You are on page 1of 25

Vasily . Lizunov @ intel .

com


Smart
Agriculture Retail: Real-Time Wearables Autonomous Cars & Virtual/Augmented
Precision Medicine &
Pricing & Inventory (Industrial/Lifestyle) 5G Connectivity Reality
Genomic Analytics

2
,
,

18% CAGR from SDN/NFV spend up to AI is the fastest growing


2017-20201 $157B through 20202 datacenter workload3
1. Source: IDC Q416 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.
3

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

Intel Data Intel Parallel Intel Media


Management, Center Manager
Intel Cache
Accelration SW Software Studio XE Server Studio
Orchestration Intel Virtual Development Intel Intel Intel
& Storage Gateway Intel RSTe
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 AVX Intel TSX

Libraries Intel HD Graphics Acceleration Library (DAAL)
Instructions Instructions Intel TXT Intel SPDK Intel ISA-L Embree

Intel Intel QAT Intel DLIA Intel VCA


Processors Chipsets Ethernet Omni-Path Network Silicon
ASICs Photonics FPGAs SSDs
Fabric Accelerators

4
Intel Xeon Scalable Processors
:

:
, Compute, storage, Network

1.6x e5/e7 v4
6

Intel Xeon Scalable Processor
Fabric Networking Accelerators SSDs Complementary
Intel QuickAssist Intel Optane SSD Intel FPGA
Intel Omni-Path Intel Ethernet DC P4800X Intel Xeon Phi
Architecture Intel AVX-512 Intel Silicon Photonics
Intel SSD DC P4600
Intel VMD
INTEGRATED OPTIONS


(e.g. Caffe*, Intel DAAL, Intel MKL, DPDK, SNAP*, SPDK)


, , PCI-E,
Intel Advanced Vector Extensions 512 (Intel AVX-512) Intel Math Kernel Library (Intel MKL) Data Plane Development Kit (DPDK)
Intel Volume Management Device (Intel VMD) Storage Performance Development Kit (SPDK) Intel Resource Director Technology (Intel RDT)
Intel Data Analytics Acceleration Library (Intel DAAL)

6
* New features in bold


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

PCI-E
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


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): 77
:



2, 4, & 8+

Enhanced PCIe

Note: Not representative of top bin die

, ,
8
2016 2017 2018

Intel Xeon Processor E7 Brickland Platform Purley Platform


High-end Skylake Cascade
E7 v3 E7 v4
Lake
RAS
Intel Xeon PLATINUM
Intel Xeon GOLD
Intel Xeon Processor E5 Grantley-EP Platform
2 E5 v3 E5-4600 v4 (4S)
Intel Xeon SILVER

performance/watt/$ E5 v3 E5-2600 v4
Intel Xeon BRONZE

E5 e7 class
9
INTEL XEON SCALABLE PROCESSOR
The Foundation for Agile, Secure Workload-Optimized Hybrid Cloud

INTEL XEON PLATINUM 81xx PROCESSORS


Most cores (28C, 56T)
Highest frequency (up to 3.6 GHz, 4C)
Most Socket flexibility (2, 4, 8+)

28 CORES
Elite IO/memory (48 PCIe 3.0 lanes, 6 memory channels)

UP TO
GREAT
Most/fastest 3SCALABLE PERFORMANCE
UPI (Ultra Path Interconnects, 10.4 GT/s)
SCALABLE PERFORMANCE
BETTER
SCALABLE PERFORMANCE


Fastest MemoryMEMORY PERFORMANCE
(DDR4-2666
Highest IntelADVANCED
AVX-512
MHz) AT LOW POWER
RASwith 2 FMA per core STANDARD RAS ENTRY HARDWARE-ENHANCED SECURITY
STANDARD RAS

2, 4 & 8 SUPPORT 3 LINKS


Intel Turbo Boost and Intel Hyper-Threading Technology
UP SOCKET UPI
WITH New Advanced RAS
TO UP TO
WORKLOAD-OPTIMIZED
Node Controller Support to assist in scaled node management
MODERATE TASKS DEDICATED TASKS
DDR4
2666 1.5 TB TOPLINE
M WITH
H UP TO
Z
MEMORY
CHANNEL BANDWIDTH
RESULT: TASKS
Best workload-optimized performance for general purpose compute across virtualized
environments and hybrid cloud deploymentsINTEL TURBO BOOST TECHNOLOGY AND
PERFORMANCE FOR GENERAL-PURPOSE
HIGHEST
AFFORDABLE, ENTRY PERFORMANCE
ACCELERATOR COMPUTE, STORAGE AND NETWORKING
INTEL HYPER-THREADING
Best choice for mission-critical, traditional TECHNOLOGY analytics, and AI/deep
DC apps, real-time FOR LIGHT-RANGE WORKLOADS
THROUGHPUT learning inference FORfor
workloads; smart path MODERATE WORKLOADS
DL training
Best performance for the most demanding storage and networking workloads
MAINSTREAM RAS: Reliability, Availability, and Serviceability

10
INTEL XEON SCALABLE PROCESSOR
The Foundation for Agile, Secure Workload-Optimized Hybrid Cloud

INTEL XEON GOLD 61xx PROCESSORS


22C (44 threads w/ HT)
2 and 4S
Up to 3.4 GHz, (6C)
3 UPI links (10.4GT/s)
DDR4-2666 MHz

28 CORES 22 CORES
Intel AVX-512 with 2 FMA

UP TO UP TO Node Controller Support

BETTER
SCALABLE PERFORMANCE SCALABLE PERFORMANCE
ENTRY
Enhanced scalability and workload optimized performance for
AT LOW POWER
general purpose/virtualized compute & hybridHARDWARE-ENHANCED
cloud SECURITY
STANDARD RAS for demanding storageSTANDARD
Improved performance RAS
and networking

UP
TO 2, 4 & 8 SUPPORT
SOCKET
3 LINKS
UPI
WITH
UP TO 2 & 4 SUPPORT
SOCKET
INTEL XEON GOLD 51XX PROCESSORS
MODERATE TASKS DEDICATED TASKS
2666 1.5 TB UP TO 3 UPI LINKS
Up to 14C (28T)
DDR4 M WITH
H UP TO
TOPLINE MEMORY
2 UPI links
- Z CHANNEL BANDWIDTH DDR4-2400 MHz
INTEL
IntelTURBO BOOST TECHNOLOGY
AVX-512 AND
with 1 FMA

HIGHEST
AFFORDABLE, ENTRY PERFORMANCE
ACCELERATOR
THROUGHPUT ADVANCED ANDRELIABILITY, AVAILABILITY
SERVICEABILITY


INTEL HYPER-THREADING
Affordable
FOR MODERATE
Suitable
AdvancedTECHNOLOGY
for a WORKLOADS
RAS and 4-socket scalability
wide range of workloads FOR LIGHT-RANGE WORKLOADS

MAINSTREAM
11
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

28 CORES UP TO 22 CORES
Intel Turbo Boost and Intel HT for multi-threaded apps

UP TO Suitable for a moderate range of workloads


UP TO 12 CORES UP TO 8 CORES
UP
TO 2, 4 & 8 SUPPORT
SOCKET
3 LINKS
WITH
UP TO 2&4
UPI INTEL XEON BRONZE 31XX PROCESSORS SOCKET
SUPPORT Up to 8C
2 SOCKET
SUPPORT 2 SOCKET
SUPPORT

48 PCI3.0
Up to 1.7GHz (8C)
DDR4
- 2666 1.5 TB
M WITH
H UP TO
Z
TOPLINE MEMORY
CHANNEL BANDWIDTH 3
UP TOEntry 2-socket
UPI LINKS configuration
DDR4-2133 MHz
2 UPI at 9.6 GT/s
DDR4
2400 M
H
Z
LANES
E

HIGHEST
Intel AVX-512 w/ 1 FMA
ACCELERATOR
ADVANCED AND SERVICEABILITY
RELIABILITY, AVAILABILITY
Reliability upgrade versus Intel Xeon processor
THROUGHPUT
E3 platform
INTEL TURBO BOOST TECHNOLOGY
HYPER-THREADING TECHNOLOGY STANDARD ANDRELIABILITY, AVAILABILITY
SERVICEABILITY

MAINSTREAM EFFICIENT ENTRY


12
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)

13
28C Top Bin 205W 28C 165W Extended life (10-Year Use) +
Highly flexible offerings optimized to
SKUs OPTIMIZED FOR HIGHEST PER-CORE PERFORMANCE
2.5G 8180 2.1G 8176
NEBS-Friendly Thermal Spec

OPTIMIZED FOR BALANCED ENERGY EFFICIENT PERF/W


24C
2.7G
205W
8168
26C
2.1G
165W
8170
address broadest array of workloads
24C 150W
26C 150W 2.1G 8160T
12C-24.75M 150W
2.0G
3.0G 8158 8164
20C
2.0G
125W Four primary SKU lineups:
4C-16.50M 105W 24C 150W 6138T
3.6G 8156
2.1G 8160
16C 125W
1. Optimized for per core performance
16C 125W 2.1G
20C
2.4G
150W
2.0G (8S entry) 8153
6130T 2. Balanced, energy efficient perf/W
6148 12C-19.25M 125W
18C 200W 22C 140W 2.6G 6126T 3. Extended life
3.0G 2.1G 6152
6154
20C 125W
14C
2.2G
105W 4. Integrated Intel Omni-Path
18C 165W 5120T
2.0G
2.7G 6150 6138
14C 85W
16C 150W
18C
2.3G
140W 1.9G 5119T New 205W SKUs:
2.6G 6140
6142
16C 125W
12C
2.1G
85W e.g. 8180 hits 2.5 GHz at 28 cores
14C 140W 2.1G 4116T
2.6G 6130
6132
14C 105W 10C 85W
12C-24.75M 165W 2.2G 5120
2.2G 4114T Great high frequency options:
3.2G
6146 8C 70W
12C-24.75M 150W
12C
2.3G
105W
2.0G 4109T
e.g. 6144 w/ 8 cores at 3.5 GHz base
5118
3.0G 6136
10C 85W SKUs w/ addl 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
6C-19.25M 115W 2.6G 4112 28C 173W
3.4G 6128 8C 85W 2.1G 8176F
2.1G 4110
4C-16.50M 105W 20C 160W 24C 160W
3.6G 5122 8C 85W 2.4G 2.1G
6148F 8160F
1.8G 4108
16C 160W 20C 135W
Core Count TDP 8C 85W 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 6130F
1.7G 6126F
M SKU available (1.5 TB/Sck)
TB/Skt) SKU # 3104

2114

STORAGE
Intel Optane DC ssds

intel 3d nand DC SSds Up to 60X better at 99% QoS1
PCIe Up to 40X faster response time1

Intel Storage
Intel Intelligent Storage Acceleration Library (ISA-L)
Intel Storage Performance Development Kit (SPDK)

Intel Platform Storage 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

15
5g-ready
26B IP networked devices/ Intel Ethernet 700 Series
Connections One Architecture. Multiple Speeds
by 2020
int. Intel Ethernet Connection X722 w/
iWARP RDMA (4x10GbE) up to 2.55x higher
Advanced features IPSec (Gbps)
Optimized SW (DPDK) for NFV

Int. Intel QuickAssist Technology


Crypto & Data Compression

2.3 ZB Global Intel Omni-Path Architecture (Fabric)
High-bandwidth & low-latency fabric
up to 2.49x higher
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.
16
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


17
SMBus Integrated IO
NC-SI MCTP
Feature Support Customer Value
Quad Port 10GbE MAC/PHY Optimized for networking capability in
Based on Intels 10GbE solution Cloud, Comms, and Storage
Interfaces: 10G (KR, SFI, XFI), 1G (KX) Single network driver on Intel platform
VF00
VF VF00
VF VF
VF
00 VF00
VF
VF 0 VF 1 VF m
Remote Direct Memory Access Routable and scalable RDMA ideal for
In-Band PF0 PF1 PFn
iWARP large segmented networks in private
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

18
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
Kernel
Avoiding application context switching
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 Network
Recommended for easy deployment and configuration, scalability and TCP/IP 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
19
20

Up to Up to
4.28x 5X

virtualization
SpecVirt
Database
OLTP Hammer DB
1.6

e5/7 v4
Up to Up to
2.7x 8.2x
Networking Supercomputing
DPDK L3 FWD LINPACK
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 www.intel.com/benchmarks. Configuration: Refer to Performance Benchmark Disclosure slide. Results have been estimated or simulated
using internal Intel analysis or architecture simulation or modeling, and provided to you for informational purposes. Any differences in your system hardware, software or configuration may affect your actual performance. *Other names and brands
may be claimed as the property of others.

21
,

,
,

,
(compute / storage / network)

(Skylake / Cascade Lake / Ice


Lake = LGA 3647)

22
Notices & Disclaimers
Intel technologies features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance
varies depending on system configuration. No computer system can be absolutely secure. Check with your system manufacturer or retailer or learn more at
intel.com.

Tests document performance of components on a particular test, in specific systems. Differences in hardware, software, or configuration will affect actual
performance. Consult other sources of information to evaluate performance as you consider your purchase. For more complete information about
performance and benchmark results, visit http://www.intel.com/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. For more complete information visit
http://www.intel.com/performance.

Cost reduction scenarios described are intended as examples of how a given Intel-based product, in the specified circumstances and configurations, may
affect future costs and provide cost savings. Circumstances will vary. Intel does not guarantee any costs or cost reduction.

No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.

Intel does not control or audit third-party benchmark data or the web sites referenced in this document. You should visit the referenced web site and
confirm whether referenced data are accurate.

2017 Intel Corporation. Intel, the Intel logo, and Intel Xeon are trademarks of Intel Corporation in the U.S. and/or other countries. *Other names and
brands may be claimed as property of others.

24