SSG EMEA Software enabling
Intel Multi-Core
the Road Intel is committed to
Bjoern Bruecher
Sr. SW Application Engineer
SSG/DRD/EMEA Enterprise Team
Intel GmbH, Munich
Bjoern.Bruecher@Intel.com
Intel: Total Platform Approach
PLATFORM
VALIDATION
CHIPSETS/
COMMS
BOARDS
SYSTEMS
INTEL CAPITAL
INDUSTRY
STANDARDS
SOLUTIONS
BLUEPRINTS
Unparalleled industry
reach and ecosystem
breadth
SOFTWARE
TOOLS
INTEL®
SOLUTIONS
SERVICES
DEVELOPER
SERVICES
SOFTWARE VENDOR
ALLIANCES
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
2
The Intel Behind Intel Software
Intel Software and Solutions Group
1 group • 6 divisions • 14 time zones • 24 major sites • 2,500 employees
170,000 registered developers, 20+ Operating Systems, Thousands of Applications
Software Development Tools
Software IP Strategy
Managed Run Time Enabling
Architecting Solutions & Services
Strategic Enabling
Parallel & Distributed Solutions
Software Enabling & Developer Services
Platform Planning
Intel Software Tools, Engineers, Expertise
SSG EMEA Software Enabling
3
for CST GmbH; Bjoern.Bruecher@Intel.com
More’s Law is Alive and Well
Transistor Count
10,000,000,000
Itanium® Family DC Processor (Montecito) = 1.7 Bil
1,000,000,000
Pentium® M Processor
Pentium® 4 Processor
Pentium® III Processor
Pentium® II Processor
10,000,000
Pentium® Processor
486™ DX Processor
1,000,000
386™ Processor
286
100,000
8086
8008
10,000
8080
4004
1970
100,000,000
1980
1990
2000
1,000
2010
“Eventually one billion transistors may crowd a single
chip – 1,000 times more than possible today.”
National Geographic, 1982
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
4
…thru Intel’s Nanotechnology Leadership
1000 nm
100 nm
10 nm
1980
1990
2000
SSG EMEA Software Enabling
2010
2020
5
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
Intel Keeps Introducing New
Si Generation Every 2 Years
www.intel.com/research/silicon
All features and dates specified are targets provided for planning purposes only and are subject to change without notice
45nm Chips currently in Production
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
6
Challenge 1: Power
Power Density Race
Sun’
Sun’s Surface
10,000
Rocket Nozzle
Power Density
(W/cm2)
1,000
Nuclear Reactor
100
Hot Plate
8086
10 4004
8085
386
8008
286
486
8080
1
’70
’80
’90
Pentium®
processors
’00
’10
Exponential of Power will not allow us to push Frequency
SSG EMEA Software Enabling
7
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
Challenge 2: Memory Latency
Memory Latency
3000
Logic
2500
GAP
MHz
2000
1500
1000
500
Memory
0
1992
1994
1996
1998
2000
2002
CPU needs to wait longer and longer accessing Memory
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
8
Challenge 3: RC Delays
D e la y (p s )
10000
Interconnect RC Delay
Clock Period
1000
100
RC delay of 1mm interconnect
10
Copper Interconnect
1
0.35
0.25
180nm
130nm
90nm
65nm
Process Technology
i486™: 16 clocks RAM access roundtrip
65 nm IC: 15 clocks to cross the chip
Signal Propagation Delay increases
SSG EMEA Software Enabling
9
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
Doubling Performance on Tight Power Budget
2f
P ~ 4f² 4P0
4x Power
Increase
f
f
P = 4P0
P0 ~ f²
Core
Die/Socket
f
f
‘only’ 2x
P ~ 2f² 2P0
Double the Cores to Double Performance
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
* Other brands and names may be claimed as the property of others.
10
Multi-Core Paradigm
Old Paradigm:
New Paradigm:
Do One Task Really, Really Fast
Do Many Tasks Simultaneously
Single Core Performance
System Performance
Relative Performance
40
1
0.8
30
Assuming 100%
parallel software
1
0.6
20
0.4
10
0.5
0.2
0.3
0
0
Large
Medium
Small
Large
Medium
Small
Frequency is NO Longer the Best indicator of Performance
SSG EMEA Software Enabling
11
for CST GmbH; Bjoern.Bruecher@Intel.com
Innovative Microarchitecture
Quad-Core Intel® Xeon® Processor 5300 Series feature …
Wide Dynamic
Execution
Intel®
Intel® NetBurst®
NetBurst® Microarchitecture
+ New
Innovations
Intel® Core™
Microarchitecture
Advanced Digital
Media Boost
Smart Memory
Access
Advanced Smart
Cache
Intelligent Power
Capability
Mobile Microarchitecture
Also used in Dual-Core Intel® Xeon® Processor 5100 Series
Common Architecture for Intel Servers, Desktop, Mobile
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
12
New Intel® Core™ Microarchitecture
FASTER
EFFICIENT
Intel®
Intel® Intelligent
Power Capability
Intel® Wide Dynamic
Execution Engine
4 instructions
per cycle
SMARTER
How do these features work?
Learn more at intel.com
http://www.intel.com/technology/architecture
/coremicro/demo/demo.htm
1
ultra fine grained
power control
Intel® Advanced Smart Cache
Intel® Smart Memory Access
2x cache size, shared
Advanced pre-fetching
compared to previous generation Dual-Core Intel® Xeon® Processor based servers
Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate performance of Intel
products as measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance.
Delivering a Performance and Power Advantage for IT
SSG EMEA Software Enabling
13
for CST GmbH; Bjoern.Bruecher@Intel.com
New Intel® Core™ Microarchitecture
Intel®
Intel® Wide
Wide Dynamic
Dynamic Execution
Execution (WIDER)
(WIDER)
NetBurst/Yonah Pipeline (Three)
Conroe/Woodcrest/Merom Pipeline (Four Lane Super Scalar Pipeline)
Conroe/Woodcrest/Merom Decodes, Executes & Retires instructions
at a sustained rate of 4 instructions per clock cycle
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
14
New Intel® Core™ Microarchitecture
Smart Cache
Intel® Advanced Smart Cache
Intel®
Intel® Advanced
Advanced Smart
Smart Cache
Cache (Efficient
(Efficient Data
Data Sharing)
Sharing)
Dynamic Cache Allocation:
Shared Cache adapts to
mismatched loads on
each core
Core
1
Core
2
L2 Cache
Dynamic Bandwidth Allocation:
High bandwidth application can
borrow L1 to L2 bandwidth from
the other core’s application.
Core
1
Core
2
L2 Cache
Core
1
2X L2 to L1 Bandwidth
Core
2
L2 Cache
2X Bandwidth to L1 caches & up to 4MB L2 cache
SSG EMEA Software Enabling
15
for CST GmbH; Bjoern.Bruecher@Intel.com
New Intel® Core™ Microarchitecture
® Advanced
Intel
Intel®
Advanced Digital
Digital Media
Media Boost
Boost (Improved
(Improved perf.
perf. for
for Multimedia
Multimedia apps)
apps)
Only
Per clock cycle
Improved Performance for Streaming SIMD Extensions (SSE, SSE2, SSE3)
(128 Bits every single clock cycle vs. 64 in previous Intel processors)
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
16
New Intel® Core™ Microarchitecture
Intel®
Intel® Intelligent
Intelligent Power
Power Capability;
Capability; Power
Power Gating
Gating
Power shut off
Intel Intelligent Power Capability optimizes energy usage,
delivering more performance per watt
SSG EMEA Software Enabling
17
for CST GmbH; Bjoern.Bruecher@Intel.com
Infrastructure Refresh with Multi-Core
Quad-Core Boosts Application Performance Across the Board
up to
5x
4.5x
Fastest Single Core Intel Xeon servers
Data as of Nov 7, 2006
Intel measured or published
Integer
Intel® Core™
Microarchitecture
Large 8MB
SAP 2 tier
4x
Java apps
Dual &
Floating pt
3x
On-Die Cache
database
Quad-Core
2x
Dedicated High
1.0
Advanced
Speed Buses
FBD Memory
64-bit Intel
Xeon single
core
Best Published Dual-Core Intel® Xeon® 5160
Intel measured Quad-Core Intel® Xeon® E5345
Intel® QuickData
Technology
Tremendous Capability – Broad IT Usages
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
18
Quad-Core for Workstations
Simultaneous
design and analysis
Interact seamlessly with large
models & assemblies
Improved Performance
for Rendering
Faster Analytics for
Financial Services
What Can Quad-Core Do For You?
SSG EMEA Software Enabling
19
for CST GmbH; Bjoern.Bruecher@Intel.com
The Magic of 45 nm
PERFORMANCE
Spectrum of Intel Architecture
ns
Optio
n
o
i
t
ra
ic s
Integ
Graph
:
e
l
p
Exam
SYSTEM
ON A CHIP
Notebook, Desktop and Server
Ultra-low Cost and
Power Optimized
Architecture
Intel® Core™ 2 and Nehalem
Architecture
Visual Computing
and HPC Optimized
Architecture
POWER
SSG EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
20
Conclusion
• Frequency is NO longer the Best indicator of
Performance
• Intel’s innovative Core Microarchitecture leads to
outstanding Performance on tight Power Budget
• Intel’s Dual-Core and Quad-Core boosts Application
Performance Across the Board
• Intel is committed to the Roadmap with a clear
Multi- / Many- Core Direction
SSG EMEA Software Enabling
21
for CST GmbH; Bjoern.Bruecher@Intel.com
Intel® Workstation Platform Roadmap
2H’
2H’2006
Xeon® DP
Workstation
2007
2008
Glidewell Platform
Harpertown
Quad-Core Xeon® processor 5300 series
Gainestown
Wolfdale-DP
Seaburg Chipset
Intel® 5000X Chipset
Tylersburg Chipset
Garlow Platform
Wyloway Platform
Foxhollow Platform
Intel® Core™ 2 Extreme
Intel® Core™ 2 Quad
Yorkfield
Bloomfield
Intel® Core™ 2 Extreme Intel® Core™ 2 Duo
Wolfdale
Bloomfield-DC
Intel® 975X Chipset
Intel® Mobile
Workstation
Thurley Platform
Stoakley Platform
Intel® Xeon® processor 5100 series
Intel® UP
Workstation
Future
Napa Platform
Santa Rosa Platform
Intel® Core™ 2 Duo
Intel® Core™ 2 Duo
Mob. Intel® 945PM Chipset
Mob. Intel® PM965
Chipset
22
SSG
EMEA Software Enabling
for CST GmbH; Bjoern.Bruecher@Intel.com
Tylersburg Chipset
Bearlake-X Chipset
Dual
Dual Core
Core
22
4+
4+ Cores
Cores