3 E5-2600 v3 SFR score SIR score SPECint*_rate SIR score E5-2600 v2 (IVB) SFR score IVB-to-HSW %gains on IC14 on IC14 (HSW) _base2006 on IC14 on IC14 on SIR, SFR score on IC 14 12C 135W 871 10C 130W 643 Up to 23%, 23% 1080 792 3.0GHz E5-2690 v2 2.6 GHz E5-2690 v3 SPECfp*_rate _base2006 830 10C 115W 626 Up to 24%, 21% 1030 12C 120W 763 2.8GHz E5-2680 v2 2.5 GHz E5-2680 v3 score on IC14 10C 115W 791 607 970 12C 120W 754 2.5GHz E5-2670 v2 Up to 22%, 24% 2.3 GHz E5-2670 v3 722 10C 95W 575 884 10C 105W 689 2.2GHz E5-2660 v2 Up to 22%, 19% 2.6 GHz E5-2660 v3 8C 95W Up to 24%, 23% 10C 105W 675 663 2.6GHz E5-2650 v2 547 824 2.3 GHz E5-2650 v3

524 8C 95W 451 Up to 34%, 28% 703 8C 90W 578 2.0GHz E5-2640 v2 2.6 GHz E5-2640 v3 489 6C 80W 666 8C 85W 2.6GHz E5-2630 v2 427 Up to 36%, 31% 2.4 GHz E5-2630 v3 561 6C 80W 417 381 Up to 22%, 22% 509 6C 85W 466 2.1GHz E5-2620 v2 2.4 GHz E5-2620 v3 6C 85W 244 4C 80W 258 Up to 25%, 24% 306 321 2.5GHz E5-2609 v2 1.9 GHz E5-2609 v3 6C 85W 181 4C 80W 266 1.8GHz E5-2603 v2 201 Up to 46%, 39% 1.6 GHz E5-2603 v3 280

1.2 1.00 1 0.9 0.87 0.79 0.78 0.79 0.8 0.75 0.69 0.71 0.65 0.6 0.59 0.6 0.56 0.51 0.48 0.49

Perofrmance 0.4 e 0.37 0.4 0.33 0.3 0.22 0.19 Relativ 0.2

0 E5-2697 E5-2699 E5-2698 E5-2697 E5-2695 E5-2683 E5-2690 E5-2680 E5-2670 E5-2660 E5-2650 E5-2640 E5-2630 E5-2620 E5-2667 E5-2643 E5-2637 E5-2623 E5-2650LE5-2630L E5-2609 E5-2603 v2 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 (12C, (18C, (16C, (14C, (14C, (14C, (12C, (12C, (12C, (10C, (10C, (8C, (8C, (6C, (8C, (6C, (4C, (4C, (12C, (8C, (6C, (6C, 2.7GHz, 2.3GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.0GHz, 2.6GHz, 2.5GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.6GHz, 2.4GHz, 2.4GHz, 3.2GHz, 3.4GHz, 3.5GHz, 3.0GHz, 1.8GHz, 1.8GHz, 1.9GHz, 1.6GHz, 130W) 145W) 135W) 145W) 120W) 120W) 135W) 120W) 120W) 105W) 105W) 90W) 85W) 85W) 135W) 135W) 135W) 105W) 65W) 55W) 85W) 85W) Source: Published results on www.spec.org as of Oct. 13, 2014. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 49 *Other names and brands may be claimed as the property of others. Intel® Xeon® Processor E5-2600 v3 Product Family Relative SPECfp*rate_base2006 Performance on IC14

1.2 1.00 1 0.94 0.91 0.86 0.86 0.84 0.83 0.82 0.75 0.8 0.74 0.74 0.71 0.66 0.63 0.61 0.62 0.6 0.51 0.51 0.46 0.41

Performance 0.35 0.4 0.3

0.2 Relative Relative

0 E5-2697 E5-2699 E5-2698 E5-2697 E5-2695 E5-2683 E5-2690 E5-2680 E5-2670 E5-2660 E5-2650 E5-2640 E5-2630 E5-2620 E5-2667 E5-2643 E5-2637 E5-2623 E5-2650LE5-2630L E5-2609 E5-2603 v2 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 (12C, (18C, (16C, (14C, (14C, (14C, (12C, (12C, (12C, (10C, (10C, (8C, (8C, (6C, (8C, (6C, (4C, (4C, (12C, (8C, (6C, (6C, 2.7GHz, 2.3GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.0GHz, 2.6GHz, 2.5GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.6GHz, 2.4GHz, 2.4GHz, 3.2GHz, 3.4GHz, 3.5GHz, 3.0GHz, 1.8GHz, 1.8GHz, 1.9GHz, 1.6GHz, 130W) 145W) 135W) 145W) 120W) 120W) 135W) 120W) 120W) 105W) 105W) 90W) 85W) 85W) 135W) 135W) 135W) 105W) 65W) 55W) 85W) 85W) Source: Published results on www.spec.org as of Oct. 13, 2014. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 50 Intel® Xeon® Processor E5-2600 v3 Product Family SPECint*_base2006 (Speed) Performance on IC14

70 63.6 62.3 63 62.5 63.7 61.3 58.9 60.5 58.1 57.6 57.7 55.6 60 52.6 53.4 53 54.8 53.1 50 45.5

40 34.1 29 30 *_base2006 *_base2006 Score

20 Cint

SPE 10

0 E5-2699 E5-2698 E5-2697 E5-2695 E5-2683 E5-2690 E5-2680 E5-2670 E5-2660 E5-2650 E5-2640 E5-2630 E5-2620 E5-2667 E5-2643 E5-2637 E5-2623 E5-2650L E5-2609 E5-2603 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 (18C, (16C, (14C, (14C, (14C, (12C, (12C, (12C, (10C, (10C, (8C, (8C, (6C, (8C, (6C, (4C, (4C, (12C, (6C, (6C, 2.3GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.0GHz, 2.6GHz, 2.5GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.6GHz, 2.4GHz, 2.4GHz, 3.2GHz, 3.4GHz, 3.5GHz, 3.0GHz, 1.8GHz, 1.9GHz, 1.6GHz, 145W) 135W) 145W) 120W) 120W) 135W) 120W) 120W) 105W) 105W) 90W) 85W) 85W) 135W) 135W) 135W) 105W) 65W) 85W) 85W)

Source as of Oct. 13, 2014: Published results on www.spec.org. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 51 Intel® Xeon® Processor E5-2600 v3 Product Family SPECfp*_base2006 (Speed) Performance on IC14

140 115 120 108 111 105 103 102 106 97.5 97.2 94.8 98.1 95.2 99.6 100 91.3 87.3 90 80

*_base2006 *_base2006 Scores 40 fp 20 SPEC 0 E5-2699 E5-2698 E5-2697 E5-2695 E5-2683 E5-2690 E5-2680 E5-2670 E5-2660 E5-2650 E5-2640 E5-2630 E5-2620 E5-2667 E5-2643 E5-2637 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 v3 (18C, (16C, (14C, (14C, (14C, (12C, (12C, (12C, (10C, (10C, (8C, (8C, (6C, (8C, (6C, (4C, 2.3GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.0GHz, 2.6GHz, 2.5GHz, 2.3GHz, 2.6GHz, 2.3GHz, 2.6GHz, 2.4GHz, 2.4GHz, 3.2GHz, 3.4GHz, 3.5GHz, 145W) 135W) 145W) 120W) 120W) 135W) 120W) 120W) 105W) 105W) 90W) 85W) 85W) 135W) 135W) 135W)

Intel® Xeon® Processor E5-2600 v3 Product Family Energy Efficiency & DDR4 Power Savings

53 Enterprise Benchmark Performance Energy Efficiency – SPECPower*_ssj2008

Generational Performance Refresh Performance 3.5 3.1 3.0 2.5 2.0 1.5 1.24 1.0

lative lative Performance 0.5 Re 0.0 E5-2660 v2 E5-2699 v3 X5675 E5-2699 v3 (10C, 2.6GHz, 95W) (18C, 2.3GHz, 145W) (6C, 3.06GHz, 95W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014. Sugon I620-G20 platform with two Intel Xeon Processor E5-2699 v3, 8x16GB DDR4-2133 DR-RDIMM, Microsoft Windows Server 2008 R2, IBM J9 VM, 10,599 overall ssj_ops/watt. Source, results submitted as of Sept. 8 2014. Dell Inc. PowerEdge T620 platform with two Intel Xeon Processor E5-2660 v2, IBM J9 VM, 8,533 overall ssj_ops/watt. Source. HP ProLiant DL360 G7* with two Intel® Xeon® Processor X5675, Source, Score: 3,329 overall ssj_ops/watt. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 54 Relative Power of DDR3, DDR3L, DDR4 DIMMs

DDR3 DDR3L DDR4 Lower is better 2 1.85 r 1.5 1.51 1.5 1.23 1 1 1

0.5

Relative DIMM Powe 0 1 DIMM/Channel 2 DIMMs/Channel Source as of Sept 2014: Internal Intel measurements for DIMM level power using16GB 2Rx4 RDIMMs on workload of 67% read, 20% page hit. DDR3 (1DPC, 2DPC): 5.63W, 7.27W. DDR3L (1DPC, 2DPC): 4.61W, 5.93W. DDR4 (1DPC, 2DPC): 3.74W, 3.91W. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 55 DIMM Level Power Savings with DDR4

DDR3 DDR3L Higher is better 0.5 0.46 0.4 0.33 0.34

Power Reduction 0.3 0.2 0.18 0.1 Relative DIMM 0 1 DIMM/Channel 2 DIMMs/Channel Source as of Sept 2014: Internal Intel measurements for DIMM level power using16GB 2Rx4 RDIMMs on workload of 67% read, 20% page hit. DDR3 (1DPC, 2DPC): 5.63, 7.27W. DDR3L (1DPC, 2DPC): 4.61, 5.93W. DDR4 (1DPC, 2DPC): 3.74, 3.91W. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 56 System Level Power Comparison of DDR3 vs. DDR4

1 DIMM/Channel 2 DIMMs/Channel 3 DIMMs/Channel Lower is better 1.1 1.08 1.08 1.06 1.05 1.04 1.02 1.01 1.02 1 1 1 0.98

Relative System Level Power Relative Level System 0.96 E5-2697 v2 -DDR3 E5-2690 v3 - DDR4 (12C, 2.7GHz, 130W) (12C, 2.6GHz, 135W)

Source as of July 2014 system level power measurements of STREAM(triad): Intel platform with two E5-2697 v2 ,16GB DDR3 RDIMMs. New Configuration: Intel® Server System R2208WTTYS with two E5-2699 v3, 16GB DDR4 RDIMMs. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 57 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Virtualization Benchmark Measurements

58 Enterprise Benchmark Performance Virtualization – SPECVirt*_sc2013

Generational Performance Refresh Performance 4.0 3.4 3.5 3.0 2.5 • Higher CPU utilization 1.70 2.0 • DDR4 Memory 1.5 ive Performance 1.0 • Faster QPI Speed

Relat 0.5 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014. Baseline: IBM System x3650 M4 platform with two Intel Xeon Processor E5-2697 v2, SPECvirt_sc2013 947.0 @ 53 VMs. Source. New configuration: Hewlett-Packard Company ProLiant DL360 Gen9 with two Intel Xeon Processor E5-2699 v3, SPECvirt_sc2013 1614 @ 95 VMs. Source. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 59 *Other names and brands may be claimed as the property of others. Enterprise Benchmark Performance Virtualization – VMmark 2.x

Generational Performance Refresh Performance 3.5 3.3 3.0 2.5 2.0 1.60 • Higher CPU utilization 1.5 • DDR4 Memory ive Performance 1.0 • Faster QPI Speed

Relat 0.5 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014. New Configuration: Fujitsu PRIMERGY RX2540 M1 platform with two Intel Xeon Processor E5-2699 v3, VMware ESXi 5.5.0 U2, VMmark v2.5.2 score: 26.48 @ 22 tiles. Source. Baseline: HP ProLiant DL380p Gen8 platform with two Intel Xeon E5-2697 v2, VMWare ESXi 5.1.0 U1, VMmark v2.5.1 score: 16.54 @ 14 tiles. Source .Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 60 *Other names and brands may be claimed as the property of others. Enterprise Benchmark Performance Virtualization – VMmark2.x

Generational Performance Refresh Performance 3.5 3.3 3.0 2.5 2.0 1.60 1.5

ive Performance 1.0

Relat 0.5 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8, 2014. Baseline: Fujitsu PRIMENERGY RX300 S6 with two Intel® Xeon® Processor X5690, VMware ESXi 4.1 U1, VMmark* v2.1.1 score: 7.59 @ 7 tiles, source. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 61 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Business Processing Benchmark Measurements

62 Enterprise Resource Planning Performance SAP Sales & Distribution* (2 Tier) on Windows

Generational Performance Refresh Performance 3.5 3.00 3.0 2.5 • Cluster on Die snoop 2.0 1.56 mode 1.5

ive Performance • More cores & larger 1.0 cache Relat 0.5 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q3'14 Launch

Source as of Sept 8 2014. results submitted or published to www.sap.com/benchmark, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 63 Enterprise Resource Planning Performance SAP Sales & Distribution* (2 Tier) on Windows

Generational Performance Refresh Performance 3.5 3.00 3.0 2.5 2.0 1.56 1.5

ive Performance 1.0

Relat 0.5 0.0 E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q3'14 Launch

Source as of Sept 8 2014. results submitted or published to www.sap.com/benchmark, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 64 Configuration Details for SAP Sales & Distribution (2-Tier) on Windows Source as of Sept 8, 2014:

. 2-socket Intel® Xeon® processor X5690 platform (6 cores, 12 threads), Microsoft Windows Server 2008 R2 Enterprise Edition, Microsoft SQL Server 2008. SAP* ERP 6.0 Enhancement Package 4. Score: 5,220 SD users. Source.

. Fujitsu PRIMERGY RX300 S8 platform with two Intel Xeon Processor E5-2697 v2 (12 cores, 24 threads), Microsoft Windows Server 2012 Standard Edition, Microsoft SQL Server 2012, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 10,240. Source.

. Fujitsu PRIMERGY RX2540 M1 platform with two Intel Xeon Processor E5-2699 v3 (18 cores, 36 threads), Microsoft Windows Server 2012, Microsoft SQL Server 2012, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 16,000. Source: Certification #2014028.

. IBM System x3650 M5 platform with two Intel Xeon Processor E5-2699 v3 (18 cores, 36 threads), Microsoft Windows Server 2012 Standard Edition, IBM DB2 10, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 16,000. Source: Certification #2014030.

65 Enterprise Resource Planning Performance SAP Sales & Distribution* (2 Tier) on Linux

Generational Performance Refresh Performance 3.50 3.30 3.00 2.50 • Cluster on Die snoop 2.00 1.60 mode 1.50

ive Performance • More cores & larger 1.00 cache Relat 0.50 0.00 E5-2697 v2 E5-2699 v3 X5680 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.33GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014. results submitted or published to www.sap.com/benchmark, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 66 Enterprise Resource Planning Performance SAP Sales & Distribution* (2 Tier) on Linux

Generational Performance Refresh Performance 3.50 3.30 3.00 2.50 2.00 1.60 1.50

ive Performance 1.00

Relat 0.50 0.00 E5-2697 v2 E5-2699 v3 X5680 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.33GHz, 130W) (18C, 2.3GHz, 145W) Q3'13 Launch Q3'14 Launch Q1'11 Launch Q3'14 Launch

Source as of Sept 8 2014. results submitted or published to www.sap.com/benchmark, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 67 Configuration Details for SAP Sales & Distribution (2-Tier) on Linux Source as of Sept 8 2014:

. Fujitsu platform with two Intel Xeon processor X5690 (6 cores, 12 threads), SuSE Linux Enterprise Server 11 SP1, MaxDB 7.8, SAP enhancement package 4 for SAP ERP 6.0, number of SAP SD benchmark users: 4875. Source: Certification #2011028.

. Dell PowerEdge R720 platform with two Intel Xeon Processor E5-2697 v2 (12 cores, 24 threads), Red Hat Enterprise Linux 6.5, SAP ASE 16, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 10,253. Source.

9. Dell PowerEdge R730 platform with two Intel Xeon Processor E5-2699 v3 (18 cores, 36 threads), Red Hat Enterprise Linux 7, SAP ASE 16, SAP enhancement package 5 for SAP ERP 6.0, number of SAP SD benchmark users: 16,500. Source as of September 11, 2014.

68 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family High Performance Computing Benchmark Measurements

69 Intel® Xeon® Processor E5-2600 v3 Product Family High Performance Computing – MPI Performance

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2690 v3 (12C, 2.6GHz, 135W) E5-2697 v3 (14C, 2.6GHz, 145W) 1.40 1.29 1.20 1.20 1.20 1.00 0.80 0.60 elative Performance elative R 0.40 0.20 0.00 SPECmpiM*_base2007

Source as of Sept. 29, 2014: Published results on www.spec.org or Intel estimates based on internal testing, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance For more information go to http://www.intel.com/performance 70 *Other names and brands may be claimed as the property of others. Benchmark Configuration Details for SPECmpiM*_base2007

Source as of Sept. 29 2014: . SGI Rackable C2112-4RP4 platform with two Intel Xeon Processor E5-2697 v2 (12 cores), 24 MPI ranks, Intel Compiler version 14.0.0.051, SPECmpiM_base2007 score: 6.51. Source. . SGI ICE X platform with two Intel Xeon Processor E5-2690 v3 (12 cores), 24 MPI ranks, Intel Compiler version 14.0.3.174, SPECmpiM_base2007 score: 7.83. Source. . Intel Corporation Endeavor with two Intel Xeon processor E5-2697 v3 (14 cores), 28 MPI ranks, Intel Compiler version 14.0.3.174, SPECmpiM_base2007 score: 7.85. Source. . Source: Intel estimates based on internal testing on E5-2699 v3 (18 cores), 36 MPI ranks, Intel Compiler version 14.0.1.106, HS mode, 8x16GB DDR4-2133 DR-RDIMM, RHEL 6.4, AVX2 enabled, Turbo enabled, THP enabled, SPECmpiM_base2007 estimated score: 8.45.

71 Intel® Xeon® Processor E5-2600 v3 Product Family High Performance Computing – OpenMP Performance

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2699 v3 (18C, 2.3GHz, 145W) 1.60 1.42 1.40 1.20 1.00 0.80 0.60 elative Performance elative R 0.40 0.20 0.00 SPECompG*_base2012

Source as of Sept. 29, 2014: New configuration: Cisco Systems Cisco UCS C220 M4 platform with two Intel Xeon Processor E5-2699 v3 (18 cores), 72 OpenMP threads, SPECompG_base2012 score: 9.67. Source. Baseline: Cisco Systems Cisco UCS C240 M3 platform with two Intel Xeon Processor E5-2697 v2 (12 cores), 48 OpenMP threads, SPECompG_base2012 score: 6.79. Source. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 72 *Other names and brands may be claimed as the property of others. GFLOPs Measured Linpack - AVXBaseFreq Max TheoreticalPeak - BaseFreq Max TheoreticalPeak Intel Theoretical Peak FLOPS Theoretical http://www.intel.com/performance assist youin computer systems, components, software, operations functions. Any and to change any ofthosefactors maycause the resultsto workloads used inperformance tests may have been optimized forperformance only on Intelmicroprocessors. Source as ofJune two E5 v3, HSW 2014: Platform with ® fully evaluating fully evaluating your contemplated purchases, including the performance product with of that when combined other p Max Theoretical Peak Performance (DP FLOPS) = 2 x(#cores) x ( Peak Performance sockets Max Theoretical Xeon (18C, 145W, 2.3/1.9GHz) E5-2699 v3 76% 1324 1094 93% ® *Other names and brands may beclaimed as theproperty of others. 1012 Processor E5 (16C, 135W, 2.3/1.9GHz) 77% E5-2698 v3 1177 93% - C1, HT disabled, BIOS 27.R01, Turbo enabled, NUMA & COD 8x16GB mode, DDR4 972 901 80% 2.6/2.2GHz) (14C, 145W E5-2697 v3 1164 94% 985 929 78% (14C, 120W, - 2.3/1.9GHz) E5-2695 v3

1030 Product Family v3 2600 95% 851 Performance Performance su tests, 806 92% 2.0/1.7GHz) E5-2683 v3 (14C,120W, 896 108% va roducts. ry. ch as 761 You You should consult other information and testsperformance to For more information go to SYSmark 821 - 2 133, 6.4,IC14.0 RHEL (12C, 135W, 2.6/2.3GHz) 81% E5-2690 v3 freq and and 998 MobileMark 92% 883 ) x(16DP FLOPS) Linpack Efficiencies Linpack

, are measured using specific 812 - AVX2, AVX2, Software MKL 11.1.1. and 79% (12C, 120W, 2.5/2.1GHz) E5-2680 v3 960 94% 806 758 85% 2.3/2.0GHz) (12C,120W, E5-2670 v3 883 95% 768 751 73 Intel® Xeon® Processor E5-2600 v3 Product Family Linpack Performance Max Theoretical Peak Performance (DP FLOPS) = 2 sockets x (#cores) x (TDP freq) x (16 DP FLOPS)

1,200 92% 100% 90% Frequency on TDP based Efficiency 85% 90% 76% 80% 81% 79% 81% 81% 1,000 77% 78% 80%

70% 800 60%

600 50% (GFLOPs) 40% max 400 Up to R 30% 1.9X 20% 200 10% 528 1012 901 929 806 821 812 758 751 673 663 536 0 0% E5-2697 v2 E5-2699 v3 E5-2698 v3 E5-2697 v3 E5-2695 v3 E5-2683 v3 E5-2690 v3 E5-2680 v3 E5-2670 v3 E5-2660 v3 E5-2650 v3 E5-2640 v3 (12C, 130W, (18C, 145W, (16C, 135W, (14C, 145W (14C, 120W, (14C,120W, (12C, 135W, (12C, 120W, (12C,120W, (10C, 105W, (10C, 105W, (8C, 90W, 2.7GHz) 2.3/1.9GHz) 2.3/1.9GHz) 2.6/2.2GHz) 2.3/1.9GHz) 2.0/1.7GHz) 2.6/2.3GHz) 2.5/2.1GHz) 2.3/2.0GHz) 2.6/2.2GHz) 2.3/2.0GHz) 2.6/2.2GHz)

Source as of June 2014: Intel internal measurements on platform with two E5-2697 v2, HT disabled, Turbo enabled, 8x8GB DDR3-1866, RHEL6.3, MKL 11.0.5. Platform with two E5 v3, HSW-C1, BIOS 27.R01, HT disabled, Turbo enabled, NUMA & COD mode, 8x16GB DDR4-2133, RHEL 6.4, IC14.0-AVX2, MKL 11.1.1. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 74 Intel Confidential Intel® Xeon® Processor E5-2600 v3 Product Family High Performance Computing Summary

Intel® Xeon® E5-2699 v3 (18C, 2.3GHz) vs. Intel® Xeon® E5-2697 v2 (12C, 2.7GHz) Up to 1.72x 2.5 Geomean Up to 2 Up to Up to Up to 1.39x 1.24x 1.25x 1.32x Geomean Geomean 1.5 Geomean Geomean ++

1 ++ ++ Sim + + ++ + ++ kernel2 kernel3 0.5 Simulation + - - previous generation previous FNR BlackScholes binomialcpu MonteCarlo Dynamic Analysis FEA Crash CFD Multiphysics MILC Amber Gaussian Gromacs Gamess Bowtie2++ Blast NAMD WRF ROMS HOMME SEIS SEIS

Relative Performance vs. Performance Relative 0 Baseline Computer Life Sciences Energy Numerical Financial E5-2697 v2 Aided Weather Services Engineering +Optimized for AVX. ++Optimized for AVX2. Source as of June 2014: Intel internal measurements on platform with two E5-2697 v2, 8x8GB DDR3-1866, RHEL6.3. Platform with two E5-2699 v3, 8x8GB DDR4-2133, RHEL 6.3, NUMA-COD mode for all except Energy- ES mode. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 75 Intel Confidential Intel® Xeon® Processor E5-2600 v3 Product Family Financial Services Applications

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2667 v3 (8C, 3.2GHz, 135W) E5-2680 v3 (12C, 2.5GHz, 120W) E5-2695 v3 (14C, 2.3GHz, 120W) E5-2697 v3 (14C, 2.6GHz, 145W) E5-2699 v3 (18C, 2.3GHz, 135W) 2.50 2.11 2.01 2.00 1.92 1.80 1.56 1.51 1.46 1.42 1.41 1.50 1.38 1.28 1.26 1.26 1.19 1.18 1.08 1.06 1.02 1.01 1.00 1.00 1.00 1.00 1.00 0.80

Relative Relative Performance 0.50

0.00 binomialcpu BlackScholes FNR MonteCarlo

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2667 v3 (8C, 3.2GHz, 135W) E5-2680 v3 (12C, 2.5GHz, 120W) E5-2695 v3 (14C, 2.3GHz, 120W) E5-2697 v3 (14C, 2.6GHz, 145W) E5-2699 v3 (18C, 2.3GHz, 135W)

1.60 1.45 1.35 1.35 1.34 1.40 1.34 1.30 1.27 1.26 1.24 1.23 1.21 1.19 1.18 1.16 1.15 1.15 1.14 1.13 1.12

1.20 1.11 1.09 1.08 1.05 1.04 1.04 1.01 1.01 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.99 0.98 0.96 0.96

1.00 0.93 0.93 0.88 0.83 0.78 0.78 0.80 0.76

0.60

Relative Performance 0.40

0.20

0.00 Amber Blast bowtie2 Gamess Gaussian Gromacs NAMD MILC

1.60 1.45 1.39 1.34 1.40 1.28 1.28 1.27 1.22 1.17 1.17 1.19 1.20 1.11 1.11 1.00 1.00 1.00 1.01 1.00 0.92 0.88

ive Performance 0.80

Relat 0.60

0.40

0.20

0.00 HOMME ROMS WRF

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2667 v3 (8C, 3.2GHz, 135W) E5-2680 v3 (12C, 2.5GHz, 120W) E5-2695 v3 (14C, 2.3GHz, 120W) E5-2697 v3 (14C, 2.6GHz, 145W) E5-2699 v3 (18C, 2.3GHz, 135W)

1.40 1.32 1.29 1.27 1.24 1.22 1.21 1.20 1.19 1.19 1.18 1.18 1.15 1.13

1.20 1.12 1.10 1.09 1.09 1.08 1.06 1.06 1.04 1.03 1.00 1.00 1.00 1.00 1.00 0.98

1.00 0.94 0.89

0.80

0.60

0.40 Relative Performance

0.20

0.00 FiniteABAQUS-exp Element DynamicABAQUS-std Analysis ComputationalFluent LS-DYNAMultiphysics CrashPamCrash Simulation Analysis Fluid Dynamics Simulation

Source as of July 2014, details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 79 Intel Confidential System Configurations for HPC Applications Intel® Xeon® E5-2697 v2 Intel® Xeon® E5-2667 v3 Intel® Xeon® E5-2680 v3 Intel® Xeon® E5-2695 v3 Intel® Xeon® E5-2697 v3 Intel® Xeon® E5-2699 v3 Sockets / Cores 2 x 12C 2 x 8C 2 x 12C 2 x 14C 2 x 14C 2 x 22C Base Freq 2.7GHz, C0 3.2GHz, R2 2.5GHz, M0 2.3GHz, C1 2.6GHz, C0 2.3GHz Turbo Mode Enabled Enabled Enabled Enabled Enabled Enabled Memory 64 GB DDR3-1867 128 GB DDR4-2133 128 GB DDR4-2133 128 GB DDR4-2133 128 GB DDR4-2133 128 GB DDR4-2133 Platform Romley-EP Server SDP Wildcat Pass Wildcat Pass Wildcat Pass Mayan City Mayan City HT HT Snoop Mode HT Snoop Mode HT Snoop Mode HT Snoop Mode HT Snoop Mode Computer Aided Engineering Finite Element Analysis ON ON ES OFF COD OFF COD ON COD ON COD Dynamic Analysis ON ON ES ON ES ON ES ON HS ON HS Computational Fluid Dynamics ON ON ES ON COD ON COD ON COD ON COD Multiphysics Simulation ON ON ES OFF COD OFF COD ON COD ON COD Crash Simulation ON ON ES ON COD ON COD ON COD ON COD Energy seis-kernel2 (Kirchhoff-PB) v1.1-12.0 ON ON ES ON COD ON COD ON COD ON COD seis-kernel3 (TTI-T4-WG) v1.1-12.0 ON OFF ES ON COD OFF ES OFF HS ON HS Financial Services binomialcpu v3.0-13.1.0_AVX / AVX2 ON ON ES ON COD ON COD ON HS ON HS BlackScholes v5.0-13.1.1_AVX / AVX2 ON ON ES ON ES ON COD ON COD ON COD FNR v4.0-12.0 ON ON ES ON COD ON ES ON COD ON ES MonteCarlo v3.0-13.1.0-AVX / AVX2 ON ON ES OFF ES ON COD OFF COD ON COD Life Sciences Amber v12-13.1.0_SSE4.2 / v12- ON ES 14.0.0_AVX2 ON OFF ES OFF ES ON HS ON ES Blast v2.2.28+_13.1.0_OPT2 ON ON ES ON ES ON COD ON COD ON COD bowtie2 v2-2.1.0.0-13.1 / AVX2 ON ON ES ON COD ON COD ON HS ON ES Gamess v01MAY2012-R1-12.1* / ON ES v01MAY2013.R1-14.0.1_AVX2 ON ON ES ON COD ON COD ON COD Gaussian g09-D.01 ON OFF ES OFF COD OFF COD OFF COD ON HS Gromacs v4.6.1-13.1.1_AVX ON ON ES ON ES ON COD ON ES ON ES NAMD v2.9-13.1.1_OPT3 / v2.9- ON ES 14.0.0_AVX2 ON OFF ES OFF COD ON ES ON COD MILC v7.7.8-13.1.1_OPT3 / v7.7.8- ON ES 14.0.0_AVX2 ON ON COD ON COD ON COD ON COD Numerical Weather HOMME v2841-20130227_AVX ON ON ES ON COD ON COD ON COD ON COD ROMS v3.0-12.0_AVX / v3.6.690- ON ES 14.0.0_AVX2 OFF OFF COD OFF COD ON COD ON COD WRF v3.1-11.1_AVX / v3.5-14.0.0_AVX2 ON ON ES OFF COD OFF COD ON COD ON COD 80 Intel Confidential Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Java, Web, Cloud Benchmark Measurements

81 Intel® Xeon® Processor E5-2600 v3 Product Family Java Performance & Benefits of JDK8

E5-2697 v2 E5-2699 v3 E5-2699 v3 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (18C, 2.3GHz, 145W) (18C, 2.3GHz, 145W) JDK7u11 JDK7u11 + 2x memory JDK8u5 + 2x memory JDK8u11 + 2x memory + optimizations 4 3.58 3.5 3 2.62 2.5 2 1.45 1.5 Relative Relative Performance 1 Up to 80% gain from JDK8 0.5 0 Preliminary SPECjbb2013-MultiJVM max-jOPs

Source as of Sept 8 2014: results published or submitted to www.spec.org, details on the next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 82 *Other names and brands may be claimed as the property of others. Java Performance & Benefits of JDK8 Configuration Details on SPECjbb*2013-MultiJVM Source as of Sept 8 2014:

. Intel Server Board S2600GZ with two Intel Xeon Processor E5-2697 v2, 128GB memory, RHEL* 6.5, Oracle Java Standard Edition 7 update 11, 29GB heap. Score: 53,168 SPECjbb2013-MultiJVM max- jOPS , 19,139 SPECjbb2013-MultiJVM critical-jOPS. Source.

. Intel® Server Board S2600WTT with two Intel® Xeon® processors E5-2699 v3, 512 GB memory, Red Hat Enterprise Linux* 6.5, Oracle Java* SE Development Kit 7u11, 240 GB heap. Score: 77,303 SPECjbb2013-MultiJVM max-jOPS and 41,226 SPECjbb2013-MultiJVM critical-jOPS. Source.

. New configuration: Intel® Server Board S2600WTT with two Intel® Xeon® Processor E5-2699 v3, 512 GB memory, Red Hat Enterprise Linux* 6.5, Java* Standard Edition 8 Update 5, source. Score: 139,511 SPECjbb*2013-MultiJVM max-jOPS, 43,107 SPECjbb*2013-MultiJVM critical-jOPs.

. HP ProLiant ML350 Gen9 platform with two Intel Xeon Processor E5-2699 v3, 512GB RAM, SUSE, Oracle Java SE Development Kit 8u11, 236GB heap. Score: 190,674 SPECjbb2013-MultiJVM max- jOPs, 47,139 SPECjbb2013-MultiJVM critical-jOPs. Source.

83 Java Performance SPECjbb*2013-MultiJVM max-jOPs Generational Performance Refresh Performance 3.5 3.02 10.0 8.7 3.0 8.0 2.5 2.0 6.0

1.5 4.0 Relative Relative Performance 1.0 Relative Performance 2.0 0.5 0.0 0.0

E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) JDK7u45 JDK8u11 JDK7u11 JDK8u11

Source as of Sept 29, 2014: results published on www.spec.org. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 84 *Other names and brands may be claimed as the property of others. Benchmark Configuration Details for SPECjbb2013-MultiJVM max-jOPS Source as of Sept 8 2014:

. Baseline configuration: Supermicro* X8DTN+ with two Intel® Xeon® Processor X5690, Oracle Java* Standard Edition 7 Update 11, source. Score: 21,709 SPECjbb*2013-MultiJVM max-jOPs, 3,587 SPECjbb*2013-MultiJVM critical-jOPS.

. Intel® Xeon® E5 v2 Family Baseline: Cisco Systems Cisco UCS C240 M3 platform with two Intel Xeon Processor E5-2697 v2, Oracle Java Standard Edition 7 update 45, 63,079 SPECjbb2013-MultiJVM max-jOPS , 23,797 SPECjbb2013-MultiJVM critical-jOPS. Source as of August 26, 2014.

. New configuration: Hewlett-Packard Company HP ProLiant ML350 Gen9 platform with two Intel Xeon Processor E5-2699 v3, Oracle Java Standard Edition 8 update 11, 190,674 SPECjbb2013- MultiJVM max-jOPS, 47,139 SPECjbb2013-MultiJVM critical-jOPS. Source.

85 Java Performance SPECjbb*2013-MultiJVM critical-jOPs Generational Performance Refresh Performance 3.0 20.0 2.45 16.3 2.5 15.0 2.0

1.5 10.0 Relative Relative Performance 1.0 Relative Performance 5.0 0.5

0.0 0.0

E5-2697 v2 E5-2699 v3 X5690 E5-2699 v3 (12C, 2.7GHz, 130W) (18C, 2.3GHz, 145W) (6C, 3.46GHz, 130W) (18C, 2.3GHz, 145W) JDK7u40 JKD8u20 JDK7u11 JDK8u20

Source as of Sept 8 2014: results published or submitted to www.spec.org. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect actual performance. Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. For more information go to http://www.intel.com/performance 86 *Other names and brands may be claimed as the property of others. Benchmark Configuration Details for SPECjbb2013-MultiJVM critical-jOPS Source as of Sept 8 2014: . Supermicro* X8DTN+ with two Intel® Xeon® Processor X5690, Oracle Java* Standard Edition 7 Update 11, source. Score: 21,709 SPECjbb*2013-MultiJVM max-jOPs, 3,587 SPECjbb*2013-MultiJVM critical-jOPS. . Hewlett-Packard Company HP ProLiant ML350p Gen8 platform with two Intel Xeon Processor E5-2697 v2, Oracle Java Standard Edition 7 update 40, 57,594 SPECjbb2013-MultiJVM max-jOPS, 32,103 SPECjbb2013-MultiJVM critical-jOPS. Source. . Cisco UCS C220 M4 platform with two Intel Xeon Processor E5-2699 v3, Oracle Java Standard Edition 8 update 20, 160,283 SPECjbb2013-MultiJVM max-jOPS, 58,478 SPECjbb2013-MultiJVM critical-jOPS. Source.

87 Intel® Xeon® Processor E5-2600 v3 Product Family mWeb Performance

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2697 v3 (14C, 2.6GHz, 145W) E5-2699 v3 (18C, 2.3GHz, 145W)

2 1.83 1.67 1.52 1.5 1.38 1.38 1.43

Relative Relative Performance 1

0.5

0 Banking eCommerce Support

Source as of June 2014 TR#1396: Intel internal measurements on platform with two E5-2697 v2, 8x8GB DDR3-1866, HT, Turbo, Prefetchers NUMA enabled, 10GbE Niantic, OS: Fedora 18, Nginx 1.4.4, PHP 5.4.16, Memcached 1.47. Platform with two E5 v3 , HT, Turbo, Prefetchers enabled, NUMA & ES mode, 8x8GB DDR4-2133, BIOS 25R02, HT, Turbo, Prefetchers NUMA enabled, 10GbE Niantic, OS: Fedora 18, Nginx 1.4.4, PHP 5.4.16, Memcached 1.47. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 88 Intel Confidential Memcached

. Open-source technology that uses the pooled DRAM from a server cluster as a storage cache, dramatically reducing latency – “networked DRAM” . Used by many cloud & web-services companies

. Frequently requested files can be delivered directly from pooled DRAM rather than forcing the server managing the request to retrieve it from storage . Memcached is an excellent stress test that shows what peak loads a network can handle

89 What is Memcached?

A distributed memory caching system based on “key-value” caching store

. Frontend sends “get” request for a “key” to memcached server

. If “key” found, memcached returns “value” for it

. If not found, the persistent db is queried & “set” request sent to store it in memcached

. Entire cache (key-value pairs) spread over many memcached servers

. Typical transactions – get, set, delete, stats, replace, incr, dec, etc.

. Most frequent transaction is “random get”

. Perf metric - RPS (requests/sec) with 99.5% queries return with response time <= 1ms

90 Intel® Xeon® Processor E5-2600 v3 Product Family Memcached Performance

E5-2697 v2 (12C, 2.7GHz, 130W) E5-2697 v3 (14C, 2.6GHz, 145W) E5-2699 v3 (18C, 2.3GHz, 145W) 1.97 2 1.8 1.60 1.6 1.4 1.2

Relative Relative Performance 1 0.8 0.6 Up to 40% of perf gain 0.4 from HSW IPC gains 0.2 0 Memcached

Source as of June 2014 TR#1427: Intel internal measurements on platform with two E5-2697 v2, 8x8GB DDR3-1866, HT, Turbo, Prefetchers NUMA enabled, RHEL 6.5, Memcached 1.6 Bag LRU, 2x10GbE Niantic, Score: 4.2M requests/sec. Platform with two E5 v3 , HT, Turbo, Prefetchers enabled, NUMA & ES mode, 8x8GB DDR4-2133, RHEL 6.5, BIOS 25R02, Memcached 1.6 Bag LRU, 2x10GbE Niantic, Score: 6.72M, 8.28M requests/sec. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. 91 Return to Table of Contents

Comparisons to AMD Opteron 6300 Series

92 Intel® Xeon® E5-2600 v3 Product Family 2S Integer Throughput vs. AMD Opteron* 6300 Series

2S Xeon® E5-2699 v3 (2.3GHz, 18C, 45M) 1,360 2S Xeon® E5-2698 v3 (2.3GHz, 16C, 40M) 1,230 2S Xeon® E5-2697 v3 (2.6GHz, 14C, 35M) 1,190 2S Xeon® E5-2690 v3 (2.6GHz, 12C, 30M) 1,080 2S Xeon® E5-2695 v3 (2.3GHz, 14C, 35M) 1,080 2S Xeon® E5-2683 v3 (2GHz, 14C, 35M) 1,070 2S Xeon® E5-2680 v3 (2.5GHz, 12C, 30M) 1,030 2S Xeon® E5-2670 v3 (2.3GHz, 12C, 30M) 970 2S Xeon® E5-2660 v3 (2.6GHz, 10C, 25M) 884 2S Xeon® E5-2650 v3 (2.3GHz, 10C, 25M) 824 2S Xeon® E5-2667 v3 (3.2GHz, 8C, 20M) 815 2S Xeon® E5-2650L v3 (1.8GHz, 12C, 30M) 767 2S Xeon® E5-2640 v3 (2.6GHz, 8C, 20M) 703 2S Xeon® E5-2643 v3 (3.4GHz, 6C, 30M) 669 2S Xeon® E5-2630 v3 (2.4GHz, 8C, 20M) 666 2S Xeon® E5-2630L v3 (1.8GHz, 8C, 20M) 556 Opteron 6386 SE, 16C, 2.8GHz,140W 539 2S Xeon® E5-2620 v3 (2.4GHz, 6C, 15M) 509 2S Opteron 6380, 16C, 2.5GHz, 115W 502 2S Opteron 6378, 16C, 2.4GHz, 115W 476 2S Opteron 6376, 16C, 2.3GHz, 115W 462 2S Xeon® E5-2637 v3 (3.5GHz, 4C, 15M) 457 2S Opteron 6348, 12C 2.8GHz, 115W 427 2S Xeon® E5-2623 v3 (3GHz, 4C, 10M) 412 2S Opteron 6370P, 16C, 2.0GHz, 99W 411 2S Opteron 6366 HE, 16C, 1.8GHz, 85W 397 2S Opteron 6344, 12C 2.6GHz, 115W 396 2S Opteron 6338P, 12C, 2.3GHz, 99W 357 2S Opteron 6328, 8C 3.2GHz, 115W 327 2S Xeon® E5-2609 v3 (1.9GHz, 6C, 15M) 306 2S Opteron 6320, 8C, 2.8GHz, 115W 299 2S Xeon® E5-2603 v3 (1.6GHz, 6C, 15M) 266 2S Opteron 6308, 4C 3.5GHz, 115W 178 0 200 400 600 800 1,000 1,200 1,400 SPECint*_rate_base2006 Source: published results on www.spec.org as of Sept 29, 2014. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. *Other names and brands may be 93 claimed as the property of others. Intel® Xeon® E5-2600 v3 Product Family 2S Floating Point Throughput vs. AMD Opteron* 6300 Series

2S Xeon® E5-2699 v3 (2.3GHz, 18C, 45M) 911 2S Xeon® E5-2698 v3 (2.3GHz, 16C, 40M) 860 2S Xeon® E5-2697 v3 (2.6GHz, 14C, 35M) 838 2S Xeon® E5-2690 v3 (2.6GHz, 12C, 30M) 792 2S Xeon® E5-2695 v3 (2.3GHz, 14C, 35M) 791 2S Xeon® E5-2683 v3 (2GHz, 14C, 35M) 774 2S Xeon® E5-2680 v3 (2.5GHz, 12C, 30M) 763 2S Xeon® E5-2670 v3 (2.3GHz, 12C, 30M) 754 2S Xeon® E5-2660 v3 (2.6GHz, 10C, 25M) 689 2S Xeon® E5-2650 v3 (2.3GHz, 10C, 25M) 675 2S Xeon® E5-2667 v3 (3.2GHz, 8C, 20M) 653 2S Xeon® E5-2650L v3 (1.8GHz, 12C, 30M) 604 2S Xeon® E5-2640 v3 (2.6GHz, 8C, 20M) 578 2S Xeon® E5-2643 v3 (3.4GHz, 6C, 30M) 571 2S Xeon® E5-2630 v3 (2.4GHz, 8C, 20M) 561 2S Xeon® E5-2630L v3 (1.8GHz, 8C, 20M) 471 2S Xeon® E5-2620 v3 (2.4GHz, 6C, 15M) 466 2S Xeon® E5-2637 v3 (3.5GHz, 4C, 15M) 422 2S Opteron 6386 SE, 16C, 2.8GHz,140W 399 2S Opteron 6380, 16C, 2.5GHz, 115W 383 2S Xeon® E5-2623 v3 (3GHz, 4C, 10M) 378 2S Opteron 6378, 16C, 2.4GHz, 115W 371 2S Opteron 6376, 16C, 2.3GHz, 115W 364 2S Opteron 6348, 12C 2.8GHz, 115W 350 2S Opteron 6370P, 16C, 2.0GHz, 99W 341 2S Opteron 6344, 12C 2.6GHz, 115W 333 2S Opteron 6366 HE, 16C, 1.8GHz, 85W 321 2S Xeon® E5-2609 v3 (1.9GHz, 6C, 15M) 321 2S Opteron 6338P, 12C, 2.3GHz, 99W 314 2S Opteron 6328, 8C 3.2GHz, 115W 290 2S Xeon® E5-2603 v3 (1.6GHz, 6C, 15M) 280 2S Opteron 6320, 8C, 2.8GHz, 115W 269 2S Opteron 6308, 4C 3.5GHz, 115W 171 0 100 200 300 400 500 600 700 800 900 1,000 SPECfp*_rate_base2006 Source: published results on www.spec.org as of Sept 29, 2014. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance Intel does not control or audit the design or implementation of third party benchmark data or Web sites referenced in this document. Intel encourages all of its customers to visit the referenced Web sites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. *Other names and brands may be 94 claimed as the property of others. Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Hadoop Performance

95 Performance Benchmarking Options for Big Data

HiBench: Micro Benchmarks . Synthetic, Illustrative benchmarks suitable for Regression testing

BigBench: End-to-End Benchmark

. Real-world, Instructive benchmarks Supply Chain Merchandising suitable for End-to-End testing Inventory management Assortment optimization Price comparison Pricing optimization . Open source standards based*

Operations Marketing . Industry consortium approved* Performance Customer segmentation . Support from 10+ ecosystem partners* Returns analysis Sentiment analysis Consumer experience * BigBench proposal accepted by TPC-TC for consideration

Increasing industry momentum toward evolving BigBench* 96 What is BigBench?

Evaluate end-to-end Big Data solution for retail scenario

. Volume – scale factor for large data set sizes

. Variety – semi/un/structured

. Velocity – periodic refresh of all data types Contains set of 30 queries targeting key retail concerns

. Performance measured by query time to completion Open source & available for download

97 Retail Analytics Areas Targeted by BigBench Queries

Supply Chain Merchandising Analyze Inventory management Assortment optimization impact of Price comparison Pricing optimization price change on sales Cross reference Marketing negative Operations customer Customer segmentation Performance reviews with Sentiment analysis stores with Returns analysis QoQ decline Consumer experience in sales

98 Scalable Hadoop Performance Solutions on Intel® Xeon® E5 v3 Platforms

Better Best Baseline Good Upgrade storage: • Add memory & • Intel® Xeon® E5 v2 Upgrade processor: • Add PCIe SSD upgrade Product Family • Intel® Xeon® E5 v3 Product with HDDs processor to E5-2699 v3 • Hard Disk Drives Family • Upgrade for storage storage to all PCIe SSDs

Analysis using BigBench provided on 10-node cluster running Cloudera Enterprise

99 Scalable Hadoop Performance on Intel® Xeon® E5 v3 Platforms Faster Insight on How to Improve Inventory Management

BigBench Query 20: Analysis of Customer Purchase & Return Habits 2 Higher is better 1.78 1.8 1.6 1.47 1.4 1.32 1.2 1 1

e Performance 0.8 0.6 0.4 Relativ 0.2 0 Baseline Good Better Best E5-2680 v2 E5-2697 v3 E5-2697 v3 E5-2699 v3 (10C, 2.8GHz, 115W) (14C, 2.6GHz,145W) (14C, 2.6GHz, 145W) (18C, 2.3GHz, 145W) 128GB memory 128GB memory 128GB memory 192GB memory HDDs HDDs HDDs + 1 PCIe SSD 2 PCIe SSDs

Source as of August 2014: Intel internal measurements on configurations shown on next slide. Query20 run time in seconds for each config: baseline = 746; good = 551; better = 516, best = 435. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance 100 *Other names and brands may be claimed as the property of others. Scalable Hadoop Performance on Intel® Xeon® E5 v3 Platforms Faster Results on How to Target Marketing to Customer Base

BigBench Query 30: Affinity Analysis of Products Commonly Viewed Together 2.5 Higher is better 2.18 2 1.68 1.44 1.5 1 1

0.5 Relative Relative Performance

0 Baseline Good Better Best E5-2680 v2 E5-2697 v3 E5-2697 v3 E5-2699 v3 (10C, 2.8GHz, 115W) (14C, 2.6GHz,145W) (14C, 2.6GHz, 145W) (18C, 2.3GHz, 145W) 128GB memory 128GB memory 128GB memory 192GB memory HDDs HDDs HDDs + 1 PCIe SSD 2 PCIe SSDs

Source as of August 2014: Intel internal measurements on configurations shown on next slide. Query30 run time in seconds for each config: baseline = 2,818; good = 1,951; better = 1,675, best = 1,290. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance 101 *Other names and brands may be claimed as the property of others. Scalable Hadoop Performance on Intel® Xeon® E5 v3 Platforms Optimize Pricing Decisions

BigBench Query 16: Analysis of Sales Impact of Price Change 2.5 Higher is better 2.17 2 1.52 1.62 1.5 1 1

0.5 Relative Relative Performance

Source as of August 2014: Intel internal measurements on configurations shown on next slide. Query16 run time in seconds for each config: baseline = 878; good = 577; better = 541, best = 403. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance 102 *Other names and brands may be claimed as the property of others. Scalable Hadoop Performance Configuration Details on BigBench Source: Intel internal measurements as of August 2014 on BigBench. https://github.com/intel-hadoop/Big-Bench.

Baseline configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208GZ4GC with two Intel® Xeon® processor E5-2680 v2 (10C, 2.8 GHz, 25 M, 115 W), 128 GB DDR3 memory, Intel® 82599ES 10Gb Ethernet Controller, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 12 x 2 TB Seagate Constellation* 7200 RPM HDDs for HDFS, Cent OS 6.5*, CDH5*.

Good configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208WTTYS with two Intel® Xeon® processor E5-2697 v3 (14C, 2.6 GHz, 35 M, 145 W), 128 GB memory (8x16GB DR-RDIMM DDR4-2133), Intel® Ethernet Controller X540-AT2, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 12 x 2 TB Seagate Constellation* 7200 RPM HDDs for HDFS, Cent OS 6.5*, CDH5*.

Better configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208WTTYS with two Intel® Xeon® processor E5-2697 v3 (14C, 2.6 GHz, 35 M, 145 W), 128 GB memory (8x16GB DR-RDIMM DDR4-2133), Intel® Ethernet Controller X540-AT2, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 12 x 2 TB Seagate Constellation* 7200 RPM HDDs for HDFS & 1 x 2 TB Intel® SSD DC P3700 for temp drive, Cent OS 6.5*, CDH5*.

Best configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208WTTYS with two Intel® Xeon® processor E5-2699 v3 (18C, 2.3 GHz, 45 M, 145 W), 192 GB DDR4 memory, Intel® Ethernet Controller X540-AT2, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 1 x 2 TB Intel® SSD DC P3700 for HDFS &1 x 2 TB Intel® SSD DC P3600 for temp drive, Cent OS 6.5*, CDH5*.

103 Scalable Hadoop Performance Solutions on Intel® Xeon® E5 v3 Platforms with Cloudera Enterprise

Best Baseline Good Better • Add memory & • Intel® Xeon® E5 v2 Upgrade processor: Upgrade storage: upgrade Product Family • Intel® Xeon® E5 v3 • Add Intel® PCIe Product Family SSD with HDDs processor to E5-2699 v3 • Hard Disk Drives • Upgrade for storage storage to all PCIe SSDs

Up to 1.42x Up to 1.58x Up to 2x performance gain performance gain performance gain vs. Baseline vs. Baseline vs. Baseline

Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance *Other names and brands may be claimed as the property of others. See next slide for configuration details. 104 Drive Big Business with your Big Data See full infographic

Accelerate big data analytics to extract actionable insights faster with: . Intel® Xeon® E5-2600 v3 Family . Intel® SSD Data Center for PCIe . Intel Ethernet Controllers 10Gb/40Gb . Cloudera Enterprise BigBench analysis completed on 10- node cluster processing 2TB of data in each configuration

4Intel internal measurements as of August 2014 on BigBench. BigBench query20: return analysis completion times for baseline configuration: 1,068 seconds; new configuration: 599 seconds. Details on next slide. 5Intel internal measurements as of August 2014 on BigBench. BigBench query30: affinity analysis completion times for baseline configuration: 2,818 seconds; new configuration: 1,290 seconds. Details on next slide. 6Intel internal measurements as of August 2014 on BigBench. BigBench query16: pricing analysis completion times for baseline configuration: 878 seconds; new configuration: 403 seconds. Details on next slide. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. Any difference in system hardware or software design or configuration may affect 105 actual performance. For more information go to http://www.intel.com/performance Drive Big Business with your Big Data Configuration Details Source: Intel internal measurements as of August 2014 on BigBench. https://github.com/intel- hadoop/Big-Bench.

Baseline configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208GZ4GC with two Intel® Xeon® processor E5-2680 v2 (10C, 2.8GHz, 25M, 115 W), 128 GB DDR3 memory, Intel® 82599ES 10Gb Ethernet Controller, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 12 x 2 TB Seagate Constellation* 7200 RPM HDDs for storage, Cent OS 6.5*, CDH5*.

Best configuration: 2 TB data set on 10-node cluster with 1 name node/resource manager, 1 Cloudera* services node, and 8 data nodes. Each node configured as Intel® Server System R2208WTTYS with two Intel® Xeon® processor E5-2699 v3 (18C, 2.3 GHz, 45M, 145 W), 192 GB DDR4 memory, Intel® Ethernet Controller X540-AT2, 1 x 800 GB Intel® SSD DC S3500 for OS drive, 1 x 2 TB Intel® SSD DC P3700 for HDFS & 1 x 2 TB Intel® SSD DC P3600 for temp drive, Cent OS 6.5*, CDH5*.

106 Return to Table of Contents

Software + Hardware Upgrade – 2x2s

107 Software + Hardware Upgrade – 2x2s

ISV solutions that work better together on Intel® Xeon® E5 v3 Product Family vs. 4-year old install base: . Operating system . Microsoft Windows Server & Red Hat Enterprise Linux . Virtualization . VMWare ESXi & Red Hat Enterprise Linux (KVM) . Database: Microsoft SQL . Enterprise Resource Planning with SAP ERP . Oracle Java Public - Data Center Optimization Deck

108 Software + Hardware Upgrade – 2x2s Detailed Configurations Operating System + System Power Virtualization . Microsoft Windows . VMWare . Server 2003 + WSM (X5690) . ESXi 4.1 + WSM (X5690) . Server 2012 R2 + HSW (E5-2699 v3) . ESXi 5.5 + HSW (E5-2699 v3) . Linux – Red Hat . Linux – Red Hat . RHEL 6.1 + WSM (X5690) . KVM (RHEL6.4) + SNB (E5-2690) . RHEL7 + HSW(E5-2699 v3) . KVM (RHEL7) + HSW (E5-2699 v3)

109 Software + Hardware Upgrade – 2x2s Detailed Configurations Database . Microsoft SQL 2008 + WSM (X5690) . Microsoft SQL 2012 + SNB (E5-2690) . Microsoft SQL 2014 + HSW (E5-2699 v3) SAP & IBM DB2 . SAP ERP 6.0 EHP4 + IBM DB2 9 + WSM (X5680) . SAP ERP 6.0 EHP5 + IBM DB2 10 + HSW (E5-2699 v3) Oracle Java . Java Standard Edition 7 + WSM (X5690) . Java Standard Edition 8 + HSW (E5-2699 v3)

110 Return to Table of Contents

Intel® Xeon® Processor E5-2600 v3 Product Family Performance Tuning Recommendations

111 Intel® Xeon® Processor E5-2600 v3 Product Family BIOS Knob Settings for Snoop Mode Configurations

BIOS Knob Snoop Mode Configuration Cluster on Die Enable Early Snoop Mode Disable Enable Early Snoop (ES) Auto Auto (Intel BIOS Default) Auto Enable Disable Auto Disable Disable Home Snoop (HS) Auto Disable Enable Disable Cluster on Die (COD) Enable Auto Not Supported - Invalid Enable Enable settings

112 Full performance tuning guide in Doc#550053 Intel® Xeon® Processor E5-2600 v3 Product Family BIOS Knob Tuning Recommendations based on E5-2697 v3 Energy Snoop Intel® HT Intel® Turbo Boost Energy/ Performance NUMA Cluster on Early Snoop Benchmark Prefetchers Efficient Mode Technology Technology Bias Mode Die Enable Mode Turbo Config Balanced Linpack* Enable Disable Enable Enable Enable Enable Disable COD Performance Balanced mWeb Enable Enable Enable Enable Enable Disable Enable ES Performance Balanced Memcached* Enable Enable Enable Enable Enable Disable Enable ES Performance Balanced SAP SD* 2-Tier Enable Enable Enable Enable Enable Enable Disable COD Performance SPECCPU*_rate Balanced Enable Enable Enable Enable Enable Enable Disable COD _base2006 Performance SPECCPU*_ Balanced Enable Enable Enable Enable Enable Disable Disable HS base2006 Performance SPECjbb*2013 Balanced Enable Enable Enable Enable Enable Disable Enable ES MultiJVM Performance SPEC Balanced jEnterprise* Enable Enable Enable Enable Enable Enable Disable COD Performance 2010 SPECmpiM* Balanced Enable Enable Enable Enable Enable Disable Disable HS 2007 Performance SPECompG* Balanced Enable Enable Enable Enable Enable Disable Disable HS 2012 Performance SPECpower*_ Enable DCU IP, Balanced Enable Enable Enable Enable Enable Disable COD ssj2008 Disable others Performance SPECvirt*_sc2013 Balanced Enable Enable Enable Enable Enable Disable Enable ES (using ESXi) Performance Balanced STREAM Enable Enable Enable Enable Enable Enable Disable COD Performance Balanced TPC-C* Enable Enable Enable Enable Enable Disable Enable ES Performance Balanced TPC-E* Enable Enable Enable Enable Enable Disable Enable ES Performance 113 Return to Table of Contents

Back Up

114 Non-AVX Turbo Boost 2.0 Frequency Bin upside by SKU† Intel® Xeon® Processor E5-2600 v3 product family: Advanced, Standard, & Basic SKUs

Processor Base Maximum Frequency in GHz (+ x00 MHz over base frequency) Cache SKU Frequency Cores (MB) (GHz) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 E5-2690 v3 2.6 12 30 +9 +9 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 n/a n/a n/a n/a n/a n/a E5-2680 v3 2.5 12 30 +8 +8 +6 +5 +4 +4 +4 +4 +4 +4 +4 +4 n/a n/a n/a n/a n/a n/a E5-2670 v3 2.3 12 30 +8 +8 +6 +5 +4 +3 +3 +3 +3 +3 +3 +3 n/a n/a n/a n/a n/a n/a E5-2660 v3 2.6 10 25 +7 +7 +5 +4 +3 +3 +3 +3 +3 +3 n/a n/a n/a n/a n/a n/a n/a n/a E5-2650 v3 2.3 10 25 +7 +7 +5 +4 +3 +3 +3 +3 +3 +3 n/a n/a n/a n/a n/a n/a n/a n/a E5-2640 v3 2.6 8 20 +8 +8 +6 +5 +4 +3 +2 +2 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2630 v3 2.4 8 20 +8 +8 +6 +5 +4 +3 +2 +2 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2620 v3 2.4 6 15 +8 +8 +5 +4 +3 +2 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2609 v3 1.9 6 15 Intel® Turbo Boost Technology not supported E5-2603 v3 1.6 6 15 Intel® Turbo Boost Technology not supported

†Max Turbo Boost Frequency based on number of 100 MHz increments above marked frequency (+1 = + 100 MHz, etc.)

Turbo frequencies can also be found in the Intel® Xeon® Processor E5-1600 & 2600 v3 Product Families Specification Update

115 AVX Turbo Boost 2.0 Frequency Bin upside by SKU† Intel® Xeon® Processor E5-2600 v3 product family: Advanced, Standard, & Basic SKUs

Processor SKU Maximum Frequency in GHz (+ x00 MHz over base frequency) AVX Cache Base Cores (MB) Freq (GHz) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

E5-2690 v3 2.3 12 30 +9 +9 +7 +7 +7 +7 +7 +7 +7 +7 +7 +7 n/a n/a n/a n/a n/a n/a E5-2680 v3 2.1 12 30 +10 +10 +8 +7 +7 +7 +7 +7 +7 +7 +7 +7 n/a n/a n/a n/a n/a n/a E5-2670 v3 2.0 12 30 +9 +9 +7 +6 +6 +6 +6 +6 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a E5-2660 v3 2.2 10 25 +9 +9 +7 +7 +7 +7 +7 +7 +7 +7 n/a n/a n/a n/a n/a n/a n/a n/a E5-2650 v3 2.0 10 25 +8 +8 +6 +6 +6 +6 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a n/a n/a E5-2640 v3 2.2 8 20 +12 +12 +10 +9 +8 +7 +6 +6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2630 v3 2.1 8 20 +11 +11 +9 +8 +7 +6 +5 +5 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2620 v3 2.1 6 15 +11 +11 +8 +7 +6 +5 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2609 v3 1.9 6 15 Intel® Turbo Boost Technology not supported E5-2603 v3* 1.3 6 15 Intel® Turbo Boost Technology not supported

*Note: SKUs without Turbo enabled (e.g. E5-2603 v3) can still operate between the non-AVX base frequency and the AVX base frequency †Max Turbo Boost Frequency based on number of 100 MHz increments above marked frequency (+1 = + 100 MHz, etc.)

Turbo frequencies can also be found in the Intel® Xeon® Processor E5-1600 & 2600 v3 Product Families Specification Update

116 Non-AVX Turbo Boost 2.0 Frequency Bin upside by SKU† Intel® Xeon® Processor E5-2600 v3 product family: Segment Optimized SKUs

Processor SKU Base Maximum Frequency in GHz (+ x00 MHz over base frequency) Cache Frequency Cores (MB) (GHz) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 E5-2699 v3 2.3 18 45 +13 +13 +11 +10 +9 +8 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 +5 +5 E5-2698 v3 2.3 16 40 +13 +13 +11 +10 +9 +8 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 n/a n/a E5-2697 v3 2.6 14 35 +10 +10 +8 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 +5 n/a n/a n/a n/a E5-2695 v3 2.3 14 35 +10 +10 +8 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 +5 n/a n/a n/a n/a E5-2687W v3 3.1 10 25 +4 +4 +2 +1 +1 +1 +1 +1 +1 +1 n/a n/a n/a n/a n/a n/a n/a n/a E5-2685 v3 2.6 12 30 +7 +7 +5 +4 +3 +2 +2 +2 +2 +2 +2 +2 n/a n/a n/a n/a n/a n/a E5-2683 v3 2.0 14 35 +10 +10 +8 +7 +6 +5 +5 +5 +5 +5 +5 +5 +5 +5 n/a n/a n/a n/a E5-2667 v3 3.2 8 20 +4 +4 +2 +2 +2 +2 +2 +2 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2650L v3 1.8 12 30 +7 +7 +5 +4 +3 +3 +3 +3 +3 +3 +3 +3 n/a n/a n/a n/a n/a n/a E5-2643 v3 3.4 6 20 +3 +3 +2 +2 +2 +2 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2637 v3 3.5 4 15 +2 +2 +1 +1 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2630L v3 1.8 8 20 +11 +11 +8 +7 +6 +5 +4 +3 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2623 v3 3.0 4 10 +5 +5 +3 +3 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a

†Max Turbo Boost Frequency based on number of 100 MHz increments above marked frequency (+1 = + 100 MHz, etc.)

Turbo frequencies can also be found in the Intel® Xeon® Processor E5-1600 & 2600 v3 Product Families Specification Update

117 AVX Turbo Boost 2.0 Frequency Bin upside by SKU† Intel® Xeon® Processor E5-2600 v3 product family: Segment Optimized SKUs

Processor SKU AVX Maximum Frequency in GHz (+ x00 MHz over base frequency) Base Cache Cores Freq (MB) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 (GHz) E5-2699 v3 1.9 18 45 +14 +14 +12 +11 +10 +9 +8 +7 +7 +7 +7 +7 +7 +7 +7 +7 +7 +7 E5-2698 v3 1.9 16 40 +14 +14 +12 +11 +10 +9 +8 +7 +6 +6 +6 +6 +6 +6 +6 +6 n/a n/a E5-2697 v3 2.2 14 35 +11 +11 +9 +8 +7 +7 +7 +7 +7 +7 +7 +7 +7 +7 n/a n/a n/a n/a E5-2695 v3 1.9 14 35 +11 +11 +9 +8 +7 +7 +7 +7 +7 +7 +7 +7 +7 +7 n/a n/a n/a n/a E5-2687W v3 2.7 10 25 +8 +8 +6 +5 +5 +5 +5 +5 +5 +5 n/a n/a n/a n/a n/a n/a n/a n/a E5-2685 v3 2.2 12 30 +10 +10 +8 +7 +6 +6 +6 +6 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a E5-2683 v3 1.7 14 35 +10 +10 +8 +8 +8 +8 +8 +8 +8 +8 +8 +8 +8 +8 n/a n/a n/a n/a E5-2667 v3 2.7 8 20 +8 +8 +6 +6 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2650L v3 1.5 12 30 +8 +8 +6 +6 +6 +6 +6 +6 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a E5-2643 v3 2.8 6 20 +7 +7 +6 +6 +6 +6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2637 v3 3.2 4 15 +4 +4 +3 +3 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2630L v3 1.5 8 20 +14 +14 +11 +10 +9 +8 +7 +6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a E5-2623 v3 2.7 4 10 +8 +8 +6 +6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a

†Max Turbo Boost Frequency based on number of 100 MHz increments above marked frequency (+1 = + 100 MHz, etc.) Turbo frequencies can also be found in the Intel® Xeon® Processor E5-1600 & 2600 v3 Product Families Specification Update 118 119