Intel Confidential Page 1 Intel 64-bit Server Technology Philip King Solution Specialist Intel...
-
date post
15-Jan-2016 -
Category
Documents
-
view
230 -
download
0
Transcript of Intel Confidential Page 1 Intel 64-bit Server Technology Philip King Solution Specialist Intel...
Intel Intel ConfidentialConfidential
Page 1
Intel 64-bit Server Technology
Philip KingSolution Specialist
Intel CorporationApril 21, 2023
Intel Intel ConfidentialConfidential
Page 2
Agenda
• Intel® Server Processors
• Intel® Xeon™ and Itanium™ Platforms
• Extended Memory 64 Technology
• Micro-architecture comparisons
• Performance and Scalability
• Dual and Multi-core plans
• 64-bit Platform selection guidance
Intel Intel ConfidentialConfidential
Page 3
IntelIntel®® Server Processor Roadmap Server Processor Roadmap
®®
Pe
rfo
rma
nc
e , R
AS
, Sc
ala
bili
ty
2002 2004
. . .. . .
Itanium® 2 Itanium® 2 ProcessorProcessor
Itanium® 2 Itanium® 2 ProcessorProcessor
Xeon™Xeon™ProcessorProcessorXeon™Xeon™
ProcessorProcessor
2003 2005
Itanium 2 Itanium 2 Processor Processor
(Madison)(Madison)
Itanium 2 Itanium 2 Processor Processor
(Madison)(Madison)
. . .. . .
NoconaNocona(8/6/04)(8/6/04)
NoconaNocona(8/6/04)(8/6/04)
PotomacPotomac(Q2’05)(Q2’05)
PotomacPotomac(Q2’05)(Q2’05)
•.18µ• 3MB iL3 cache
• .13µ• 6MB iL3 cache
• 90nm• 24MB iL3 cache•Dual-core
• HT• .13µ•2MB iL3 cache
• HT; EM64T• 1MB iL2 cache•> 3.6 Ghz; 90nm;
2006-Beyond
MontecitoMontecito(Q3’05)(Q3’05)
MontecitoMontecito(Q3’05)(Q3’05)
Itanium 2 Itanium 2 Processor Processor
(Madison Refresh)(Madison Refresh)
Itanium 2 Itanium 2 Processor Processor
(Madison Refresh)(Madison Refresh)
• .13µ• 9MB iL3 cache
• HT; EM64T• 4MB+ iL3 cache•> 3 Ghz; 90nm;
Future Direction:
• Larger Caches
• 65nm process
• Multiple cores
• Multi-threading
• Virtualization
• >1 billion trans.
Intel Intel ConfidentialConfidential
Page 4
Intel® Xeon™ Platform
2004 Enhancements2004 Enhancements
2003 & Prior 2003 & Prior EnhancementsEnhancements
Future Future EnhancementEnhancement
Hyper-Threading Hyper-Threading TechnologyTechnology
Intel® Netburst® Intel® Netburst® micro-architecturemicro-architecture
SSE & SSE2 SSE & SSE2 instructionsinstructions
VirtualizationVirtualization User defined power User defined power
thresholdsthresholds Dual core CPUsDual core CPUs Fully buffered Fully buffered
DIMMsDIMMs
New SSE3 instructionsNew SSE3 instructions PCI Express*PCI Express* 800MHz FSB800MHz FSB 64-bit extension technology64-bit extension technology
Power management technologyPower management technology Demand Based Switching (DBS)Demand Based Switching (DBS)
DDR2 memoryDDR2 memory Higher capacity & performance w/ lower powerHigher capacity & performance w/ lower power
Intel, Netburst and Xeon 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.
Intel Intel ConfidentialConfidential
Page 5
Intel® Itanium® Platform
2005 Planned Enhancements2005 Planned Enhancements
2004 & Prior 2004 & Prior EnhancementsEnhancements
EPIC architectureEPIC architectureEnhanced Machine Enhanced Machine
Check ArchitectureCheck ArchitectureFMAC for floating-point FMAC for floating-point
leadershipleadershipLargest on-die resources Largest on-die resources
for demanding workloadsfor demanding workloads
Multi-coreMulti-core VirtualizationVirtualization I/O and memory I/O and memory Enhanced RASEnhanced RAS Cost parity with Xeon processor Cost parity with Xeon processor
based platforms (via common based platforms (via common platform infrastructure)platform infrastructure)
Up to 2X higher performance than Up to 2X higher performance than Xeon based platforms Xeon based platforms
Dual-coreDual-core Multi-threadingMulti-threading Power management technologyPower management technology
Demand Based Switching (DBS)Demand Based Switching (DBS)Automatic Control Power Consumption (ACPC)Automatic Control Power Consumption (ACPC)
Enhanced system bus BandwidthEnhanced system bus Bandwidth
Foxton Technology (performance feature)Foxton Technology (performance feature) Pellston Technology (cache reliability)Pellston Technology (cache reliability) PCI Express*PCI Express* Fully-buffered DIMMS (FBD) , DDR2Fully-buffered DIMMS (FBD) , DDR2
Future Future EnhancementsEnhancements
And lower power than today
Intel Intel ConfidentialConfidential
Page 6
What is Intel® Extended Memory 64 Technology?
Evolutionary IA-32 architectural enhancements to support Evolutionary IA-32 architectural enhancements to support extended memory past 4 GBextended memory past 4 GB
Evolutionary IA-32 architectural enhancements to support Evolutionary IA-32 architectural enhancements to support extended memory past 4 GBextended memory past 4 GB
Additional RegistersAdditional Registers8-SSE & 8-Gen Purpose8-SSE & 8-Gen Purpose
Additional RegistersAdditional Registers8-SSE & 8-Gen Purpose8-SSE & 8-Gen Purpose
64-bit Integer Support64-bit Integer Support(Double Length)(Double Length)
64-bit Integer Support64-bit Integer Support(Double Length)(Double Length)
Extended MemoryExtended MemoryAddressabilityAddressability
64-bit Pointers, 64-bit 64-bit Pointers, 64-bit RegistersRegisters
Extended MemoryExtended MemoryAddressabilityAddressability
64-bit Pointers, 64-bit 64-bit Pointers, 64-bit RegistersRegisters
++ ==With Intel® With Intel®
EM64TEM64TSupport for flat virtualSupport for flat virtual
Address spaceAddress space
Support for flat virtualSupport for flat virtualAddress spaceAddress space
64-bit64-bit64/6464/64
64-bit64-bit64/6464/64
Compatibility ModeCompatibility Mode64 OS / 32 Apps64 OS / 32 Apps
64-bit Mode64-bit Mode64 OS / 64 Apps64 OS / 64 Apps
Compatibility ModeCompatibility Mode64 OS / 32 Apps64 OS / 32 Apps
64-bit Mode64-bit Mode64 OS / 64 Apps64 OS / 64 Apps
Legacy ModeLegacy Mode32 OS / 32 Apps32 OS / 32 Apps
Legacy ModeLegacy Mode32 OS / 32 Apps32 OS / 32 Apps
FeaturesFeatures
++
ModesModes
Intel Intel ConfidentialConfidential
Page 7
CISC
RISC
Superscalar
EPICEPIC
Per
form
ance
Time
• Largest, most demanding workloads requires new approach
• Benefits from the experience of past architectures
• A convergence of the best minds in the industry
• Maximizes instructions executed in parallel
• Multiple execution units and issue ports
• Large and fast on-die cache • 128 general registers, 128 floating point
registers, 8 branch• Efficient management engine
– Register stack engine– 4 GB page size
• Modular• Able to seamlessly add
execution resources, issue ports
Massive on-chip Massive on-chip resourcesresources
Performance Performance through parallelismthrough parallelism ScalableScalable
IBM 370, VAX 11*IBM 370, VAX 11*Age : 20+Age : 20+
Sun SPARC, MIPS R4000*Sun SPARC, MIPS R4000*Age : 10-15+Age : 10-15+
IntelIntel®® ItaniumItanium®®
ProcessorProcessor Age : 2+Age : 2+
IBM PowerPC*IBM PowerPC*Age : 9+Age : 9+
Next Enterprise Architecture
* Source: * Source: Computer Organization and Architecture, 1999 W. StallingsComputer Organization and Architecture, 1999 W. Stallings
Intel Itanium architecture built from the ground up to meet the needs of the most demanding applications
Intel Intel ConfidentialConfidential
Page 8
Characteristics of High-end ProcessorsHigh-end Processors Require Significant Resources, Capabilities
AMD Opteron*AMD Opteron*
1 MB on-1 MB on-diedie
cachecache72 72 RegistersRegisters
2 / 4 MB 2 / 4 MB page sizepage size
6 MB 6 MB 9 MB 9 MB on-die cacheon-die cache
4 GB 4 GB page sizepage size
264264RegistersRegisters
ItaniumItanium®® 2 2 processorprocessor
1 MB1 MBon-die on-die cachecache4 MB 4 MB
page sizepage size
160160RegistersRegisters
Sun Sun UltraSPARC UltraSPARC
IIIi* IIIi* Up to Up to 1.5 MB1.5 MBon-die on-die cachecache72 72
RegistersRegisters
IBM IBM Power* 4Power* 4
16 MB16 MBpage sizepage size
Source: IBM.comSource: IBM.com Source: sun.comSource: sun.com Source: AMD.com,Source: AMD.com,
Intel Intel ConfidentialConfidential
Page 9
Hyper-ThreadingHyper-ThreadingTechnologyTechnology
Intel Enterprise Micro-ArchitecturesXeon™ EM64TEM64T
6.4 GB/s6.4 GB/s
1TB 40 bit1TB 40 bit
1 MB1 MB
2 2x Integer2 2x Integer
1 1x Integer, 1 1x Integer, 1 MMx & SSE1 MMx & SSE
2 2 FloatingFloating
PointPoint
~3.6 GHz~3.6 GHz
Itanium® 2 Processor 6M
6.4 GB/s6.4 GB/s
1024 TB1024 TB
88
1 2 3 4
Memory AddressingMemory Addressing
1 2 3 4 5 6 7 8 9 1011
System Bus BandwidthSystem Bus Bandwidth
On-die CacheOn-die Cache
Pipeline StagesPipeline Stages
On-die RegistersOn-die Registers
Execution UnitsExecution Units
Core FrequencyCore Frequency
Issue PortsIssue Ports
Performance via Megahertz Performance via Parallelism
On-die multi-threadOn-die multi-thread
264 Application Registers264 Application Registers+ 64 Predicate Registers*+ 64 Predicate Registers*
6 Instructions / Cycle6 Instructions / Cycle
24 Registers24 Registers
2020
Hyper-ThreadingHyper-ThreadingTechnologyTechnology
3 Instructions / Cycle3 Instructions / Cycle
6 MB6 MB
Instructions / ClkInstructions / Clk
6 Integer, 6 Integer, 3 Branch3 Branch
2 FP, 2 FP, 1 SIMD1 SIMD
2 Load and 2 Load and 2 Store2 Store
** Intel’s EPIC technology includes 64 single-bit predicate registers Intel’s EPIC technology includes 64 single-bit predicate registers to accelerate loop unrolling and branch intensive code execution. to accelerate loop unrolling and branch intensive code execution.
1.5 GHz1.5 GHz
Intel Intel ConfidentialConfidential
Page 10
Intel® Server Processor PlatformsPerformancePerformance 1 1
On track to deliver 1.5-2X better performance On track to deliver 1.5-2X better performance than Intel® Xeon™ processorthan Intel® Xeon™ processor
While achieving platform cost parity While achieving platform cost parity via common platform infrastructurevia common platform infrastructure
All products, dates and information arepreliminary and subject to change without notice.
Platform CostPlatform Cost 22
‘‘0404 ’’07+07+
+30%-60% or +30%-60% or higher in ’04higher in ’04
+0% in ’07++0% in ’07+
Source : www.ioncomputers.com, www.dell.com ION SR4004, (4) Xeon processors 2.8 GHz, 2MB cache, 24 GB system memory, 36 GB HDD, no OS - $34,616ION I2X4, (4) Itanium 2 processors 1.5 GHz, 6 MB cache, 24 GB system memory, 36 GB HDD, no OS - $44, 950Dell PowerEdge 2650, (2) Xeon processors 3.2 GHz, 1 MB cache, 4 GB system memory, 36 GB HDD, no OS - $6,143Dell PowerEdge 3250, (2) Itanium 2 processors 1.4 GHz, 1.5 MB cache, 4 GB system memory, 36 GB HDD, no OS - $9,499
1 Data based on Intel projections.2 ‘04 Price based on comparable OEM systems, HW only for enterprise and technical computing applications.
‘‘0404 ’’07+07+
+30%-50%+30%-50%in ‘04in ‘04
+50%-100%+50%-100%in ’07+in ’07+
Intel Intel ConfidentialConfidential
Page 11
Multi-Core Technology
• Intel™ manufacturing leadership (90nm, 65nm) enables leading multi-core• Itanium™ architecture has smaller core size – enabling up to 2x more
cores per die than IA-32 for higher performance at same cost
2004 2005 2007 Single Core Dual Core Multi-Core
+ Cache
+ Cache
Cache
Core
4 or more cores
+ Cache
CoreCache
2 or more cores
+ Cache
2X more
cores
All products, dates and features are preliminary and subject to change without notice
ItaniumItanium® ® architecture expected to enable up to 2x more architecture expected to enable up to 2x more cores per processor than Xeon processors by 2007cores per processor than Xeon processors by 2007
ItaniumItanium® ® architecture expected to enable up to 2x more architecture expected to enable up to 2x more cores per processor than Xeon processors by 2007cores per processor than Xeon processors by 2007
Intel Intel ConfidentialConfidential
Page 12
Which is the right Architecture?Intel® Xeon™ Processor*Intel® Xeon™ Processor* Intel® Itanium® 2 Processor FamilyIntel® Itanium® 2 Processor Family
• Performance leadership now and getting greater in the future
• Mainframe class reliability features• Best scalability• Over 1600 Applications ported• End to end 64 bit computing
• Best Price/performance• High availability• Scalable• Broadest 32bit S/W availability• Large install base, but all 32bit• 64bit addressability only
Intel Xeon ProcessorIntel Xeon Processor Intel Itanium 2 ProcessorIntel Itanium 2 Processor
64 bit 64 bit AddressabilityAddressability
Legacy Appsneed need 64bit 64bit
addressability addressability
*By 2005 all Xeon processors will have memory extensions
High ComputeServers
don’t need don’t need 64bit 64bit
addressability addressability
Leave applications as 32BitLeave applications as 32Bit Port to EM64T or Port to EM64T or Port to IPFPort to IPF
64 bit 64 bit AddressabilityAddressability
Intel Intel ConfidentialConfidential
Page 13
Attribute Intel® Xeon™ processor family Intel® Itanium® processor familyPerformance Best for workgroup, workstation, web
server and IA-32 legacy applications• Out of order execution• Netburst uArchitecture• High frequency (3GHz+)
Best for large database and technical computing workloads• EPIC architecture • Large integrated L3 cache• FMACs / FP engine
Physical Addressing
Expanded memory for legacy solutions• 40 bits (1 TB)
Memory capacity for largest SMP• 50 bits (1024 TB or 1 PB)
Ecosystem Best choice of legacy IA-32 solutions• Mature ecosystem of 32-bit apps, tools, OS• Nascent 64-bit extension ecosystem:
Systems & solutions for largest scalability and highest reliability solutions• Production 64-bit applications, tools & OS • Range of large SMP platforms
RAS Reliable data integrity• ECC & parity coverage on major data arrays
Data integrity & high availability to replace RISC / mainframe at a fraction of cost• ECC & parity on most data arrays & FSB ECC• Enhanced machine check architecture• Fault-tolerant, error-hardened OEM platforms
Architecture Enhanced architecture support• 16 general purpose & 16 SSE registers• 64-bit system MSRs• 64-bit pointers
Architected for enterprise & future headroom •EPIC architecture w/ predication & speculation•264 application registers•64-bit pointers
Platform Leading edge technologies for high performance & balanced platform• DDR2, Faster buses• PCI Express*
End user investment protection• Stable platform (2-3 years+)• Same bus / socket through Montecito• High bandwidth bus; future DDR2 & PCI-Express*
Platform Feature / Benefit Comparison
Intel Intel ConfidentialConfidential
Page 14
T H A N K Y O U
Intel Intel ConfidentialConfidential
Page 15
B A C K U P
Intel Intel ConfidentialConfidential
Page 16
Other 64-bit platform considerations
• Some tools available and will grow over time
• Effort needed to porting to 64 bit is the same regardless of architecture
• Strong 32-bit eco-systems in place, but needs to expand to EM64T
• OEMs are ramping volume quickly
• Best price point for servers
• Native 32 bit environment (drivers still need be ported to 64bit environment)
• Mature Tools and large selection
• Porting effort is about the same as on EM64T
• Eco-system in place and still growing
• OEMs shipping in volume
• Processor price delta between Xeon and IPF (top speed) is <$400 per CPU
• IA32-EL provides 32-bit execution environment – perf. 50% of Xeon top speed
Xeon Processor with EM64TXeon Processor with EM64T Itanium 2 ProcessorItanium 2 Processor
Intel Intel ConfidentialConfidential
Page 17
Server Architecture Reliability Comparisons
Characteristic Itanium® 2 Processor
IBM Power*
Intel® Xeon™ MP Processor
Sun Ultra-
Sparc*
Intel® Xeon™
ProcessorOpteron
Error recovery on data bus (ECC or retry) 2005 Internal soft error logic check 2005 2004
Lockstep support Bad/poisoned data containment Cache Reliability (Pellston) 2005 Memory SDEC, retry on double-bit Memory spares Partitioning node core node node
Electrical isolated partitions node node node
Intel Intel ConfidentialConfidential
Page 18
0
100
200
300
400
500
600
700
4P 32P
+33%
+100%
Itanium® 2 processor 6M scales better than Xeon™ processor MP 3GHz
1 4P: Source www.tpc.org: IBM x365 102,667.42 tpmC, $3.52/tpmC, available: 3/31/2004, with 4 Intel® Xeon processors MP at 3GHz, each with 4MB iL3 cache, running, Microsoft Windows Server 2003 Enterprise Server, Microsoft SQL Server 2000 EE SP3, 32GB memory. Itanium® 2 processor results of 136,111 tpmC at $4.09/tpmC on HP Integrity Server rx5670 with 4 Itanium® 2 processors 6M at 1.5GHz, each with 6MB L3 cache, 96GB memory, Red Hat Enterprise Linux Advanced Server 3 and Oracle Database 10g Enterprise Edition; TPC-C availability date 3/5/04.
2 32P: Source www.tpc.org: Unisys ES7000 Orion 540 Enterprise Server 304,148.50 tpmC @ $6.18/tpmC, availability 4/28/04 with 32 Intel® Xeon™ processors MP at 3GHz, each with 4MB iL3 cache, running Microsoft* Windows* Server 2003 Datacenter Edition, Microsoft* SQL Server 2000 Enterprise Edition. Itanium® 2 processor 6M results on NEC Express5800/1320Xd C/S w/Express5800/120Rf-2, 609,467tpmC, $6.78/tpmC, with thirty two (32) Intel Itanium® 2 processors 6M at 1.5 GHz, running SUSE LINUX Enterprise Server 9 and Oracle Database 10g Enterprise Edition, with 512 GB RAM; Available: 9/1/2004.Results as of 4/29/04. 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. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, reference www.intel.com/procs/perf/limits.htm or call (U.S.) 1-800-628-8686 or 1-916-356-3104
IA Scalability Performance Comparison: Transaction Processing (TPC-C)
Xeon™ Processor MP 3GHz 4MB L3
Itanium® 2 Processor1.5GHz 6MB L3
tpm
C in
th
ou
san
ds
*Other names and brands may be claimed as property of others