POWER7 Processors: The Beat Goes On: Joel M. Tendler, Executive IT Architect

POWER7 Processors: The Beat Goes On
Joel M. Tendler, Executive IT Architect

jtendler@us.ibm.com
Acknowledgment: This material is based upon work
supported by the Defense Advanced Research Projects
Agency under its Agreement No. HR0011-07-9-0002
IBM Power Systems value proposition
Deliver business value by leveraging technology
Performance Flexibility Reliability Affordability
+
–
. . . the highest value at the lowest risk

with leading technology
2
45nm
Approaching 20 Years of POWER Processors

65nm
Next Gen.
POWER7
RS64IV Sstar 130nm -Multi-core
POWER6TM
RS64III Pulsar 180nm -Ultra High Frequency
.18um
RS64II North Star
.25um POWER5TM
-SMT
RS64I Apache .35um
BiCMOS .5um
POWER4TM Major POWER® Innovation
.5um -Dual Core
Muskie A35 .22um -1990 RISC Architecture
.5um -1994 SMP
-Cobra A10 -1995 Out of Order Execution
-64 bit
-1996 64 Bit Enterprise Architecture
POWER3TM -1997 Hardware Multi-Threading
.35um -630 -2001 Dual Core Processors
-2001 Large System Scaling
-2001 Shared Caches
.72um -2003 On Chip Memory Control
POWER2TM
P2SC -2003 SMT
.25um -2006 Ultra High Frequency
RSC .35um -2006 Dual Scope Coherence Mgmt
1.0um -2006 Decimal Float/VSX
.6um
-2006 Processor Recovery/Sparing
604e
-2009 Balanced Multi-core Processor
-603
POWER1 -2009 On Chip EDRAM
-AMERICA’s
-601
1990 1995 2000 2005 2010

* Dates represent approximate processor power-on dates, not system availability
3
POWER Roadmap – The Only Reliable Server Roadmap

2001 2004 2007 2010
POWER4 POWER5 POWER6 POWER7*
5GHz Alti Advanced

2 Cores -
Vec Core Design
2.3GHz 2.3GHz
1.9 1.5+
Core Core L2 Cache Cache
1.5+ GHz 1.5+ GHz GHz GHz Advanced Advanced
1+ GHz Core1+ GHzCore Core Core
Shared L2 System Features System Features
Core Core Distributed
Shared L2 Switch 4-8 cores / die
Shared L2 Highly threaded cores
Distributed Distributed Switch

Shared L2 Switch Very High Frequencies 4-5GHz
Enhanced Virtualization
2.3 GHz POWER5+ Advanced Memory Subsystem
Distributed Switch Enhanced Scaling Altivec Vector SIMD instructions
Simultaneous Multi-Threading (SMT) Instruction Retry/Alternate
Enhanced Distributed Switch Processor Recovery
Enhanced Core Parallelism Decimal Floating Point
Chip Multi Processing
Improved FP Performance Dynamic Energy Management
- Distributed Switch
Increased memory bandwidth Partition Mobility
- Shared L2
Micropartitions Memory Protection Keys
Dynamic LPARs (32)
Virtualized IO Advanced Memory Sharing
Upgrades to be available
First Dual core chip in industry First Quad core in industry Fastest chip in industry
For Power 570 & Power 595
BINARY COMPATIBILITY
*All statements regarding IBM's future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.
4
POWER7 Processor Chip

¾ 567mm2 Technology: 45nm lithography, Cu, SOI, eDRAM
¾ 1.2B transistors
Equivalent function of 2.7B
eDRAM efficiency
¾ Eight processor cores
12 execution units per core
4 Way SMT per core
32 Threads per chip
256KB L2 per core
¾ 32MB on chip eDRAM shared L3
¾ Dual DDR3 Memory Controllers
100GB/s Memory bandwidth per chip sustained
¾ Scalability up to 32 Sockets
360GB/s SMP bandwidth/chip
20,000 coherent operations in flight
¾ Advanced pre-fetching Data and
Instruction
* Statements regarding SMP servers
¾ Binary Compatibility with POWER6 and prior systems do not imply that IBM will introduce
a system with this capability.
5
Computer Systems Memory: DRAM
¾ Dynamic RAM
Uses MOSFETs and Capacitors
Charge , i.e., “0” or “1”, decays over time
(capacitor discharges) losing its state
To be useful, cell needs periodic refreshing
(dynamic)
Volatile memory (data lost when memory is
not powered)
¾ Each cell stores 1 bit
Memory cell: 1xMOSFET + 1xCapacitor
Diagram courtesy of Wikipedia

¾ Comparison with SRAM
Less components, much more dense
Higher latency, variable access timings
4x4 Dynamic RAM memory
Charge refresh & amplification complexity
7
Computer Systems Memory: SRAM
¾ Static RAM
Uses transistors only, e.g., MOSFET
Diagram courtesy of Wikipedia

Charge does not need to be refreshed (static)
Uses bi-stable latching circuitry
Volatile memory (data lost when memory is
not powered)
Static RAM memory cell: 6 x MOSFET
¾ Each cell stores 1 bit
6 or more MOSFETs
¾ Comparison with DRAM
Uses more components per bit
Lower latency
Higher frequency access
Simpler access interface
8
POWER7 Design Principles: Traditional Performance View
10000
Multiple optimization Points

1000
¾ Balanced Design
100
Multiple optimization points
Improved energy efficiency
10
RAS improvements
¾ Improved Thread Performance
1
Dynamic allocation of resources Thread Core Socket 32 Chip
System
Shared L3
¾ Increased Core parallelism Balanced View
4 Way SMT
10000
Aggressive out of order execution 32 Chip
System POWER6
¾ Extreme Increase in Socket
POWER7
Throughput 1000
System Thruput
Continued growth in socket

bandwidth 100
Socket
Balanced core, cache, memory

improvements Core
¾ System 10
Thread
Scalable interconnect
Reduced coherence traffic 1
* Statements regarding SMP servers do not imply that IBM will Single Thread Performance
introduce a system with this capability.
10
Graphs for illustration purposes only (Not actual data)
POWER7 Design Principles:

Flexibility and Adaptability
¾ Cores:
4, 6, and 8-core offerings with up to 32MB of L3 Cache
Dynamically turn cores on and off, reallocating energy
Dynamically vary individual core frequencies, reallocating energy
Dynamically enable and disable up to 4 threads per core
¾ Memory Subsystem:
4 or 8 channel configurations
¾ System Topologies:
Standard, half-width, and double-width SMP busses supported
¾ Multiple System Packages
2/4s Blades and Racks High-End and Mid-Range Compute Intensive
Single Chip Organic Single Chip Glass Ceramic Quad-chip MCM
1 Memory Controller 2 Memory Controllers 8 Memory Controllers

3 4B local links 3 8B local links 3 16B local links (on MCM)
2 8B Remote links
11
* Statements regarding SMP servers do not imply that IBM will introduce a system with this capability.
POWER7: Core
¾ Execution Units DFU
2 Fixed point units
ISU
2 Load store units VSX
4 Double precision floating FPU
FXU
point
1 Vector unit
1 Branch
1 Condition register IFU
1 Decimal floating point unit CRU/BRU
6 Wide dispatch/8 Wide Issue
LSU
¾ Recovery Function Distributed
¾ 1,2,4 Way SMT Support
¾ Out of Order Execution
256KB L2
¾ 32KB I-Cache
¾ 32KB D-Cache
¾ 256KB L2
Tightly coupled to core
12
Computer Systems Memory: Hierarchy

Instructions
Microprocessor
Integrated Circuit OutputData
InputData
Typical Memory Hierarchy

¾ Memory wall
Registers
Differential in “performance” between
Processor intimacy, smaller capacity
Increased Latency, greater capacity

{
addressable memory and the Instruction cache
microprocessor L1 cache
¾ Implement a memory hierarchy and Data cache
intelligence to reduce access
L2 cache
times/latency
¾ Combination of memory-cell L3 cache
technology and “distance” from the
processor DIMM Buffer ASIC
Addressable memory
13
Challenge: Beating Physics to Realize Multi-core Potential
Local SMP Links

Core Core Core Core
L2 Cache L2 Cache L2 Cache L2 Cache
Mem Ctrl L3 Cache and Chip Interconnect Mem Ctrl
Remote SMP + I/O Links

Core Core Core Core
POWER7™ is an 8-core, high performance Server chip. A solid chip is a good start.
But to win the race, you need a balanced system. POWER7 enables that balance.
14
Challenge: Beating Physics to Realize Multi-core Potential

Need to Amplify Effective Compute Throughput Potential
Socket Throughput
to Close Gap and
Achieve Potential
Socket Throughput Limitation

(Physical signal economics)
Multi-core evolution
15
Trends in Server Evolution

Single Image Virtualized/Cloud
Emerging Entry Server - A simple matter of riding
Virtualized/Cloud Platform the multi-core trend?
- Add more cores to the die,
Enabled by: 8-core 8-core
beef up some interfaces,
- Technology and scale to a large SMP?
- Innovation
Driven by: 8-core 8-core

- IT Evolution
Time
- Economics 2 to 4 socket
16 to 32-way SMP Server
Traditional Entry Server Traditional High-End Server

Single Image Platform Virtualized Consolidation Platform
2-core 2-core
2-core 2-core
2 to 4 socket 8 to 32 socket
4 to 8-way SMP Server 16 to 64-way SMP Server do not imply that IBM will introduce
16
Single Image Virtualized/Cloud
Emerging Entry Server - A simple matter of riding
Virtualized/Cloud Platform the multi-core trend?
- Add more cores to the die,
beef up some interfaces,
- Technology and scale to a large SMP?
- Innovation
Not so simple:
Driven by: 8-core 8-core - Emerging entry servers
Sim
- IT Evolution have characteristics similar
Time
- Economics 2 to 4 socket to traditional high-end
i
lar
16 to 32-way SMP Server
large SMP servers
Ch
all
Traditional Entry Server Traditional High-End Server
en
Single Image Platform Virtualized Consolidation Platform
ge
Achieving solid virtual
machine performance
2-core 2-core
requires a Balanced
System Structure.
2-core 2-core
17
Single Image Virtualized/Cloud UltraScale Cloud
Emerging Entry Server Emerging High-End Server
Virtualized/Cloud Platform UltraScale Cloud Platform

- Technology
- Innovation
Driven by: 8-core 8-core

- IT Evolution
Time
- Economics 2 to 4 socket 8 to 32 socket

16 to 32-way SMP Server 64 to 256-way SMP Server
Traditional Entry Server Traditional High-End Server Same enablers and

Single Image Platform Virtualized Consolidation Platform driving factors apply
at larger scale
2-core 2-core
2-core 2-core
18
Challenge: How does POWER7 maintain the Balance?

Socket Throughput
to Close Gap and
Achieve Potential
Cache Hierarchy Technology

and Innovation

19
Cache Hierarchy Technology and Innovation
Cache Hierarchy Rqmt Challenge

for POWER® Servers for Multi-core POWER7
Core ... Core POWER4TM, POWER5TM, and

POWER6TM systems derive huge
benefit from high bandwidth access
Low Low
Latency Latency to large, off-chip cache.
2M to 4M 2M to 4M
per Core per Core But socket pin count constraints
Cache Cache
footprint footprint
prevent scaling the off-chip cache
interface to support 8 cores.
Large, Shared, 30+ MB

Cache footprint
much closer than
Local Memory
20
Solution: High speed eDRAM on the processor die
Conventional IBM ASIC IBM Custom Custom Custom

Memory DRAM eDRAM eDRAM Dense SRAM Fast SRAM
Dense, low power Off On High Area/power

Low speed/bandwidth uP uP High speed/bandwidth
Chip Chip
Conventional Large, Off-chip On-processor On-processor Private core

Memory DIMMs 30+ MB Cache 30+ MB Cache Multi-MB Cache Sub-MB Cache
Industry Standard Caching and Memory Technologies:

Conventional DIMMs, Dense and Fast SRAM’s.
IBM’s POWER Servers have leveraged large off-chip

eDRAM caches in POWER4, 5, and 6.
With POWER7, IBM introduces on-processor, high-speed,

custom eDRAM, combining the dense, low power attributes
of eDRAM with the speed and bandwidth of SRAM.
21
Local SMP Links

Core Core Core Core

Core Core Core Core
22

for POWER Servers for Multi-core POWER7
Core ... Core

Need to satisfy both caching
requirements with one cache.
Low Low
Latency Latency
2M to 4M 2M to 4M
per Core per Core
Cache Cache
footprint footprint

Cache footprint
much closer than
Local Memory
23
Solution: Hybrid L3 “Fluid” Cache Structure
Core Core Core Core Core Core Core Core
Private Shared
Private Shared Private
Private
Large, Shared Private
Private 32M L3 Cache Private
Shared Private
Private
- Keeps multiple footprints at ~3X lower latency than local memory.
Working Set
Footprints
24
Cloned Private Shared

Private Cloned
Shared Private
Private
Large, Shared Private
Shared Fast, Local Fast, Local Private
L3 Region Private
L3 Region
- Keeps multiple footprints at ~3X lower latency than local memory.

- Automatically migrates private footprints (up to 4M) to fast local
region (per core) at ~5X lower latency than full L3 cache.
Working Set
- Automatically clones shared data to multiple private regions.
Footprints
25
Private Private Private

Private Private
Private Private
Private Private
Private Private Private Private
Large, Shared
Private Private Private
32M L3 Cache Private
Private Fast, Local Private
Private Private Private Private
L3 Region
- Enables a subset of the cores to utilize the entire large shared

L3 cache when the remaining cores are not using it.
26
Local SMP Links

Core Core Core Core

Fast Local
L3 Region

Core Core Core Core
27

for POWER Servers for Multi-core POWER7
Core ... Core Low power, dense eDRAM

value enhanced with
low latency, high bandwidth,
Low Low
Latency Latency
fast SRAM structures
2M to 4M 2M to 4M
per Core per Core
Cache Cache
footprint footprint IBM Custom Custom
eDRAM Fast SRAM
Dense, low power High Area/power

Lower speed/bandwidth High speed/bandwidth

Cache footprint
much closer than
Local Memory
On-processor Private core
30+ MB Cache Sub-MB Cache
28
Solution: L2 “Turbo” Cache
L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo

Cache Cache Cache Cache Cache Cache Cache Cache
Private Private
Cloned Private Cloned
Private
Private
Cloned Cloned
Large, Shared
L3 Region Private
L3 Region
- L2 “Turbo” cache keeps a tight 256K working set with extremely

low latency (~3X lower than local L3 region) and high bandwidth,
reducing L3 power and boosting performance.
29
Local SMP Links

Core Core Core Core

Core Core Core Core
30

Cache Hierarchy Summary
L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo L2 Turbo

Cache Cache Cache Cache Cache Cache Cache Cache
Private Private
Cloned Private Cloned
Private
Private
Cloned Cloned
Large, Shared
L3 Region Private
L3 Region
Cache Level Capacity Array Policy Comment

L1 Data 32K Fast SRAM Store-thru Local thread storage update
Private L2 256K Fast SRAM Store-In De-coupled global storage update
Fast L3 Region Up to 4M eDRAM Partial Victim Reduced power footprint (up to 4M)
Shared L3 32M eDRAM Adaptive Large 32M shared footprint
31

Socket Throughput
to Close Gap and
Achieve Potential
Advances in Memory Subsystem

and Innovation

32
Advances in memory subsystem

Memory subsystem requirement for POWER7 processor-based servers
POWER7 Requirements
Core
10-20GB/s ¾ Core:
sustained
bandwidth 10GB/s to 20GB/s sustained
per core
memory bandwidth per core
16GB to 32GB of cache
¾ Socket:
4 times growth in memory
bandwidth & capacity
¾ System:
16-32GB
Energy storage per Packaging more memory into
constraints core
similar volume, with similar energy
and cooling constraints
33

Multi-faceted Solution
POWER7 Chip 1) Dual Integrated DDR3 Controllers
- Massive 16KB scheduling window
per POWER7 chip insures high
channel and DIMM utilization
- Sparse access acceleration
Memory Memory - Advanced Energy Management
Controller Controller - Numerous RAS advances
2) Eight high speed 6.4 GHz channels

- New low power differential signaling
- Sustained 100+ GB/s per socket
Advanced
Buffer 3) New DDR3 buffer chip architecture
Chip - Larger capacity support (32 GB / core)
- Energy Management support
- RAS enablement
4) DDR3 DRAMs
- Supports 800, 1066, 1333, and 1600
34
* Statements regarding memory subsystem features do not imply that IBM will introduce a system with these capabilities.
Local SMP Links

Core Core Core Core

Core Core Core Core
35

Socket Throughput
to Close Gap and
Achieve Potential
Advances in Off-Chip Signaling

Technology

and Innovation

36
Advances in Off-chip Signaling Technology
1) Enhanced Signal-ended “Elastic Interface” Technology

2) New high speed, low power Differential Technology
Interface Signal Type Info Width Frequency Bandwidth
Off-chip Cache none none none none
Memory Channels Differential 28 bytes 6.4 Ghz 180 GB/s
I/O Bridge Single-ended 20 bytes 2.5 Ghz 50 GB/s
SMP Interconnect Single-ended 120 bytes 3.0 Ghz 360 GB/s
Total Bandwidth 590 GB/s

(Note that bandwidths shown are raw, peak signal bandwidths)
- Moving L3 onto POWER7 along with advances in signaling

technology enables significant raw bandwidth growth for both
memory and I/O subsystems. Note that advanced scheduling
improves POWER7’s ability to utilize memory bandwidth.
37

Socket Throughput
to Close Gap and
Exploit Long Term Investment
Achieve Potential
in Coherence Innovation

Technology

and Innovation

38
Exploit Long Term Investment in Coherence Innovation
Local SMP Links

Core Core Core Core
L2 Cache L2 Cache Remote SMP + I/O Links L2 Cache L2 Cache
Core Core Core Core
Using local and remote SMP links, up to 32 POWER7 chips are connected
39

Up to 32 POWER7 chips form a massive SMP system. do not imply that IBM will introduce
40
Coherence Protocol Features POWER7 Exploitation

- De-coupled global storage - POWER Servers can drive massive
coherence throughput.
- De-serialized storage updates - 32-chip system can manage over
allowing storage updates to be
reordered 20,000 concurrently reordered
coherent storage operations
- De-centralized coherence ~4X more POWER6 systems
resolution, and bounded latency - Minimal tracking overhead per
broadcast transport layer. operation.
- Decentralized coherence resolution, - Low latency intervention,

advanced cache states, optimized - High performance locking constructs
on-chip transport, and broadcast - robust scaling.
free barriers.

do not imply that IBM will introduce
41

Challenge: As system size grows, Coherence broadcast traffic increases
Compute Throughput
~5X
POWER6 High-End Server Global Scope POWER7 High-End Server
Virtualized Consolidation Platform Coherence UltraScale Cloud Platform
Broadcast
450
Compute Global Coherence
Throughput 1X GB/s Throughput
320
Global Coherence GB/s
Throughput
42

Solution: Speculative limited scope Coherence broadcast
- In 2003, recognized emerging trend
- Developed Dual-Scope Broadcast Coherence Protocol for POWER6
- Utilizes 13 cache states and integrated scope indicator in memory
POWER6 High-End Server Global Scope POWER7 High-End Server

Virtualized Consolidation Platform Coherence UltraScale Cloud Platform
Broadcast
Nodal Scope
Speculative
Coherence
Broadcast
Provides value for POWER6 Necessity for POWER7

- Latency reduction - 450 GB/s must grow to 1.6 TB/s
- Near Perfect Scaling for extreme to match POWER6 scaling
memory intensive workloads - 450 GB/s 3.6 TB/s theoretical peak
- Ultra-dense packaging (Power 575) - 3.6 TB/s 14.4 TB/s* Statements
with chip scope
regarding SMP servers
43
Summary: POWER7 maintains the Balance
Achieves extreme Multi-core Compute Throughput Potential

throughput while providing
Balance and SMP scaling
Exploit Long Term Investment
by building on a foundation
in Coherence Innovation
of solid innovation.

Technology

and Innovation

IBM POWER chips uniquely positioned

to excel given the emerging trends:
1) History of large SMP leadership
2) Storage Architecture economics
Multi-core evolution 3) High density packaging leadership
44
POWER7: Performance Estimates

POWER7 Continues Tradition of Excellent
Scalability Core Performance
¾ Core performance increased by:

Re-pipelined execution units
POWER6 SMT2
Reduced L1 cache latency
POWER7 SMT4
Tightly coupled L2 cache
Additional execution units
More flexible execution units
Increased pipeline utilization with SMT4 and Floating Pt. Integer Commercial
aggressive out of order execution
Chip Performance
¾ Chip Performance Improved Greater then 4X:
High performance on chip interconnect
Improved storage architecture
Dual high speed integrated memory controllers POWER6
POWER7 SMT4
¾ System
Achieves extreme Multi-core throughput while
providing Balance and SMP scaling by building
Floating Pt. Integer Commercial
on a foundation of solid innovation
Advanced SMP links will provide near linear * Performance estimates relate to processor
scaling for larger POWER7 systems. only and should not be used to estimate
projected server performance.
45
Energy Management: Architected Idle Modes
Two Design Points Chosen for Technology 4 PowerPC Architected States
Nap (optimized for wake-up time) RV Winkle

Power gate
Turn off clocks to execution units
Wake-Up Latency
Reduce frequency to core
Caches and TLB remain coherent
Fast wake-Up Sleep
Sleep (optimized for power reduction) Nap

Doze
Purge caches and TLB
Turn off clocks to full core and caches
Reduce voltage to V-retention
Energy Reduction
Leakage current reduced substantially
Voltage ramps-up on wake up
No core re-initialization required
46
Adaptive Energy Management: Energy ScaleTM

¾ Chip FO4 Tuned for Optimal Performance/Watt
in Technology
¾ DVFS (Dynamic Voltage and Frequency
Slewing)
-50% to +10% frequency slew independent SPECPower: Mean System Power per Load
per core Level
Frequency and voltage adjusted based on:
Work load and utilization.
On board activity monitors
¾ Turbo-Mode
Up to 10% frequency boost
AC Power
Leverages excess energy capacity from:
Non worst case work loads
Idle cores
Vmin
¾ Processor and Memory Energy Usage can be
independently Balanced.
Real time hardware performance monitors
100 80 60 40 20 0
used.
Load Level (%)
On board power proxy logic estimates power
¾ Power Capping Support
Allows budgeting of power to different parts
of system
47
Power Systems – Reliability,

Availability, Serviceability (RAS)
48
OS Downtime Comparison Survey

400 participants in 27 countries
Hours
10
9
8
7
6
5
4
3
2
1
0
Win2000 Win2003 RHEL SOLARIS HP-UX SUSE AIX
The Yankee Group “2007-2008 Global Server Operating Systems Reliability Survey” as quoted in “Windows Server: The New King of Downtime” by Mark
Joseph Edwards at www.windowsitpro.com/article/articleid/98475/windows-server-the-new-king-of-downtime.html, March 5, 2008 and in
http://www.sunbeltsoftware.com/stu/Yankee-Group-2007-2008-Server-Reliability.pdf
49
ITIC Survey says Power Systems with AIX deliver 99.997% uptime
- 54% of IT executives and managers say that they require 99.99% or better availability for their applications
¾ Power Systems with AIX delivers

the best RAS of UNIX, Linux,
Windows choices
1. Availability: The least amount of
downtime
15 minutes a year
2.3 times better than the
closest UNIX competitor
more than 10X better than
Windows
2. Reliability: The fewest
unscheduled outages
less than one outage per year
3. Serviceability: The fastest patch
time
11 minutes to apply a patch
Source: Network World, dated July 14, 2009, reports on the 2009 ITIC Global Server Hardware & Server OS Reliability Survey Results
© 2009 IBM Corporation

50
POWER7: Reliability and Availability Features

Dynamic Oscillator
Failover OSC0 OSC1
Fabric Bus Interface to other Chips and
Fabric Interface Nodes
¾ ECC protected
¾ Node hot add /repair
Core Recovery
BUF ¾ Leverage speculative execution resources to
enable recovery
BUF ¾ Error detected in GPRs FPRs VSR, flushed
and retried
¾ Stacked latches to improve SER
BUF
Alternate Processor Recovery

BUF ¾ Partition isolation for core checkstops
L3 eDRAM
¾ ECC protected
X8 Dimms IO Hub ¾ SUE handling
¾ Line delete
¾ 64 Byte ECC on Memory ¾ Spare rows and columns
Corrects full chip kill on X8 dimms
Spare X8 devices implemented GX IO Bus
¾ Dual memory chip failures do not cause PCI ¾ ECC protected
outage Bridge ¾ Hot add
¾ Selective memory mirror capability to recover
partition from dimm failures InfiniBand® Interface
¾ Hardware assisted scrubbing ¾ Redundant paths
¾ SUE handling
¾ Dynamic sparing on channel interface PCI Adapter * Statements regarding SMP servers
¾ PowerVM Hypervisor protected from full DIMM do not imply that IBM will introduce
failures a system with this capability.
51
Power Systems Benefits
¾ IBM Power Systems have a consistent, reliable history of executing

on schedule allowing customers to confidently plan for the future
¾ IBM Power Systems offer highest performance reducing the need
for additional resources
¾ IBM Power Systems are designed for performance with high
reliability and availability
Moving towards Continuous Availability – hardware and software
Reduced and shorter outages lower costs and improve SLAs
¾ Virtualization capabilities intrinsic to Power Systems design allows
improved service and lower costs by consolidating
POWER7 systems increased to up to 1000 partitions / system
POWER7 systems designed to leverage, exploit and enhance current
PowerVM capabilities
52
Summary
Power Systems™ continue strong
7th Generation Power chip:
Balanced Multi-Core design
EDRAM technology
SMT4
Greater then 4X performance in similar power
envelope as previous generation
Scales to 32 socket, 1024 threads balanced
system
Building block for peta-scale PERCS project
Achieves extreme Multi-core throughput while
providing Balance and SMP scaling by Power7 High Volume Card
building on a foundation of solid innovation
POWER7 Systems Running in Lab

AIX®, IBM i, Linux® all operational

53
Joel M. Tendler, Executive IT Architect

jtendler@us.ibm.com
Acknowledgment: This material is based upon work
supported by the Defense Advanced Research Projects
Agency under its Agreement No. HR0011-07-9-0002
Trademarks
The following are trademarks of the International Business Machines Corporation in the United States, other countries, or both.
Not all common law marks used by IBM are listed on this page. Failure of a mark to appear does not mean that IBM does not use the mark nor does it mean that the product is not
actively marketed or is not significant within its relevant market.
Those trademarks followed by ® are registered trademarks of IBM in the United States; all others are trademarks or common law marks of IBM in the United States.
For a complete list of IBM Trademarks, see www.ibm.com/legal/copytrade.shtml:
*, AS/400®, e business(logo)®, DBE, ESCO, eServer, FICON, IBM®, IBM (logo)®, iSeries®, MVS, OS/390®, pSeries®, RS/6000®, S/30, VM/ESA®, VSE/ESA,
WebSphere®, xSeries®, z/OS®, zSeries®, z/VM®, System i, System i5, System p, System p5, System x, System z, System z9®, BladeCenter®
The following are trademarks or registered trademarks of other companies.
Adobe, the Adobe logo, PostScript, and the PostScript logo are either registered trademarks or trademarks of Adobe Systems Incorporated in the United States, and/or other countries.
Cell Broadband Engine is a trademark of Sony Computer Entertainment, Inc. in the United States, other countries, or both and is used under license therefrom.
Java and all Java-based trademarks are trademarks of Sun Microsystems, Inc. in the United States, other countries, or both.
Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both.
Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered trademarks of Intel
Corporation or its subsidiaries in the United States and other countries.
UNIX is a registered trademark of The Open Group in the United States and other countries.
Linux is a registered trademark of Linus Torvalds in the United States, other countries, or both.
ITIL is a registered trademark, and a registered community trademark of the Office of Government Commerce, and is registered in the U.S. Patent and Trademark Office.
IT Infrastructure Library is a registered trademark of the Central Computer and Telecommunications Agency, which is now part of the Office of Government Commerce.
* All other products may be trademarks or registered trademarks of their respective companies.
Notes:
Performance is in Internal Throughput Rate (ITR) ratio based on measurements and projections using standard IBM benchmarks in a controlled environment. The actual throughput that any user will
experience will vary depending upon considerations such as the amount of multiprogramming in the user's job stream, the I/O configuration, the storage configuration, and the workload processed.
Therefore, no assurance can be given that an individual user will achieve throughput improvements equivalent to the performance ratios stated here.
IBM hardware products are manufactured from new parts, or new and serviceable used parts. Regardless, our warranty terms apply.
All customer examples cited or described in this presentation are presented as illustrations of the manner in which some customers have used IBM products and the results they may have achieved. Actual
environmental costs and performance characteristics will vary depending on individual customer configurations and conditions.
This publication was produced in the United States. IBM may not offer the products, services or features discussed in this document in other countries, and the information may be subject to change without
notice. Consult your local IBM business contact for information on the product or services available in your area.
All statements regarding IBM's future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.
Information about non-IBM products is obtained from the manufacturers of those products or their published announcements. IBM has not tested those products and cannot confirm the performance,
compatibility, or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products.
Prices subject to change without notice. Contact your IBM representative or Business Partner for the most current pricing in your geography.
55

POWER7 Processors: The Beat Goes On: Joel M. Tendler, Executive IT Architect

Uploaded by

Copyright:

Available Formats

POWER7 Processors: The Beat Goes On: Joel M. Tendler, Executive IT Architect

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

POWER7 Processors: The Beat Goes On: Joel M. Tendler, Executive IT Architect

Uploaded by

Copyright:

Available Formats

POWER7 Processors: The Beat Goes On

Joel M. Tendler, Executive IT Architect

IBM Power Systems value proposition

Deliver business value by leveraging technology

Performance Flexibility Reliability Affordability

. . . the highest value at the lowest risk

Approaching 20 Years of POWER Processors

1990 1995 2000 2005 2010

POWER Roadmap – The Only Reliable Server Roadmap

5GHz Alti Advanced

Distributed Distributed Switch

POWER7 Processor Chip

Computer Systems Memory: DRAM

Diagram courtesy of Wikipedia

Computer Systems Memory: SRAM

Diagram courtesy of Wikipedia

POWER7 Design Principles: Traditional Performance View

Multiple optimization Points

 Continued growth in socket

 Balanced core, cache, memory

POWER7 Design Principles:

1 Memory Controller 2 Memory Controllers 8 Memory Controllers

Computer Systems Memory: Hierarchy

Typical Memory Hierarchy

Processor intimacy, smaller capacity

Increased Latency, greater capacity

Challenge: Beating Physics to Realize Multi-core Potential

Local SMP Links

L2 Cache L2 Cache L2 Cache L2 Cache

Mem Ctrl L3 Cache and Chip Interconnect Mem Ctrl

Remote SMP + I/O Links

Core Core Core Core

Challenge: Beating Physics to Realize Multi-core Potential

Socket Throughput Limitation

Trends in Server Evolution

Driven by: 8-core 8-core

Traditional Entry Server Traditional High-End Server

- Economics 2 to 4 socket to traditional high-end

Enabled by: 8-core 8-core

Driven by: 8-core 8-core

- Economics 2 to 4 socket 8 to 32 socket

Traditional Entry Server Traditional High-End Server Same enablers and

Challenge: How does POWER7 maintain the Balance?

Cache Hierarchy Technology

Socket Throughput Limitation

Cache Hierarchy Technology and Innovation

Cache Hierarchy Rqmt Challenge

Core ... Core POWER4TM, POWER5TM, and

Large, Shared, 30+ MB

Cache Hierarchy Technology and Innovation

Solution: High speed eDRAM on the processor die

Conventional IBM ASIC IBM Custom Custom Custom

Dense, low power Off On High Area/power

Conventional Large, Off-chip On-processor On-processor Private core

Industry Standard Caching and Memory Technologies:

IBM’s POWER Servers have leveraged large off-chip

With POWER7, IBM introduces on-processor, high-speed,

Cache Hierarchy Technology and Innovation

Local SMP Links

Continued growth in socket

Balanced core, cache, memory