presentation of the 16th list
TRANSCRIPT
1
Presentation of the 16th List
Hans Werner Meuer, University of Mannheim Erich Strohmaier, University of TennesseeJack J. Dongarra, University of TennesseHorst D. Simon, NERSC/LBNL
SC2000, Dallas, TX, November 4-10, 2000 BoF Session, November 8, 5:30pm
TOP500 BoFTOP500 BoF
• Presentation of the 16th Top500 ListPresentation of the 16th Top500 List• TOP500 Background Information
and general Market Overview.• Geographical Distribution• Manufacturers• Performance Development• Architectures, Technology• Future TOP500 Ideas and Plans
2
TOP500 Motivation
• Basis for analysing the HPC market
• Quantification of observations
• Detection of trends
– Market
– Architecture
– Technology
TOP500TOP500- Listing of the 500 most powerfulComputers in the World
- Yardstick: Rmax from LINPACK MPPAx=b, dense problem
- Updated twice a yearSC‘xy in the States in NovemberMeeting in Mannheim, Germany in June
- All data available from www.top500.org
8N_J
7FYJ
955�UJWKTWRFSHJ
3
,Q������D�FRPSXWDWLRQ�WKDW�WRRN���IXOO�\HDU�WR�FRPSOHWHFDQ�QRZ�EH�GRQH�LQ���PRQWK�
Fastest Computer Over Time
TMCCM-2(2048)
Fujitsu VP-2600
Cray Y-MP (8)
05
101520253035404550
1990 1992 1994 1996 1998 2000
Year
GF
lop/
s
,Q������D�FRPSXWDWLRQ�WKDW�WRRN���IXOO�\HDU�WR�FRPSOHWHFDQ�QRZ�EH�GRQH�LQ���GD\V�
Fastest Computer Over Time
HitachiCP-PACS
(2040)IntelParagon
(6788)
FujitsuVPP-500
(140)TMC CM-5(1024)
NEC SX-3(4)
TMCCM-2(2048)
Fujitsu VP-2600
Cray Y-MP (8)
050
100150200250300350400450500
1990 1992 1994 1996 1998 2000
Year
GF
lop/
s
4
,Q������D�FRPSXWDWLRQ�WKDW�WRRN���IXOO�\HDU�WR�FRPSOHWHFDQ�WRGD\�EH�GRQH�LQ���KRXU�
Fastest Computer Over TimeASCI White
Pacific(7424)
ASCI BluePacific SST
(5808)
SGI ASCIBlue
Mountain(5040)
Intel ASCI Red
(9152)Hitachi
CP-PACS(2040)
IntelParagon(6788)
FujitsuVPP-500
(140)
TMC CM-5(1024)
NEC SX-3
(4)
TMCCM-2(2048)
Fujitsu VP-2600
Cray Y-MP (8)
Intel ASCIRed Xeon
(9632)
0500
100015002000250030003500400045005000
1990 1992 1994 1996 1998 2000
Year
GFl
op/s
- Author since 6/1993 and “11/2000”
Jack Dongarra Horst Simon
TOP500 AuthorsTOP500 Authors- Started in spring 1993 by:
Hans W. Meuer and Erich Strohmaier
5
TOP500 Status
• 1. List in June 1993
• Accepted by users and manufacturers
• 16. List on November 2, 2000
6
TOP500 list TOP500 list -- Data shownData shown• Manufacturer Manufacturer or vendor
• Computer Type indicated by manufacturer or vendor
• Installation Site Customer
• Location Location and country
• Year Year of installation/last major update
• Customer Segment Academic,Research,Industry,Vendor,Class.
• # Processors Number of processors
• Rmax Maxmimal LINPACK performance achieved
• Rpeak Theoretical peak performance
• Nmax Problemsize for achieving Rmax
• N1/2 Problemsize for achieving half of Rmax
� 1ZRUOG���� 3RVLWLRQ�ZLWKLQ�WKH�723����UDQNLQJ
TOP10Rank Manufacturer Computer
Rmax
[TF/s]Installation Site Country Year Area of
Installation# Proc
1 IBM ASCI WhiteSP Power3
4.93 Lawrence LivermoreNational Laboratory
USA 2000 Research 8192
2 Intel ASCI Red 2.38 Sandia National Laboratory USA 1999 Research 9632
3 IBM ASCI Blue PacificSST, IBM SP
604E
2.14 Lawrence LivermoreNational Laboratory
USA 1999 Research 5808
4 SGI ASCI BlueMountain
1.61 Los AlamosNational Laboratory
USA 1998 Research 6144
5 IBM SP Power3375Mhz
1.42 IBM/Naval OceanographicOffice (NAVOCEANO)
USA 2000 Research 1336
6 IBMSP Power3
375Mhz1.18
National Centers forEnvironmental Predicition
USA 2000 Research 1104
7 Hitachi SR8000-F1 1.04 Leibniz Rechenzentrum Germany 2000 Academic 112
8 IBM SP Power3375MHz 8way
0.93 San Diego SupercomputerCenter
USA 2000 Academic 1152
9 Hitachi SR8000-F1 0.92 High Energy AcceleratorResearch Organization/ KEK
Japan 2000 Research 100
10 Cray Inc. T3E 1200 0.89 Government USA 1998 Classified 1084
7
TOP10Rank Manufacturer Computer
Rmax
[TF/s]Installation Site Country Year Area of
Installation# Proc
1 IBM ASCI WhiteSP Power3
4.93 Lawrence LivermoreNational Laboratory
USA 2000 Research 8192
2 Intel ASCI Red 2.38 Sandia National Laboratory USA 1999 Research 9632
3 IBM ASCI Blue PacificSST, IBM SP
604E
2.14 Lawrence LivermoreNational Laboratory
USA 1999 Research 5808
4 SGI ASCI BlueMountain
1.61 Los AlamosNational Laboratory
USA 1998 Research 6144
5 IBM SP Power3375Mhz
1.42 IBM/Naval OceanographicOffice (NAVOCEANO)
USA 2000 Research 1336
6 IBMSP Power3
375Mhz1.18
National Centers forEnvironmental Predicition
USA 2000 Research 1104
7 Hitachi SR8000-F1 1.04 Leibniz Rechenzentrum Germany 2000 Academic 112
8 IBM SP Power3375MHz 8way
0.93 San Diego SupercomputerCenter
USA 2000 Academic 1152
9 Hitachi SR8000-F1 0.92 High Energy AcceleratorResearch Organization/ KEK
Japan 2000 Research 100
10 Cray Inc. T3E 1200 0.89 Government USA 1998 Classified 1084
Architectures
Single Processor
SMP
MPP
SIMD Constellation Cluster - NOW
0
100
200
300
400
500
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-9
5
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-9
8
Jun-9
9
Nov-99
Jun-0
0
Nov-00
CluMPs
Y-MP C90
Sun HPC
Paragon
CM5T3D
T3E
SP2
Cluster ofSun HPC
ASCI Red
CM2
VP500
SX3
8
Excerpt from TOP500
Rank Manufacturer Computer Rmax [GF/s]
Installation Site Country Year Area of Installation
# Proc
… … … … … … … …
80 IBM LosLobos 237 University of New Mexico USA 2000 Academic 512
… … … … … … … … …
84 Self-made Cplant Cluster 232 Sandia Nationla Lab USA 1999 Research 580
… … … … … … … … …
96 HPTi ACL-276 196 NOAA USA 1999 Research 276
… … … … … … … … …
126 Self-made CLIC PIII 143 Technical University of
Chemnitz Germany 2000 Academic 528
… … … … … … … … …
413 Self-made NT SuperCluster
63 NCSA USA 1999 Academic 256
Performance Development
88.0 TF/s
1.167 TF/s
59.7 GF/s
4.94 TF/s
0.4 GF/s
55.1 GF/s
Jun-
93
Nov-93
Jun-
94
Nov-94
Jun-9
5
Nov-9
5
Jun-9
6
Nov-9
6
Jun-9
7
Nov-9
7
Jun-9
8
Nov-98
Jun-
99
Nov-99
Jun-
00
Nov-00
Intel XP/S140Sandia
Fujitsu’NWT’ NAL
SNI VP200EXUni Dresden
Hitachi/TsukubaCP-PACS/2048
IntelASCI Red
Sandia
IBMASCI White
LLNLN=1
N=500
SUM
IBM SP PC604e130 processors
Alcatel
1 Gflop/s
1 Tflop/s
100 Mflop/s
100 Gflop/s
100 Tflop/s
10 Gflop/s
10 Tflop/s
9
Performance Development
ASCI
Earth Simulator
Jun-9
3
Jun-9
4
Jun-9
5
Jun-9
6
Jun-9
7
Jun-9
8
Jun-9
9
Jun-0
0
Jun-0
1
Jun-0
2
Jun-0
3
Jun-0
4
Jun-0
5
N=1
N=500
Sum
1 GFlop/s
1 TFlop/s
1 PFlop/s
100 MFlop/s
100 GFlop/s
100 TFlop/s
10 GFlop/s
10 TFlop/s
Continents
USA/Canada
Europe
Japan
others
0
100
200
300
400
500
Jun-
93
Nov-9
3
Jun-9
4
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-9
7
Nov-97
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-00
10
Continents - Performance
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Jun-
93
Nov-9
3
Jun-9
4
Nov-94
Jun-9
5
Nov-9
5
Jun-9
6
Nov-96
Jun-
97
Nov-97
Jun-9
8
Nov-98
Jun-
99
Nov-9
9
Jun-
00
Nov-00
Per
form
ance
USA/Canada
other
Japan
Europe
Europe - Countries
Germany
UK
France Scandinavia
BeneluxSwitzerland
others
0
50
100
150
Jun-
93
Nov-9
3
Jun-
94
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-0
0
11
Producers
USA
JapanEurope
0
100
200
300
400
500
Jun-
93
Nov-9
3
Jun-
94
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-0
0
Producers - Performance
0%10%20%30%40%50%60%70%80%90%
100%
Jun-
93
Nov-9
3
Jun-
94
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-0
0
Perf
orm
an
ce
USA
Japan
Europe
12
Manufacturer
Cray
SGI
IBM
Sun
Convex/HP
TMC
Intel
FujitsuNEC
Hitachiothers
0
100
200
300
400
500
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-95
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-98
Jun-9
9
Nov-99
Jun-0
0
Nov-00
Manufacturer
Cray
SGI
IBM
SunConvex/HP
TMC
Intel
Fujitsu
NEC
Hitachiothers
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-95
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-98
Jun-9
9
Nov-99
Jun-0
0
Per
form
ance
13
www.top500.org
Kflops per Inhabitant
183
111 10190 89 74
49 29
509
0
100
200
300
400
500
600
US
A
Sca
ndin
avia
Ger
man
y
UK
Japa
n
Sw
itzer
land
Fran
ce
New
Zea
land
Luxe
mbo
urg
14
Customer TypeVendor
Research
Industry
Academic
Classified
0
100
200
300
400
500
Jun-
93
Nov-9
3
Jun-
94
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-0
0
Customer Types - Performance
Research
Industry
Academic
ClassifiedVendor
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Jun-
93
Nov-9
3
Jun-
94
Nov-9
4
Jun-
95
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-
98
Nov-9
8
Jun-
99
Nov-9
9
Jun-
00
Nov-0
0
Pe
rfo
rma
nc
e
15
Industrial Customer Segments
Engineering
Commercial
Unknown
0
50
100
150
200
250
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-95
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-98
Jun-9
9
Nov-99
Jun-0
0
Nov-00
Processor Type
Scalar
Vektor
SIMD
0
100
200
300
400
500
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-9
5
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-9
8
Jun-9
9
Nov-99
Jun-0
0
Nov-00
16
Chip Technology
CMOS/off the shelf
CMOS/ proprietary
ECL
0
100
200
300
400
500
Jun-
93
Nov-9
3
Jun-9
4
Nov-9
4
Jun-9
5
Nov-9
5
Jun-
96
Nov-9
6
Jun-
97
Nov-9
7
Jun-9
8
Nov-98
Jun-
99
Nov-9
9
Jun-
00
nov00
0
Chip Technology
Alpha
PowerHP
intel
MIPS
SUNother COTS
proprietary
0
100
200
300
400
500
Jun-9
3
Nov-9
3
Jun-9
4
Nov-9
4
Jun-9
5
Nov-9
5
Jun-9
6
Nov-9
6
Jun-9
7
Nov-9
7
Jun-9
8
Nov-9
8
Jun-9
9
Nov-9
9
Jun-0
0
Nov-0
0
17
TOP Centers
• Two approaches (No ‘classified’ centers):
• Sum up all TOP500 systems for each site– Snapshot only
• “Try” on historical perspective:– Sum up all systems over all lists all lists * 0.5 years* 0.5 years
= Total deliverable Linpack flop years
– This give a ‘discounted’ credit for history
TOP10 Centers - “snap-shot”Rank Site
∑ Rmax
[TF/S]Current Best Country
1 Lawrence Livermore National Laboratory 7.7 1, 3, 32, 36 USA
2 Sandia National Laboratory 2.38 2, 408 USA
3 Los Alamos National Laboratory 2.36 4, 19, 401 USA
4 Naval Oceanographic Office(NAVOCEANO)
2.17 5, 21, 326 USA
5 US Army Research Laboratory (ARL) 1.32 28, 138, 253-258, 424, 425
USA
6 Charles Schwab 1.28 15, 34 USA
7 ECMWF 1.20 12, 90, 214 UK
8 National Center for EnvironmentalPrediction
1.18 9 USA
9 Leibniz Rechenzentrum 1.14 11, 196 Germany
10 University of Tokyo 1.11 13, 85 Japan
18
TOP10 Centers - “integrated”
Rank Site∑ t*Rmax
[TFYears]Current Best Country
1 Sandia National Laboratory 8.0 2, 408 USA
2 Lawrence Livermore National Laboratory 7.5 1, 3, 32, 36 USA
3 Los Alamos National Laboratory 6.4 4, 19, 401 USA
4 University of Tokyo 3.3 13, 85 Japan
5 Naval Oceanographic Office(NAVOCEANO)
2.9 5, 21, 326 USA
6 United Kingdom Meteorological Office 2.7 26, 29 UK
7 Forschungszentrum Juelich (FZJ) 2.1 40, 83 Germany
8 NERSC/LBNL 2.0 42, 60 USA
9 Oak Ridge National Laboratory 1.9 17, 70, 402 USA
10 ECMWF 1.8 12, 90, 214 UK
Performance Distribution
0
1
2
3
4
5
1 10 100 1000Rank
Per
form
ance
[TF
lop/
s]
19
Cumulative Performance
0102030405060708090
100
1 10 100 1000Rank
Pe
rfo
rma
nc
e [
TF
lop
s]
88 TF/s
Rank 1/2
Performance Distribution
Rank of 1/2 TOP500 Performance
0
10
20
30
40
50
60
70
Jun-9
3
Nov-93
Jun-9
4
Nov-94
Jun-9
5
Nov-95
Jun-9
6
Nov-96
Jun-9
7
Nov-97
Jun-9
8
Nov-98
Jun-9
9
Nov-99
Jun-0
0
Nov-00
20
What’s new with the TOP500?
•• HPLHPL: High Performance LinpackAntoine Petitet and Clint Whaley, ICL, UTK
– icl.cs.utk.edu/hpl
– Needs only• MPI
• BLAS or VSIPL
– Highly scalable and efficient for the whole range of system sizes we see
What’s new with the TOP500?
• New database interface at www.top500.org– More categories and variables to select or list
– experimental cross-linked lists
• Ad’s corner(Ad Emmen)
• Architecture descriptions(Aad van der Steen)
21
What’s next with the TOP500?
• What’s now?
• New application areas of ‘Supercomputers’
• New architectures for ‘Supercomputers’
• ‘Self-made’ systems
Implications
• Commercial DB systems– “don’t care about floating point”
• New architectures - ‘compute farms’– “not designed to run Linpack”
• “Self-made” cluster– no vendor to measure Linpack
– every system is different
22
TOP500 Plans
• Talks with experts since 3 years on how to ‘resolve’ these issues
• Q: “How can we provide better information for different application and architecture domains?”
• A: “Make additional lists”
TOP500 Plans
• Keep current TOP500– Continuity, trends, 16 existing editions
– 17th: Supercomputer’2001Supercomputer’2001Mannheim/Heidelberg, Germany, June 21-23, 2001
• The current TOP500 will remain ‘center’ of future activities
23
TOP500 Plans
• Assemble additional lists for special purposes - first try:– “TOP100 (commodity) cluster”
– Jointly with TFCC
– But what is a cluster ???
– Which benchmark?• Initially without
TOP500 Plans
• Collaborate with research centers (NERSC, ICL, ...) which develop new benchmarks
• Feedback about benchmarks
• TOP500.org will provide – free access to benchmark results and
– possibilities for ‘re-ranking’