Broadwell Skylake Next Gen* NEW Intel NEW Intel NEW Intel Microarchitecture Microarchitecture Microarchitecture

Home , CNVi, Coffee Lake, Goldmont, Intel Hub Architecture, Xeon Phi

15 лет доступности

IOTG is extending the product availability for IOTG roadmap products from a minimum of 7 years to a minimum of 15 years when both processor and chipset are on 22nm and newer process technologies. - Xeon Scalable (w/ chipsets) - E3-12xx/15xx v5 and later (w/ chipsets) - 6th gen Core and later (w/ chipsets) - Bay Trail (E3800) and later products (Braswell, N3xxx) - Atom C2xxx (Rangeley) and later - Не включает в себя Xeon-D (7 лет) и E5-26xx v4 (7 лет)

2 IOTG Product Availability Life-Cycle

15 year product availability will start with the following products: Product Discontinuance • Intel® Xeon® Processor Scalable Family codenamed Skylake-SP and later with associated chipsets Notification (PDN)† • Intel® Xeon® E3-12xx/15xx v5 series (Skylake) and later with associated chipsets • 6th Gen Intel® Core™ processor family (Skylake) and later (includes Intel® Pentium® and Celeron® processors) with PDNs will typically be issued no later associated chipsets than 13.5 years after component • Intel Pentium processor N3700 (Braswell) and later and Intel Celeron processors N3xxx (Braswell) and J1900/N2xxx family introduction date. PDNs are (Bay Trail) and later published at https://qdms.intel.com/ • Intel® Atom® processor C2xxx (Rangeley) and E3800 family (Bay Trail) and late

Last

7 year product availability Time Last Last Order Ship Last

15 year product availability Time Last Last Order Ship

L-1 L L+1 L+2 L+3 L+4 L+5 L+6 L+7 L+8 L+9 L+10 L+11 L+12 L+13 L+14 L+15 Years

Introduction of component family † Intel may support this extended manufacturing using reasonably Last Time Order/Ship Periods Component family introduction dates are feasible means deemed by Intel to be appropriate. Future business, Publication of the PDN will typically be available from your Intel Sales Representative. manufacturing, or market conditions may change, which may cause followed by: 12 month last time order Refresh SKU offerings launched after the Intel to revisit the EOL/PDN timing. Intel may make changes to period, followed by 6 month last time component family will follow the original dates and schedule at any time without notice ship period. component family introduction date.

3 Какие модели будут доступны 15 лет? https://ark.intel.com/

4 Зачем использовать embedded parts?

- Долгая доступность (15 лет) - Разные т.н. USE CONDITIONS:

Use Conditions Temp Range Reliability DPM

Automotive 15Y Reliability at 11.4% Active, • Advanced Driver Assistance Systems (ADAS) Automotive Grade 2, T = -40⁰ to 105⁰ C <10 la Tj-max = 110o - 125o C, • Software Defined Cockpit (SDC)

10Y Reliability at 100% active, CT – Commercial Temp: T = 0o to 70o C Industrial la Tj-max = 100o - 110o C <500 ET – Extended Temp: T = -40o to 85o C la Base Frequency only – No core or graphics turbo

o o CT – Commercial Temp: Tla = 0 to 70 C 10Y Reliability at 100% active, Communications o o o o <500 ET – Extended Temp: Tla = -40 to 85 C Tj-max = 100 - 105 C

5Y Reliability at 80% Active, Embedded Broad Market CT – Commercial Temp: T = 0o to 70o C <500 la Tj-max = 100o - 105o C

5Y Reliability at 80% Active, Server/Enterprise CT – Commercial Temp: T = 0o to 70o C <500 la Tj-max sku dependent

5Y Reliability up to 30% Active, PC/Client/Tablet CT – Commercial Temp: T = 0o to 70o C <500 la Tj-max sku dependent,

5 АНАЛИТИКА, ОБЛАКА НОВЫЕ СЕТИ СУПЕРКОМПЬЮТЕРЫ, УМНЫЕ ДАННЫЕ

18% CAGR from SDN/NFV spend up to AI is the fastest growing 2017-20201 $157B through 20202 datacenter workload3

1. Source: IDC Q4’16 Cloud Infrastructure Tracker 2. Source: Technology Business Research, Sept 2015 3. Source: Amalgamation of Intel financials, analyst data and Intel analysis, Intel revenue includes FPGAs. 6 «Строительные блоки» для решения текущих задач

Analytics & Technical Intel Scalable Machine Learning Intel / Cloudera Computing System Framework Software

Intel® Data Intel® Cache Intel® Parallel Intel® Media Management, Center Manager Accelration SW Software Studio XE Server Studio Orchestration Intel® Virtual Development Intel® Intel® Intel® Intel® RSTe & Storage Gateway Tools Inspector VTune CoFluent™ Wind River

Intel ® AES-NI Intel® SHA Intel® MKL Library Intel® DPDK Intel® MPI Library Instructions Extensions SW & OpenGL/DX on Intel® Data Analytics Instructions … Intel® HD Graphics Acceleration Library (DAAL) … Intel ® AVX Intel ® TSX Libraries Instructions Instructions Intel ® TXT Intel® SPDK Intel® ISA-L Embree

Intel® QAT Intel® DLIA Intel® VCA Intel® Network Silicon Processors Chipsets Ethernet Omni-Path ASICs Photonics FPGAs SSDs Fabric Accelerators

7 Intel® Xeon® Scalable Processors ФУНДАМЕНТ ДАТАЦЕНТРА БУДУЩЕГО: ГИБКАЯ И НАДЕЖНАЯ АРХИТЕКТУРА

Производительность Оптимизировано под все нагрузки: ПРОСТОТА Там, где надо Compute, storage, Network ВНЕДРЕНИЯ

В среднемувеличение производительности в 1.6x раза против e5/e7 v4

6 * New features in bold

ПРОИЗВОДИТЕЛЬНОСТЬ ЯДРА ПРОИЗВОДИТЕЛЬНОСТЬ СОКЕТА Faster application response time Increased multi-tenancy & VM density for cloud More choices to optimize software license costs Application performance for parallelizable workloads ▪ New core micro-architecture ▪ More cores ▪ New lower latency cache memory hierarchy design ▪ New High-speed Interconnects = much greater inter-CPU bandwidth ПАМЯТЬ И ШИНА PCI-E УСКОРЕНИЕ СПЕЦ ВЫЧИСЛЕНИЙ High bandwidth connections for networking Application specific instructions and accelerators and memory performance ▪ Intel® AVX-512 (2X FLOPs/clock) for HPC, Analytics, Database, & ▪ More memory bandwidth via added memory channels Security ▪ Int. Intel® QuickAssist Technology: faster crypto & data compression ▪ Significantly more PCIe IO bandwidth versus prior gen ▪ Intel® FPGA: Big speedups on server/storage/networking algorithms ▪ Richer chipset & IO support options БЕЗОПАСНОСТЬ ГИБКОСТЬ Built in performant security for trusted infrastructure Optimized architecture for Hybrid-Cloud world ▪ Up to 2X security algorithm performance improvements ▪ Enhanced Intel® RunSure Technology ▪ Root of trust with Intel® TXT-OTA, Intel® PTT and Intel® Boot Guard ▪ Enhanced Intel® Virtualization Technology (VT-x): MBE & TSC ▪ New Key Protection Technology (w/ integrated Intel® QAT & Intel PTT) ▪ New Intel® VMD for improved management of NVMe SSD drives

• Mode-based Execute Control (MBE): • Timestamp Counter Scaling (TSC): 99 «СЕТЕВАЯ» архитектура: ▪ Максимальная производительность ▪ Гарантированно низкие задержки при передачи данных ▪ Оптимизировано для работы с общей памятью

▪ Масшабируется эффективно для 2, 4, & 8+ сокет конфигур МОНОЛИТНОЕ ЯДРО МОНОЛИТНОЕ Enhanced PCIe

Note: Not representative of top bin die

Много вычислительных ядер, быстрая память, малые задержки

10 2016 2017 2018

Intel® Xeon® Processor E7 Brickland Platform Purley Platform

High-end сервера с Skylake Cascade Lake масшабированием по сокетам и E7 v3 E7 v4 поддержкой RAS Intel® Xeon® PLATINUM Intel Xeon GOLD Intel® Xeon® Processor E5 Grantley-EP Platform Intel Xeon SILVER Сбалансированные 2х сокет E5 v3 E5-4600 v4 (4S) системы с лучшим соотнош Intel Xeon BRONZE performance/watt/$ E5 v3 E5-2600 v4

Общая платформа для построения E5 и e7 –class решений

11 INTEL® XEON® SCALABLE PROCESSOR The Foundation for Agile, Secure Workload-Optimized Hybrid Cloud INTEL® XEON® PLATINUM 81xxPROCESSORS • Most cores (28C, 56T) • Highest frequency (up to 3.6 GHz, 4C) • Most Socket flexibility (2, 4, 8+) • Elite IO/memory (48 PCIe 3.0 lanes, 6 memory channels) • UP TO Most/fastest 3SCALABLE UPI (Ultra PERFORMANCE Path Interconnects, 10.4SCALABLE GT/s) PERFORMANCE SCALABLE PERFORMANCE 28 CORES • Fastest MemoryMEMORY (DDR4 PERFORMANCE-2666 MHz) BETTER AT LOW POWER HARDWARE-ENHANCED SECURITY •GREATHighest Intel®ADVANCED AVX-512 RAS with 2 FMA per core STANDARD RAS ENTRY STANDARD RAS • Intel® Turbo Boost and Intel® Hyper-Threading Technology UP SOCKET WITH UPI • New Advanced RAS TO 2, 4 & UP TO 8 SUPPORT 3 LINKS • NodeWORKLOAD Controller-OPTIMIZED Support to assist in scaled node management MODERATE TASKS DEDICATED TASKS M TOPLINE MEMORY DDR4 H WITH RESULT: TASKS 2666 Z UP TO 1.5 TB CHANNEL BANDWIDTH • Best workload-optimized performance for general purpose compute across virtualized environmentsPERFORMANCE FOR GENERAL and hybrid-PURPOSE cloud deploymentsINTEL® TURBO BOOST TECHNOLOGY AND AFFORDABLE, ENTRY PERFORMANCE • INTEL® HYPER-THREADING TECHNOLOGY ACCELERATOR BestCOMPUTE, choice STORAGE for AND mission NETWORKING-critical, traditional DC apps, real-time analytics,FOR and LIGHT AI/deep-RANGE WORKLOADS HIGHEST THROUGHPUT learning inference workloads; smart pathFOR for MODERATE DL training WORKLOADS • Best performance for the most demanding storage and networking workloads MAINSTREAM RAS: Reliability, Availability, and Serviceability

12 INTEL® XEON® SCALABLE PROCESSOR The Foundation for Agile, Secure Workload-Optimized Hybrid Cloud

INTEL® XEON® GOLD 61xxPROCESSORS • 22C (44 threads w/ HT) • 2 and 4S • Up to 3.4 GHz, (6C) • 3 UPI links (10.4GT/s) • DDR4-2666 MHz • Intel® AVX-512 with 2 FMA • Node ControllerSCALABLE Support PERFORMANCE UP TO • Enhanced scalability and workload optimized performanceSCALABLE PERFORMANCE for UP TO 28 CORES 22 CORES AT LOW POWER HARDWARE-ENHANCED SECURITY BETTERgeneral purpose/virtualized computeENTRY & hybrid cloud • Improved performanceSTANDARD RAS for demanding storageSTANDARD and networking RAS UP SOCKET WITH UPI SOCKET TO 2, 4 & 8 SUPPORT UP TO 3 LINKS 2 & 4 SUPPORT INTEL® XEON® GOLD 51XXPROCESSORS DEDICATED TASKS • UpMODERATE to 14C (28T) TASKS M WITH TOPLINE MEMORY DDR4 H UP TO UPI LINKS • 2 UPI links - 2666 Z UP TO 1.5 TB CHANNEL BANDWIDTH 3 • DDR4-2400 MHz • INTEL®Intel® TURBO AVX BOOST-512 TECHNOLOGY with 1 FMA AND AFFORDABLE, ENTRY PERFORMANCE INTEL® HYPER-THREADING TECHNOLOGY ACCELERATOR RELIABILITY, AVAILABILITY • Affordable Advanced RAS and 4-socketFOR scalability LIGHT-RANGE WORKLOADS HIGHEST THROUGHPUT ADVANCED AND SERVICEABILITY • SuitableFOR MODERATE for a WORKLOADSwide range of workloads MAINSTREAM

13 INTEL® XEON® SCALABLE PROCESSOR

INTEL® XEON® SILVER 41xxPROCESSORS Up to 12C, 24T Up to 2.2 GHz (10C) DDR4-2400 MHz 2 UPI at 9.6 GT/s Intel® AVX-512 w/ 1 FMA Intel® Turbo Boost and Intel® HT for multi-threaded apps UP TO 28 CORES SuitableUP for TO a moderate22 CORES range of workloads UP TO 12 CORES UP TO 8 CORES

UP SOCKET WITH UPI INTEL® XEON® BRONZE 31XXSOCKET PROCESSORS SOCKET SOCKET TO 2, 4 & 8 SUPPORT UP TO 3 LINKS 2 & 4 SUPPORT 2 SUPPORT 2 SUPPORT Up to 8C Up to 1.7GHz (8C) M M DDR4 WITH TOPLINE MEMORY Entry 2-socket configuration DDR4 H PCIE H UP TO UP TO UPI LINKS 48 - 2666 Z 1.5 TB CHANNEL BANDWIDTH 3 DDR4-2133 MHz 2400 Z LANES 3.0 2 UPI at 9.6 GT/s Intel® AVX-512 w/ 1 FMA ACCELERATORReliability upgrade versus Intel® Xeon® processorRELIABILITY, AVAILABILITYE3 platform TURBO BOOST TECHNOLOGY RELIABILITY, AVAILABILITY HIGHEST THROUGHPUT ADVANCED AND SERVICEABILITY INTEL HYPER-THREADING TECHNOLOGY STANDARD AND SERVICEABILITY MAINSTREAM EFFICIENT ENTRY

14 2S Purley Platform Configurations

2S-2UPI Configuration with integrated Intel® Omni-Path Fabric

Intel® UPI SKL/ SKL/ CLX CLX

DMI ** x4 LBG 3x16 3x16 PCIe* PCIe*

1x100G 1x100G Intel® Omni-Path Fabric Intel® Omni-Path Fabric

DDR4 DIMMs DDR4/Apache Pass (Apache Pass support planned for Cascade Lake Server) ** PCIe* uplink connection for Intel® QuickAssist Technology and Intel® Ethernet Example DIMM population shown; please look up Apache Pass customer collateral for specific rules on DDR4/Apache Pass DIMM populations

15 4S & 8S Purley Platform Configurations 8S Configuration LBG LBG

4S-2UPI Configuration 4S-3UPI Configuration Intel® SKL/ UPI SKL/ CLX CLX

Intel® Intel® UPI UPI SKL/ SKL/ SKL/ SKL/ CLX CLX CLX CLX SKL/ SKL/ CLX CLX

LBG LBG LBG LBG

SKL/ SKL/ SKL/ SKL/ CLX CLX CLX CLX SKL/ SKL/ CLX CLX

** ** DMI DMI LBG 3x16 LBG 3x16 LBG 3x16 LBG 3x16 PCIe* PCIe* PCIe* PCIe*

SKL/ SKL/ DDR4 DIMMs CLX CLX DDR4/Apache Pass (Apache Pass support planned for Cascade Lake Server) LBG Optional PCIe* uplink connection for Intel® QuickAssist Technology and Intel® Ethernet DMI ** Example DIMM population shown; please look up Apache Pass customer collateral for specific rules on LBG 3x16 ** DDR4/Apache Pass DIMM populations LBG 3x16 PCIe* PCIe*

16 Product Numbering Convention for Intel® Xeon® Scalable Processors

Intel® Xeon® Platinum 8 1 # # a a processor Intel® Xeon® processor Intel® Xeon® Gold 6 1 # # a a processor E7 Family (4/8S+) Intel® Xeon® Gold 5 1 # # a a processor Intel® Xeon® processor Intel® Xeon® Silver 4 1 # # a a processor E5 Family (2S, 4S) Intel® Xeon® Bronze 3 1 # # a a processor Integrations and Optimizations (if applicable) • F = Fabric • T = High Tcase/Extended Reliability

Memory Capacity • No Suffix = 768GB per socket SKU Level • M = 1.5TB per socket • 8 = Platinum • 6, 5 = Gold • 4 = Silver Processor Generation Processor SKU • 3 = Bronze • 1 = 1st Gen (Skylake-SP) • (ex. 20, 34…)

17 28C Top Bin 205W 28C 165W Extended life (10-Year Use) + 2.5G 8180 2.1G 8176 NEBS-Friendly Thermal Spec • Highly flexible offerings optimized to 24C 205W 26C 165W 2.1G address broadest array of workloads 2.7G 8168 8170 24C 150W 2.1G 12C-24.75M 150W 26C 150W 8160T 3.0G 2.0G 8164 8158 20C 125W • Four primary SKU lineups: 2.0G 4C-16.50M 105W 24C 150W 6138T 3.6G 2.1G 8160 8156 16C 125W 1. Optimized for per core performance 20C 150W 16C 125W 2.1G 6130T 2. Balanced, energy efficient perf/W 2.4G 2.0G (8S entry) 8153 6148 12C-19.25M 125W 22C 140W 2.6G 6126T 3. Extended life 18C 200W 2.1G 3.0G 6154 6152 CORE PERFORMANCE CORE 14C 105W 4. Integrated Intel® Omni-Path - 20C 125W 2.2G 5120T 18C 165W 2.0G 2.7G 6138 6150 14C 85W 18C 140W 1.9G 16C 150W 2.3G 5119T • New 205W SKUs: 2.6G 6140 6142 12C 85W 16C 125W • e.g. 8180 hits 2.5 GHz at 28 cores 14C 140W 2.1G 4116T 2.1G 6130 2.6G 6132 14C 105W 10C 85W 12C--24.75M 165W 2.2G 2.2G 4114T • Great high frequency options: 3.2G 5120 6146 12C 105W 8C 70W 2.0G • e.g. 6144 w/ 8 cores at 3.5 GHz base 12C-24.75M 150W 2.3G 5118 4109T 3.0G 6136 10C 85W • SKUs w/ add’l cache per core vs. default 12C-19.25M 125W 2.4G 5115 2.6G 6126 12C 85W 8C-24.75M 150W 2.1G 4116 3.5G 6144 10C 85W 8C-24.75M 130W 2.2G 4114 3.2G 6134 4C 85W with integrated Intel® Omni-Path Architecture Fabric 2.6G 6C-19.25M 115W 4112 28C 173W 3.4G 6128 8C 85W 2.1G 8176F

OPTIMIZED FOR BALANCED ENERGY EFFICIENT PERF/W EFFICIENT ENERGY BALANCED FOR OPTIMIZED 2.1G 4C-16.50M 105W 4110

SKUs OPTIMIZED FOR HIGHEST PER HIGHEST FOR OPTIMIZED SKUs 20C 160W 24C 160W 3.6G 5122 8C 85W 2.4G 6148F 2.1G 8160F 1.8G 4108 16C 160W 20C 135W 8C 85W Core Count TDP 2.6G 6142F 2.0G 6138F 1.7G 3106 Base Non-AVX Core Frequency 12C-19.25M 135W 16C 135W 6C 85W 2.6G 2.1G 1.7G 6126F 6130F “M” SKU available (1.5 TB/SktSck)) SKU # 3104

2118 Volume Management Device (VMD)

Ethernet Controller w/ crypto accelerator & RDMA Новая платформа для 5g-ready сетей

26B IP networked devices/ Intel® Ethernet 700 Series Приватность данных Connections • One Architecture. Multiple Speeds • int. Intel Ethernet Connection X722 w/ by 2020 iWARP RDMA (4x10GbE) up to 2.55x higher • Advanced features IPSec (Gbps) • Optimized SW (DPDK) for NFV Int. Intel® QuickAssistTechnology • Crypto & Data Compression Шифрация передачи Intel® Omni-Path Architecture (Fabric) 2.3 ZB Global up to higher • High-bandwidth & low-latency fabric 2.49x IP Traffic for HPC clusters (100 Gb / sec) TLS Web Proxy (Gbps) Annually by 2020

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, SW, operations and functions. Any change to any of those factors may cause the results to vary. Consult other information/performance tests to assist you in fully evaluating your contemplated purchases, including perfoof 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 referenced benchmark data are accurate/reflect performance of systems available for purchase. 20 SMBus Integrated IO NC-SI MCTP Feature Support Customer Value Quad Port 10GbE MAC/PHY • Optimized for networking capability in • Based on Intel’s 10GbE solution Cloud, Comms, and Storage • Interfaces: 10G (KR, SFI, XFI), 1G (KX) • Single network driver on Intel® platform VFVF0 VFVF0 VFVF0 VFVF0 VF00 VF10 … 0 VFm0 Remote Direct Memory Access • Routable and scalable RDMA ideal for PF PF … PF • iWARP large segmented networks in private In-Band 0 1 n and public clouds Mgmt. PCIe* with SR-IOV Network Virtualization Offloads • Abstract the network for cloud Queue Mgmt & Scheduling • Flexible Filters (ATR, Flow Director) flexibility

• NVGRE, IPinGRE, VXLAN, MACinUDP • Enhanced programmability and Q0 Q1 ... Qm application affinity Protocol Engine

Standards Based Virtualization • Broad OS enablement VEB & DCB Traffic Classifiers • SR-IOV: 4 Physical/128 Virtual Function • VEB (Virtual Ethernet Bridge) 4x10G MACs Power Management • Energy Efficient Ethernet Manageability • Common transport for LAN-to-BMC 4x10GbE/1GbE • BMC Pass-Through (control & network) with RDMA(iWARP) • Interfaces: NC-SI, SMBus, and MCTP

21 Without With RDMA is a network performance optimization iWARP iWARP •Enables direct app-to-app communication across nodes •Bypasses the OS stack & kernel Application •Provides direct channel for remote memory application access

RDMA offers lower latency, high throughput by: IO Lib User •Avoiding application context switching Kernel OS Stack Avoid Copies •Placing data directly in application buffers (zero-copy DMA) & Kernel Sys Driver •Moving protocol processing off the CPU S/W H/W Intel® Ethernet Connection X722 is featured with iWARP (Internet Wide- PCIe area RDMA Protocol) RDMA: Basic iWARP iWARP Network TCP/IP Network •Recommended for easy deployment and configuration, scalability and Controller Controller congestion control Flow Ethernet Flow RDMA-aware applications or messaging layers are required •Windows*/Linux* Drivers supported for Storage, Messaging Middleware and HPC application categories 22 Intel® Ethernet Architecture Evolution

500 Series 700 Series 800 Series 900 Series Niantic Fortville Columbiaville** Gemini Mountain** Highly Advanced/ Customizable FPGA CPU Cores Proprietary Internal Fabric

Extensible

Pipeline Workload

Programmable Pipeline

Programmability Partially Programmable Pipeline

Fixed Function Fixed Pipeline Traditional Port Speed 10GbE 10/25/40GbE 10/25/50/100GbE 10/25/50/100/200GbE (General IT) Laying the Foundation for supporting a broader portfolio of network protocols

**These programs are in the planning stage. Features are subject to change. Intel Confidential 23 Intel® Ethernet Adapter Roadmap Summary Intel Ethernet 800 Series E810 (100GbE QSFP28) Enhanced Virtualization Intel Ethernet 700 Series and Network Overlays XL710 (40GbE QSFP+) XXV710 (25GbE SFP28) Network Virtualization Cloud and Network Overlays Acceleration Virtualization Overlays

X710 (10GbE SFP+) X722 (10GbE SFP+) X520 10GbE SFP+ Cloud and Network X710 (10GBASE-T) RDMA and Network World’s Best Selling Virtualization Overlays Quad-Port 10GBASE-T Virtualization Overlays 10GbE CNA X540 10GBASE-T X550 10GBASE-T EthernetAdapters World’s 1st 2nd Generation Single Chip 10GBASE-T Single Chip 10GBASE-T . . . 2014 2015 2016 2017 2018 2019

Avoton Broadwell DE Denverton Lewisburg Columbia Park

SoC Integrated Integrated Integrated Integrated 2x100GbE with RDMA

4x1/2.5GbE 2x10GbE 4x10GbE 4x10GbE with RDMA IP / / IP

Intel Confidential 24 Intel® VMD is a CPU-integrated device to aggregate NVMe SSDs into a storage volume and enables other storage services such as RAID • Intel® VMD is an “integrated end point” that Intel Intel Intel stops OS enumeration of devices under it VMD VMD VMD • Intel® VMD maps entire PCIe* trees into its own address space (a domain) • Intel® VMD driver sets up and manages the domain (enumerate, event/error handling), but out of fast IO path

Eliminates additional components to provide a full-feature storage solution

25 ПОДДЕРЖКА 24 NVMe на примере 2U 2S платформы

GRANTLEY to PURLEY CHANGE  2U 8x25 Combo HSBP now supports NVMe in all 8 Drive Slots  OCuLink being used for NVMe vs. miniSAS HD  More options for PCIe SSD AIC

Datacenter Group Intel Confidential – For Use Under NDA Only Released New release 17.10 SPDK Architecture Jan ‘18

iSCSI vhost-scsi NVMe-oF* vhost-blk Integration Storage Target Target Target Target Protocols SCSI NVMe RocksDB Block Device Abstraction (BDEV) BlobFS Logical Ceph Storage 3rd Party GPT Volumes Services Blobstore Linux Ceph Virtio scsi NVMe PMEM QEMU Async IO RBD (vhost) BDev QoS

NVMe Devices Core Intel® QuickData Application Drivers * * Technology NVMe-oF NVMe Framework Initiator PCIe Driver Driver

Data Center Group Intel Confidential – For Use Under NDA Only SPDK Community

• SPDK community webpage: http://www.spdk.io/community/ • Github : https://github.com/spdk/spdk - Centralized place for the code management • Trello : https://trello.com/spdk - Place to find SPDK’s ongoing or future work • GerritHub : https://review.gerrithub.io/#/q/project:spdk/spdk+status:open - SPDK Code Reviews • IRC: https://freenode.net/ we’re on #spdk – Post technical questions • Mailing List: https://lists.01.org/mailman/listinfo/spdk - Available for usage and development questions as well as design discussions

Data Center Group Intel Confidential – For Use Under NDA Only Новая платформа для гибридных облачныхSTORAGE инфраструктур • БыстрыеIntel® Optane™ DC ssds БЫСТРЫЙ И НАДЕЖНЫЙ емкиеintel® 3d nandDC SSds • ▪ Up to 60X better at 99% QoS1 • БольшеPCIeввода вывода ▪ Up to 40X faster response time1 • Intel® Storage ПО • Intel® Intelligent Storage Acceleration Library (ISA-L) • Intel® Storage Performance Development Kit (SPDK) • Intel® Platform Storage расширения 6x Optane SSD Up to 5.2x higher IOPS • Intel® Volume management device (intel® VMD) + Intel® SPDK Up to 3.3x lower latency • Intel® virtual raid on cPU(Intel® VROC)

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 complete information visit http://www.intel.com/performance Configurations: Tested by Intel as of Jun 2017; test versus out of box 4x NVMe SSD 29

31 ИТАК,

Революция в использовании данных меняет ВСЕ индустрии

Новые датацентры должны быть мощнее, безопаснее и более конфигурируемы, чтобы справиться с требованием бизнеса

Новая платформа представляет мощность, безопасность и конфигурируемость (compute / storage / network)

Общий сокет и стабильность платформы (Skylake / Cascade Lake / Ice Lake = один сокет LGA 3647) дает возможности масшабирования

32 Purley with Cascade Lake Overview Cascade Lake: Up to 28Core with Intel HT Technology Purley 2S Configuration CPU 48 lanes (Drop in to Purley Platform @ 70W-205W) DDR4/AEP DIMM 48 lanes DDR4/AEP DIMM PCIe PCIe

Frequency and architecture improvements and Cascade UPI Cascade New Capabilities DDR-T/Apache Pass DIMM Support on select SKUs Lake Lake FPGA FPGA

DMI DMI Socket Socket P LBG LBG DDR4 DIMM Scalability 2S, 4S, & glueless 8S (>8S via xNC support) DDR4 or AEP DIMM 6 channels DDR4 RDIMM and LRDIMM per CPU/ / 12 DIMMs per Purley 4S Configuration

socket, 16Gb DDR4 based DIMMs supported, up to 2666 MT/s DDR4/AEP DIMM 48 lanes 48 lanes DDR4/AEP DIMM Memory 2DPC, up to 2933 MT/s 1DPC and PCIe PCIe

Apache Pass DIMM Support on select SKUs Cascade UPI Cascade UPI Up to 3 links per CPU Lake Lake (Ultra Path Interconnect) x20, speed: 9.6 and 10.4 GTS

PCIe PCIe Gen 3: 48 lanes per CPU (bifurcation support: x16, x8, x4) LBG LBG Discrete Omni Path adapter (100Gb/s) [Integrated Fabric SKUs DDR4/AEP DIMM 48 lanes 48 lanes DDR4/AEP Host Fabric PCIe PCIe DIMM available on Skylake only] FPGA Integrated Intel Arria® 10 FPGA in multi-chip package Cascade UPI Cascade Lake Lake IE, Intel QAT, eSPI, Integrated Intel Ethernet Connection: up to PCH – Lewisburg 4x10Gb/1Gb ports, Up to 14 SATA 3, Up to 14 USB 2.0, Up to 10 DMI DMI USB 3.0, Up to 20 ports PCIe* 3.0 (8 GT/s) LBG LBG New capabilities/Changes relative to Skylake/Purley in Bolded DDR4 or AEP DIMM DDR4 DIMM Blue Data Center Group Intel Confidential – For Use Under NDA Only 33 Apache Pass Production System Requirements

NEW NEW NEW Cascade BIOS with Purley Platform Lewisburg Lake CPU Reference PCH Platform (AEP-enabled Code Apache Pass stepping) Requirements

• No board changes • Provides cache & required1 memory control

Software NEW NEW Block Driver and Persistent Memory 2 Requirements Management Libraries Aware File System

1. No board changes required assuming OEM system was designed to be compliant with PDG Guidelines 2. Dependent on operating mode

Data Center Group Intel Confidential – For Use Under NDA Only 34 Cascade Lake VNNI Instruction Set

Vector Neural Network Instruction (VNNI) on Cascade Lake will accelerate Intel® Xeon® Scalable processor performance on Deep Learning and AI workloads • VNNI : A new set of Intel® Advanced Vector Extension (Intel® AVX-512) instructions for Intel® Xeon® Platinum 8000 and Gold 6000 series processors • 8-bit (int8) new instruction • 16-bit (int16) new instruction • No hardware changes are required to support VNNI on Purley Platform • Minimal OS/VMM enabling if Intel® AVX-512F (foundation) state pre-exists • SW development support will be enabled in MKL-DNN (Math Kernel Library – Deep Neural Networks)

Data Center Group Intel Confidential – For Use Under NDA Only 35 Datacenter ai

Intel® Stratix® 10 FPGA Intel® Xeon® Processor Scalable Family Intel® Nervana™ Neural Network Processor* Flexible acceleration Foundation for AI Deep learning by design Accelerate the widest Begin your journey with Accelerate the most intensive range of AI and other deep learning deployments with workloads & configurations the AI you need this custom-built processor on the chip you know

*Future product that was formerly codenamed the “Crest family” All performance positioning claims are relative to other processor technologies in Intel’s AI datacenter portfolio All products, computer systems, dates, and figures specified are preliminary based on current expectations, and are subject to change without notice.

Data Center Group Intel Confidential – For Use Under NDA Only Intel® FPGA Data Center Platform Options Enabled By The Acceleration Stack for Intel® Xeon® CPU with FPGAs PCIe Acceleration Cards Server Platform Option with In-Package FPGA

▪ System flexibility with Intel Xeon CPU SKU options ▪ Coherent interface benefits software developers ▪ Dedicated local memory ▪ Superior performance for bandwidth & latency sensitive applications ▪ Can be slotted into 1U servers

Choose the Intel FPGA platform matched to your application needs

1 Data Center Group UPI = Intel® Ultra Path Interconnect Intel Confidential – For Use Under NDA Only 37 2HSSI = High Speed Serial Interface Learn More, Get Started Today

Resource Location Intel® Acceleration Solutions Site www.intel.com/accelerators

Intel® FPGA Acceleration Hub www.intel.com/fpgaaccelerationhub

Learn more about the Acceleration Stack www.altera.com/solutions/acceleration- hub/acceleration-stack.html Start developing for Intel FPGAs using OPAE 01.org/OPAE

(Video) “Getting Started With Open www.brighttalk.com/webcast/10773/275799 Programmable Acceleration Engine” (Whitepaper) “The Open Programmable 01.org/sites/default/files/downloads/opae/open- Acceleration Engine (OPAE)” programmable-acceleration-engine-paper.pdf

Data Center Group Intel Confidential – For Use Under NDA Only 38 Xeon D

Broadwell (-DE, -DE NS) Skylake (-D)

39 Intel® Xeon® Processor D Workload Positioning

Enterprise Public Cloud HPC Big Data Storage Network

Data Mining Search CAE, Real Time Media Modeling Analytics Processing Simulation

Business Virtualized EDA Batch Scale-out / Deep Packet Processing Hosting Analytics distributed Inspection database Collaboration, Content Digital No SQL Enterprise Edge Routing Email Delivery Content database SAN / NAS Creation

Application Dynamic Financial Data caching Hot Storage Edge Development Web Serving Risk Security / Scales to low TDP, high Analysis Firewall performance, 10 Data center GbE and PSE edge File & Print, Dedicated Gov’t / Scale up and Warm Wireless workloads: Security, Sys Hosting Defense / distributed Storage Access High density at low power Management Academic database resulting in good TCO Dynamic Web Memory Geo Large object Cold Storage Wireless Serving Caching Sciences; store Base Station Siesmic

Intel Confidential Intel and the Intel logo are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. Other names and brands may be claimed as the property of others. All products, dates, and figures are preliminary and are subject to change without any notice. Copyright © 2015, Intel Corporation. 40 Intel® Xeon® Processor D Product Cadence 22nm Process Technology 14nm Process Technology 10nm Process Technology

Haswell Broadwell Skylake Next Gen* NEW Intel NEW Intel NEW Intel Microarchitecture Microarchitecture Microarchitecture

Future Future Xeon®-D *

Skylake-D Q1 ’18 100G QAT**, 4x10GbE

Broadwell-DE-NS Q3 ’17 Broadwell-DE + 40G QAT** + 2x10GbE

Q1 ’15, Q4 ‘15 Broadwell-DE • *Product in Planning • ** Intel® QuickAssist Technology Density Integration Acceleration

Datacenter Group Intel® Confidential – For Use Under NDA Only 41 Intel® Xeon® Processor D-1500 Platform Overview

CPU Intel® Xeon® Processor • Up to 16C BDW Xeon© SMT (14nm) До 16 ядер D-1500 Family • Targeted TDP ~20W - 65W • Intel® Xeon® Features Intel® x8 PCIE2 Communications До 2.7GHz Chipset 89xx Series Memory (Coleto Creek) • 2 Memory Channels, Broadwell x24 PCIE3 2 DIMMS/channel • DDR3L/DDR4 Cores • 128GB Max Capacity

x6 SATA 3 IOs

Advanced Advanced x4 USB3 Integrated IOs FW, BIOS, • x24 PCIE 3.0, x8 PCIE2.0 x4 USB2 Drivers • x6 SATA3 Server Uncore Legacy IOs • x4 USB 3.0, x4 USB 2.0 BIOS • x2 10 GbE Intel® Ethernet

EDK2

x2 10GbE SPS Firmware DDR4/3L Intel® Other Features 128GB Max CS Drivers Ethernet • Storage/Network Environment LAN Drivers (Reliability, Temp, Availability) • Validated with external Crypto ECC UDIMM, OS/VMM SODIMM, Intel® Enabling/Testing accelerator (Coleto Creek) RDIMM Switches • Non-Transparent Bridging (NTB), Asynchronous DRAM self-refresh (ADR), Intel® QuickData Technology

Intel® HW Intel® SW 3rd Party HW

Intel Confidential Intel and the Intel logo are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. Other names and brands may be claimed as the property of others. All products, dates, and figures are preliminary and are subject to change without any notice. Copyright © 2015, Intel Corporation. Xeon® D-2200 (Bakerville/Skylake-D) Overview CPU Skylake-D Socket & TDP FCBGA 45 mm x 52.5 mm; Up to 110W Block Diagram Scalability 1S SoC Cores Up to 18C Up to 4 DDR4 channels* (up to 2DIMMs / channel) Ethernet QAT PCIe Gen3 Memory RDIMM/LRDIMM, 1-2 DPC @2133-2666MT/s (4x1, 4x10) 2x16 Up to 512GB memory capacity

Up to 4 ports 10 GbE/1GbE Integrated Ethernet w/ iWARP DDR4 DDR4 Ethernet/LAN 2133-2666 IntelIntel®® 2133-2666 RDMA (Supports KR/KX/SFI) ® 1MB L2 IntelXeon™Xeon™ CoreCore DDR4 Xeon™ Core

Crypto: Up to ~100 Gb/s IPSec/SSL, L2 1MB DDR4 Accelerators 2133-2666 Intel® 2133-2666 Compression: Up to ~100Gb/s, Xeon™ Core (QAT) L2 1MB Public Key: Up to RSA2K 100K Ops USB2.0 x4 PCIe x32 PCIe Gen3 + up to 20 PCIe Gen3 lanes via Flexible IO Innovation PCIe Gen 3 SATA3 USB3.0 Engine

20 High-Speed Flexible Lanes configured as PCIe or SATA or x20 x14 x4 USB3 Legacy Intel® 20 Flexible HSIO Lanes Flexible IO Up to 20 lanes of PCIe 3.0 (Bifurcation support: x4, x2) or IO ME

Up to 14 lanes of SATA Gen 3 x20 Up to four ports of USB Gen 3

Other IO Up to 4 USB 2.0, eSPI/LPC, SMBus, Integrated Clocking New capabilities relative to BDW-DE in Bolded Blue *Varies with SKUs Manageability Intel® Manageability Engine (Intel® ME), Innovation Engine (IE)

Datacenter Group LAUNCHEDIntel® Confidential – For Q1’18Use Under NDA !!! Only 43 Gen Over Gen Comparison Grangeville with Grangeville NS with Bakerville with Spec Broadwell-DE SoC Broadwell-DE-NS SoC Skylake-D SoC

SOC TDP 20-65W ~20-65W ~60-110W

Cores Up to 16C with Intel® HT Technology Up to 8C with Intel® HT Technology Up to 18C with Intel® HT Technology

Package Size 37.5mm x 37.5mm 37.5mm x 37.5mm 45mm x 52.5mm

Key Target Network/Storage/Compute Network Storage/Compute/Network

AVX AVX-256 AVX-256 AVX-512 New Instructions

LLC: 1.5MB/core LLC: 1.5MB/core LLC: 1.375MB/core Cache MLC: 256k/core MLC: 256k/core MLC: 1MB/core 2 channels DDR4 per SOC 2 channels DDR4 per SOC 4 channels DDR4 per SOC Memory RDIMM, LRDIMM, SODIMM, UDIMM RDIMM, LRDIMM, SODIMM, UDIMM RDIMM, LRDIMM, SODIMM* Up to 128G memory capacity Up to 128G memory capacity Up to 512GB memory capacity

Ethernet Up to 2 lanes of 10GbE/1GbE Up to 4 lanes of 10GbE/1GbE Up to 4 lanes of 10GbE/1GbE with RDMA

PCIe 3.0 (2.5, 5.0, 8.0 GT/s) PCIe 3.0 (2.5, 5.0, 8.0 GT/s) PCIe 3.0 (2.5, 5.0, 8.0 GT/s) PCIe 24xGen3 + 8xGen2 lanes 24xGen3 + 8xGen2 lanes 32xGen3 lanes + Up to 20xGen3 over High Speed IO

SATA 6xSATA ports 6xSATA ports Up to: 14xSATA

Crypto/ Via External Controller Integrated Intel® QuickAssist Technology: Up to 40G Integrated Intel® QuickAssist Technology: Compression Coleto Creek/Cave Creek or similar Crypto/20G Compression, 40kOps PKE 2K Up to 100G Crypto/Compression + 100kOps PKE 2K New microarchitecture, memory subsystem and integration in Skylake generation Datacenter Group Intel® Confidential – For Use Under NDA Only *Supported, not validated. 44 See Memory POR Section. Details of Intel® QuickAssist Functions Look Aside HW Accelerator for Security & Compression / Decompression Co-processing services

Symmetric Cryptographic Functions: PCIe 3.0 PCIe 3.0 DMI Gen3 x4 • Ciphers: x8 x16 • AES, 3DES, (A)RC, Kasumi, SNOW3G, ZUC • Hash functions: Integrated Accelerator & PCH • MD5, SHA-1/2/3 USB3 • Authentication: Intel® QuickAssist Technology SATA3 • HMAC, AES-XCBC-MAC • Kasumi, SNOW3G, ZUC Crypto GPIO • Cipher-Hash/Authenticate Combined Operation 4x10 GbE/ Compression SPI & eSPI • Key Derivation / Pseudo Random Functions 1 GbE x1 GbE with RDMA SMBus

Public Key Functions: Innovation 10 GbE/1 GbE Engine • RSA, DH, DSA Ports (IE) • ECDSA, ECDH PCIe • Intel® Key Protection Technology Gen3 Root Compression/Decompression: • Deflate Lossless LZ77 & Huffman coding

Datacenter Group Intel® Confidential – For Use Under NDA Only 45 Пример использования

46 Atom C - uServer история

Atom C2000 Atom C3000

47 Atom C2000

До 8 ядер До 2.4GHz

48 Примеры использования

Network appliance Cold storage

49 50 Intel ® Data center SoCRoadmap

2015 2016 2017+

Intel® Xeon® D-1500 Product Family Next Gen (Broadwell-DE) Xeon® SoCs

Intel® Atom® Processor C2000 Intel® Atom® Processor Product Family C3000 Product Family (Denverton / Denverton NS)

Leadership 64-bit datacenter SoC roadmap targeting lower power, high density design points

Intel Confidential – For Use Under NDA Only 51 Estimated DenvertonPerformance General Purpose Server Performance: SPECint*_rate_base2006

Intel Confidential – For Use Under NDA Only 52 Harrisonville Platform overview Atom C3000

Core Ingredients

SSD Recommendations Denverton/Denverton NS • Intel SSD Data Center Family for SATA • SATA 2.5in x 7mm • M.2 Boot (PCIe* and SATA)

Atom® ME QAT Cores (Up to 16C) IE Low Power Mem Integrated IO FW, BIOS, Drivers Ctrl. NIC ME

SPS FW legend (SDV) Innovation Engine MCTP Open Source SW

uEFI BIOS Intel HW 4X10/2.5/1GbE, Gen3 IO, I/O & LAN Drivers Intel SW/FW Legacy IO OS/VMM Enabling/Testing rd До 16 ядер DDR 4 3 party FW До 2.2GHz

Intel Confidential – For Use Under NDA Only 53 Xeon® Processor D-1500 vs. Atom® Processor C3000

Intel® Xeon® processor D-1500 Intel® Atom® processor C3000 (Broadwell-DE) (Denverton, Denverton-NS) Production Mar 2015, Nov. 2015* July-August 2017 Availability CPU core Broadwell Xeon core Goldmont Atom® core TDP 20-65W 8.5-32W Cores # Up to 16 cores Up to 16 cores Cache 1.5MB L3 cache per core ≤8C L2 is 2MB/Core; >8C is 2MB/Core Pair Performance** Up to 3.7x Atom® C2750 Up to 2.6x Atom® C2750 ECC SODIMM, UDIMM, RDIMM ECC SODIMM, UDIMM, RDIMM Memory Type 2CH DDR4, DDR3L 2CH DDR4 Max Mem. Cap. 128GB 128GB Integrated Ethernet 2x 10 GbE Intel® Ethernet 4x10 GbE Intel® Ethernet Flexible High Speed No Yes I/O x24 PCIE gen 3, x8 PCIE gen 2 20 lanes of configurable HSIO I/O x6 SATA3 16 lanes PCIe gen 3, x16 SATA3,x4 USB3 x4 USB 3.0, x4 USB 2.0 + fixed eMMC Quick Assist (QAT) Discrete - Coleto Creek*** Integrated QAT**** Innovation Engine No Yes

*Depend on SKUs ** Comparing top bin on-roadmap SKUs Based on SPECint*_rate_base2006 projection *** Coleto Creek: DH89xxCL includes Chipset 8900 Series SKUs where 8925 ≤ SKU ≤ 8955 **** QAT on select Denverton-NS SKUs only

Intel Confidential – For Use Under NDA Only 54 Ну и совсем немного про MEHLOW(coffee lake xeon)

Intel Confidential – For Use Under NDA Only 55 Branding update - Product Numbering Conventions

Broadwell-EX Intel® Xeon Processor Scalable Family (Skylake/Purley) Intel® Xeon® processor E7 - 8 8 # # a V4 Intel® Xeon® Platinum 8 1 # # a processor Intel® Xeon® processor E7 - 4 8 # # a V4 Intel® Xeon® Gold 6 1 # # a processor

Broadwell-EP New Intel® Xeon® Gold 5 1 # # a processor Intel® Xeon® processor E5 - 4 6 # # a V4 Intel® Xeon® Silver 4 1 # # a processor Intel® Xeon® processor E5 - 2 6 # # a V4 Intel® Xeon® Bronze 3 1 # # a processor

Level Generation = SKU # (E.g. 20) = Integrations/Optimizations a Broadwell-EP 1S workstation Skylake / Basin Falls (if applicable, E.g. F = Fabric) Intel® Xeon® processor E5 - 1 6 # # a V4 New Intel® Xeon® W- 2 1 # # processor

Broadwell-DE Skylake-DE Intel® Xeon® processor D- 1 5 # # Change Intel® Xeon® D- 2 2 # # processor

Greenlow (Kaby Lake CPU) Mehlow (Coffee Lake CPU) Intel® Xeon® processor E3 - 1 2 # # a V6 New Intel® Xeon® E- 2 1 # # a processor

# of Cores G = integrated graphics Knights Landing Intel® Xeon Phi™ 7 2 # # a processor

Datacenter Group Intel Confidential – For Use Under NDA Only Mehlow Platform with Intel® Xeon® E processor

2016 2017 2018 2019

Xeon E3 Skylake –E ≤4c Kaby lake -E ≤4c Coffee Lake-E: ≤6c Entry Greenlow: LGA SKL PCH-H, 2ch 4 DIMM Mehlow: LGA, CNL PCH, Server & DDR4/3 2ch 4 DIMM WS

Mehlow platform will result in up to 30-50% improvement on multi-threaded workloads*

Source: Intel. Configuration: Windows* 10, dual-channel DDR4-2400, 7200RPM HDD, 4 phase VR (~$1 BOM adder) Note: Values are based on estimates.

Datacenter Group Intel Confidential – For Use Under NDA Only 57 Intel® Xeon® E Delivering Value to Multiple Segments

Small Business Workstation Equip your SMB with a server, allowing Creativity from anywhere – for increased security and reliability Optimized HD graphics for reliable, with small incremental investment. stable performance with your professional workloads

Coffee Lake-S

Storage Internet of Things Energy efficient; Faster I/O Best Perf/Watt delivery for IoT bandwidth with graphics for Entry applications in the public sector, Enterprise. Delivering value to medical, and industrial software defined storage (SDS) and production SAN/NAS.

Datacenter Group Intel Confidential – For Use Under NDA Only Intel® Xeon® E processor SKU Stack

4 Core 6 Core

E-2186G 6C/12T 3.8/4.7 6+2 HT 95W

E-2176G 6C/12T E-2174G 4C/8T 3.7/4.7 6+2 HT 80W

3.8/4.7 4+2 HT 71W pGFX E-2146G 6C/12T No pGFX E-2144G 4C/8T 3.5/4.5 6+2 HT 80W 3.6/4.5 4+2 HT 71W E-2136 6C/12T E-2134 4C/8T 3.3/4.5 6+0 HT 80W 3.5/4.5 4+0 HT 71W E-2126G 6C/6T E-2124G 4C/4T 3.3/4.5 6+2 No HT 80W 3.4/4.5 4+2 HT 71W

E-2124 4C/4T 3.3/4.3 4+0 No HT 71W Please contact your local customer QNR rep for reliability E-2104G 4C/4T differences between Xeon® E, Xeon® D, and Xeon® SP. 3.2 4+2 No HT 65W

* GT0 and GT2 graphics can be used for pedestal workstation, server, and AMT-server. Note: iGFX not supported in intel® Server Platforms Services (server) SKU Decoder Brand String, #Cores/Threads, Frequency, Core+Graphics, Hyper threading, Test TDP Note: Subject to change without notice. Datacenter Group All SKUs can be Intelused Confidential in either – Forserver Use Under or workstation NDA Only platforms Coffee lake Core i3, Pentium, and Celeron CPU Support Desktop CPU Intel® Core i3-8350K Intel® Core i3-8300 Intel® Core i3-8100 Intel® Core i3-8300T Intel® Core i3-8100T Intel® Pentium G5600 Intel® Pentium G5500 All SKUs and features Intel® Pentium G5400 are PRELIMINARY and can change without Intel® Celeron G4920 notice Intel® Celeron G4900 Intel® Pentium G5500T Intel® Pentium G5400T

Note: Intel® Celeron G4900T Non-Xeon CPUs do not support Intel® vPro™ or Intel® AMT Please reference the client Dashboard for CFL-S sample/prod schedules, CDI# 570312 Intel® Core/Pentium/Celeron SKU stack – Desktop Platform 5Q Roadmap link

Datacenter Group Intel Confidential – For Use Under NDA Only 60 Intel® Xeon® E LGA1151 Package & Socket

Not drawn to scale. Do not use for design purposes. ▪ CFL-S will use the same LGA1151 package & socket as KBL-S. LGA1151 – Maximizes design reuse and socket cost structure. 37.5mm x 37.5mm ▪ No support for forward or backward compatibility with legacy platforms.

15.5mm

Datacenter Group Intel Confidential – For Use Under NDA Only Integrated Intel® Wireless-AC (Jefferson Peak) M.2 2230 Jefferson Peak Platform Pwr ▪ CFL delivers integrated Wi-Fi & Bluetooth* 5. Cannon Lake PCH CRF CNVio (Wi-Fi) Pulsar Module MAC – Wi-Fi/BT MAC integrated into CNL PCH. PCH Resources GPIOs (BT), Clock (Power, I/O, Key E Key Wi-Fi/BT Chassis) PCIe (Wi-Fi) PHY+RF Integrated – Separate companion PHY+RF (CRF) (CNVi) 802.11ac R2 + Bluetooth* 5 module with Platform Pwr USB (BT)

2x2 (160 MHz Channels) and 1x1 M.2 2230 Thunder Peak options. Platform Pwr Cannon Lake PCH Discrete CNVio (Wi-Fi) ▪ Standard M.2, hybrid Key E pinout allows Pulsar Module MAC PCH Resources GPIOs (BT), Clock support for integrated (CNVi) and discrete. (Power, I/O, Chassis) E Key Wi-Fi/BT – Routing CNVi alone drives I/O savings. PCIe (Wi-Fi) Discrete Platform Pwr USB (BT) ▪ Soldered-down option also available. Integrated Intel® Wireless-AC delivers ▪ Driver & BIOS auto-detect CNVi or discrete. a flexible, compelling and competitive ▪ . WiFi/BT solution.

Datacenter Group Intel Confidential – For Use Under NDA Only 2017 2018 Освещена только часть линейки продукции для встраиваемых решений https://solutionsdirectory.intel.com/

https://www.intel.com/content/www/us/en/partner/solutions-alliance/program-overview.html Embedded SSDProduct Family Dedicated to IoT! Expanded Capacity Range Flexible feature sets Data center As low as 64 GB Low Power Intel® Remote Secure Erase Embedded PLI Extended Temperature Multiple form factors Extended supply life cycle

M.2 80 & 110mm 3yr* Client + Form/Fit/Function Continuity 2.5”

*NSG product availability for Embedded products is three years from when the product(s) is first available on the Intel SSD 5Q roadmap to when the product(s) last ships.

Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice. 64 Intel Technology Innovations Fill the Memory and Storage Gap Performance and Capacity for Every Need

Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice. 6565 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice. 66 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice.67 67 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice.68 68 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice.69 69 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice.70 70 Internet of Things Group Intel Confidential – NDA Platform Roadmap. All dates and plans are subject to change without notice.71 71 КОНКУРЕНТЫ

72 Intel® Xeon® Scalable Momentum Industry Adoption Growing Cloud Service Providers Enterprise & Gov’t Comms Service Providers

Datacenter Marketing Group 73 System Pricing Comparison

Supermicro 6029U-TRT Supermicro 2023US-TR4 • 12x 16GB DDR4-2666 • 16x 16GB DDR4-2666 • Quad 1GbE • Quad 1GbE • 1 SATA HDD • 1 SATA HDD http://www.thinkmate.com/system/ultra- http://www.thinkmate.com/system/ultra- superserver-6029u-trt superserver-6029u-trt Processor List Price 2S Intel Xeon-SP 8160 2S AMD EPYC 7601 2 x $4702/processor 2 x $4200/processor System Price $16148 $16280

• Both systems configured with 1 DIMM per channel to get full memory bandwidth. Systems priced online at www.Thinkmate.com on 10/12/2017. 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.

Datacenter Marketing Group 74 Internal Measured Performance: Enterprise and Cloud Workloads

Relative Performance

2 Higher is better

1,75

1,68 1,64 1,8 1,64

1,6 1,45

1,37

1,31

1,29 1,27

1,4 1,27

1,22 1,22

1,18 1,18

1,15 1,05

1,2 1,05 1,00 1 0,8 Network Storage Network 0,6 bound bound bound 0,4 workload workload workload 0,2 0 2S AMD EPYC Server Side Java Node.js PHP Runtime DPDK L3 Fwding HammerDB NoSQL Online Memory Server 7601 Javascript Database Database Transaction Caching Virtualization (32C, 2.2GHz) Processing

2S AMD EPYC 7601 2S Intel Xeon Platinum 8160 2S Intel Xeon Platinum 8180 (32C, 2.2GHz) (24C, 2.1Ghz) (28C, 2.5Ghz)

• EPYC perf variability on DPDK L3 Forwarding, HammerDB, and Memory Caching – see slides 13, 15

Datacenter Marketing Group 75 Performance per Core

Relative Performance Higher is better This is based on public SPECint*_rate_base2006 Performance / Number of Cores 1,8

1,6 1,53 1,38 1,4

1,2

0,8

0,6

0,4

0,2

0 2S AMD EPYC 7601 2S Intel Xeon Platinum 8160 2S Intel Xeon Platinum 8180 (32C, 2.2GHz) (24C, 2.1GHz) (28C, 2.5GHz)

Datacenter Marketing Group 76 Internal Measured Performance: HPC Workloads - Single Node

Relative Performance Compute sensitive apps

3,5 Higher is better 3,16

3 2,69 2,5 Memory sensitive apps

1,65

1,63

1,59

1,52

1,47

1,46 1,35

1,5 1,23

1,10

1,00

0,88

0,86

0,85 0,81

1 0,75 0,73

0,5

0 2S AMD EPYC STREAM-OMP OpenFOAM MILC NAMD GROMACS LAMMPS Black-Scholes Monte Carlo Double Precision 7601 Triad v4.1 v7 v2.12 v2016.2 v17-Aug-17 Kernel Kernel Matrix (32C, 2.2GHz) Multiplication 2S AMD EPYC 7601 2S Intel Xeon Gold 6148 2S Intel Xeon Platinum 8160 DGEMM (32C, 2.2GHz) (20C, 2.4GHz) (24C, 2.1GHz)

• All measurements above are single node. Expect Intel performance to improve on cluster measurements - more typical of HPC deployments • EPYC tested with GCC, AOCC/LLVM and Intel compiler. The best AMD result was used for each workload • Intel® Xeon Phi™ processors are a good option for customers who have memory-bandwidth-bound workloads

Datacenter Marketing Group 77 Экосистема не готова – Open Source проекты не оптимизированы Нужно по крайней мере 3-5 лет для оптимизации (2020+) Microsoft НЕ МОЖЕТ эмулировать x86 код на ARM архитектуре

Datacenter Marketing Group 78 99% серверов с GPGPU – на Xeon Интел в процессе разработки продуктов для рынка Nervana и MobileEye продукты – ASIC или AP+ASIC. FPGA проекты

Datacenter Marketing Group 79