Computing: the next decade
Lennart JohnssonDepartment of Computer Science
University of Houston
TransistorsTransistors Circa 1947
First Junction Transistor, 1951
http://www.pbs.org/transistor/science/info/junctw.html
First Integrated circuits 1958-591958-1959 Integrated Circuits
IC Technology
Technology and Economics in the Semiconductor Industry, G.D. Hutcheson and J.D. Hutcheson, Scientific American, October 1997
WaferSilicon Wafer
Wafer sizes
Tom Garrett, Intel Program Briefing, May 11, 2001,http://www.intel.com/research/silicon/D1C.pdf
IC Technology, Intel
Robert Chau, Gerald Marcyk, http://www.intel.com/research/silicon/20nmpressfoils.pdf
Intel Processor complexity
Gerald Marcyk, http://www.intel.com/research/silicon/MarcykIDF0801.pdf
Chip complexity
Technology and Economics in the Semiconductor Industry, G.D. Hutcheson and J.D. Hutcheson, Scientific American, October 1997
IC Production Technology, 2007Demonstrated 2001
20 nm, 1.5 THz transistors demonstrated 2001Will enable 1 billion transistor chips operating at 20 GHz 2007
Robert Chau, Gerald Marcyk, http://www.intel.com/research/silicon/20nmpressfoils.pdf
Transistors, IntelProduction 2001 Production 2005
Production 2007
Paolo Gargini, http://www.intel.com/research/silicon/PaoloIDF022601.pdfRobert Chau, Gerald Marcyk, http://www.intel.com/research/silicon/20nmpressfoils.pdf
On-Chip Interconnection
Paolo Gargini, http://www.intel.com/research/silicon/PaoloIDF022601.pdf
Extreme Ultraviolet Lithography
Gary Stix, Scientific American, April 2001, http://scientificamerican.com/2001/0401issue/0401innovations.html
UV vs EUV lithography
Charles Gwyn, Intel, Cahners MDR Microprocessor Forum, October 11, 2000,http://www.intel.com/research/silicon/euv_mpf_101100.pdf
Lithography – EUV/DUV
Charles Gwyn, Intel, Cahners MDR Microprocessor Forum, October 11, 2000,http://www.intel.com/research/silicon/euv_mpf_101100.pdf
Transistor Sizes vs. DNA
Gerald Marcyk, Intel Developer Forum, Fall 2001,http://www.intel.com/research/silicon/MarcykIDF0801.pdf
PIV die
SIA Roadmap – DRAM Capacity
0
20
40
60
80
100
120
Gbi
ts
1999 180 nm
2000 2001 130 nm
2002 2003 .
2004 2005 .
2008 60 nm
2011 40 nm
2014 30 nm
DRAM: Gbits per Chip; High Volume & High Density
DRAM Gbits/chip atproduction
DRAM Gbits/chip atintroduction
SIA Roadmap – DRAM Cell Sizes
0
0.05
0.1
0.15
0.2
0.25
0.3
Cel
l siz
e m
icro
n2
1999 180 nm
2000 2001 130 nm
2002 2003 .
2004 2005 .
2008 60 nm
2011 40 nm
2014 30 nm
DRAM: Cell factor x 10-3 and cell size micron2
DRAM Cell factor x 10-3
DRAM Cell areamicron2
SIA Roadmap – Chips sizes
Chip Sizes
0
100
200
300
400
500
600
700
800
1999 180 nm
2000 .
2001 130 nm
2002 .
2003 .
2004 90 nm
2005 .
2008 60 nm
2011 40 nm
2014 30 nm
mm
2
DRAM Chip Size at Introduction, mm2
DRAM Chip Size at Production, mm2
MPU, High-volume Microprocessor, Cost-Performance, at Introduction, mm2MPU, High-volume Microprocessor, Cost-Performance, at Production, mm2MPU, Low-volume Microprocessor, High-Performance, at Production, mm2ASIC, at Production, mm2
SIA Roadmap – Transistor density
0
5000
10000
15000
20000
25000
Mtra
nsis
tors
/cm
2
1999 180nm
2000 .
2001 130nm
2002 .
2003 .
2004 90 nm
2005 .
2008 60 nm
2011 40 nm
2014 30 nm
Transistor Density/cm2
DRAM, transistor density at introduction(Mtransistors/cm2)DRAM, transistor density at production(Mtransistors/cm2)SRAM, transistor density at introduction(Mtransistors/cm2)Logic, transistor density at introduction(Mtransistors/cm2)MPU, Cost-Performance, transistor density atintroduction (Mtransistors/cm2)MPU, Cost-Performance, transistor density atproduction (Mtransistors/cm2)MPU, High-Performance, transistor density atproduction (Mtransistors/cm2)ASIC at production, (Mtransistors/cm2)
SIA Roadmap – Functions per chip
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
1999 180 nm
2000 2001 130 nm
2002 .
2003 .
2004 90 nm
2005 .
2008 60 nm
2011 40 nm
2014 30 nm
Functions per chip, Mtransistors
DRAM, at Production
DRAM, at Introduction
MPU, Cost-Performance, atProduction MPU, Cost-Performance, atIntroduction MPU, High-Performance, atProduction ASIC, at Production
SIA Roadmap – Clock rates
0.000
2.000
4.000
6.000
8.000
10.000
12.000
14.000
16.000
GH
z
1999 180nm
2000 .
2001 130nm
2002 .
2003 .
2004 90 nm
2005 .
2008 60 nm
2011 40 nm
2014 30 nm
Clock rates
On Chip local clock, High-Performance
On-Chip, across chip clock,High-Performance
On-Chip, across chip clock,High-Performance, ASIC
On-Chip, across chip clock,Cost-Performance
Chip to Board (off-chip)speed, High-Performance,Reduced-Width, MultiplexedbusChip to Board (off-chip)speed, High-Performance,Peripheral bus
SIA Roadmap - Costs
05
1015202530354045
$/M
tran
sist
or
1999 180nm
2001 130nm
2003 .
2005 100nm
2011 40 nm
Cost/Mtransistor
DRAM, cost x 100, at Introduction,$/MtransistorsDRAM, cost x 100, at Production,$/MtransistorsMPU, Cost-Performance, atIntroduction, $/MtransistorsMPU, Cost-Performance, atProduction, $/MtransistorsMPU, High Performance, atProduction, $/Mtransistors
Relative IC densities
• DRAM 1.00• SRAM 0.13• Logic 0.03
Semiconductor Roadmap - IBM
http://www.research.ibm.com/0.1um
Networking
ARPA Net, 1969
ARPA Net, 1980
Internet2 - Abilene
The Internet, 1999
Total number of Internet Hosts
http://www.isc.org/ds/hosts.html
Network Proliferation
Internet around the World
Interplanetary Internet
Interplanetary Internet
155 missions, and counting
Interplanetary Internet
HumanExploration
Earth Science Deep Space Science
In-depth Planetary Exploration
Progressive, planned deployment of reusable communications infrastructure
2001 2003 2005 2007 2009 2011
Capabilities
Fiberoptics
All optical networks
Capabilities - OpticsNetwork Junction of large fiber bundles
Assume:50 fibers,100 wavelengths,40 Gb/s per wavelengthSwitching or
RoutingNode
(50/2) * 100* 40 = 100 Tb/s in each direction
This node would require200 Tb/s input + 200 Tb/s output capacity.
The dominant traffic in the network is IPLAN, WAN, SAN(?) - Convergence
Pervasive computing
The world wide computer• Internet Provides Connectivity• Web Provides Hyperlinked File System• Distributed Storage Moving from SAN to NAS• Peer-to-Peer Computing Provides Vast CPU Power• Result--The Distributed Global Computer
– Storage everywhere– Scalable computing– Wireless Interfaces Greatly Outnumber PC Interfaces
“When the Network is as fast as the computer’s internal links, the machine disintegrates across the Net into a set of
special purpose appliances”-Gilder Technology Report June 2000
Peer-to-Peer Computing
Larry SmarrComputer Science
& Engineering
Exponential Growth inScale of PC Parallel Computers
1
10
100
1000
10000
100000
1000000
10000000
1992 1994 1996 1998 2000 2002 2004
NASA Beowulf (1GF)
Sandia ASCI Red (1TF)
SETI@home (10TF)
Megacomputers (PFs)
1,000,000x in Only Ten Years
Entrap.com’s PrimeNet Grew to a Teraflop in Only Two YearsThe Great Mersenne Prime (2P-1) Search (GIMPS) Found
the First Million Digit Prime www.entropia.comSustained Throughput Today of 1.3 TF (30,000 PCs) = 47 Cray T916s
Larry SmarrComputer Science
& Engineering
Peer-to-Peer Computing
Larry SmarrComputer Science
& Engineering
SETI@home was the Breakthrough toPC Internet Computing
• Running on 500,000 PCs, ~1000 CPU Years per Day– 428,438 CPU Years so far
• Sophisticated Data & Signal Processing Analysis• Distributes Datasets from Arecibo Radio Telescope
Next Step-Allen Telescope Array
Computing Platforms 2001 ⇒ 2030
• Personal Computers O[$1000]– 109 Flops/sec in 2001 ⇒ 1015 – 1017 Flops/sec by 2030
• Supercomputers O[$100,000,000]– 1013 Flops/sec in 2001 ⇒ 1018 – 1020 Flops/sec by 2030
• Number of Computers [global population ~1010]– SCs ⇒ 10-8 –10-6 per person ⇒ 102 – 104 systems– PCs ⇒ .1x – 10x per person ⇒ 109 – 1011 systems– Embedded ⇒ 10x – 105x per person ⇒ 1011 – 1015 systems– Nanocomputers ⇒ 0x – 1010 per person ⇒ 0 – 1020 systems
• Available Flops Planetwide by 2030– 1024 – 1030 Flops/sec [assuming classical models of computation]
Peer-to-Peer Computing
Pervasive Computing - Sun
Shahin Kahn
Pervasive Computing - Oxygen
http://oxygen.lcs.mit.edu
Pervasive Computing - InfoSphere
http://www.cc.gatech.edu/projects/infosphere/SBBDkeynote.pdf
Pervasive Computing - InfoSphere
http://www.cc.gatech.edu/projects/infosphere/SBBDkeynote.pdf
Pervasive Computing - InfoSphere
http://www.cc.gatech.edu/projects/infosphere/SBBDkeynote.pdf
Storage
Disk drive evolution - IBM
http://www.almaden.ibm.com/sst/
Disk storage roadmap – IBM
85 90 95 100 105 110 115
Availability Year
0.01
0.1
1
10
100
1000
10000A
real
Den
sity
, Gbi
ts/in
2IBM Advanced Storage Roadmap
Superparamagnetic Effect
1 Gbit/in2 Demo 3 Gbits/in2 Demo
5 Gbits/in2 Demo12.1 Gbits/in2
20.3 Gbits/in2 Deskstar 40GVUltrastar36LZX
Enhanced Magnetic Disk Drive
Advanced Storage Technology/Holography
2000 05 10 15
Microdrive 10K RPM Integrated Head/Suspension Giant MR Head/Pico Slider Ramp Load/Unload, Glass Substrates No-ID MR Head/Nano-slider PRML Data Channel Thin Film/High Coercivity DisksSmall Form Factor
Lab Demos3.5 Inch FF2.5 Inch FF>10 Inch FF
ED GROCHOWSKI at ALMADEN
advr
dmp2
0a.p
rz
35.3 Gbits/in2 Travelstar30GT
http://www.almaden.ibm.com/sst/
Disk storage areal density - IBM
1970 1980 1990 2000 2010
Year
0.01
0.1
1
10
100
1000
10000
100000
Are
al D
ensi
ty, M
bits
/inch
2Areal Density of Magnetic HDD and DRAM
25% = 2X per 3 years40 260 1.5100 1
25% CGR
60% CGR
40% CGR
1M4M
16M
64M256M
1024M
3380E3380K
Travelstar 6GN
Ultrastar 36XPUltrastar 18XP
Ultrastar 2XPUltrastar XP
Corsair3390-2
Ultrastar 36ZX
Ed Grochowski at Almaden
AREA
L98E
.PR
Z
Travelstar 25GSUltrastar 36LZX
100% CGR
http://www.almaden.ibm.com/sst/
Disk storage - supermagnetism
http://www.almaden.ibm.com/st/projects/patternedmedia/
Disk storage – AFC technology
http://www.research.ibm.com/resources/news/20010518_pixie_dust.shtml
Disk storage - cost
IBM 9.1GB Ultra 2XP
1980 1985 1990 1995 2000 2005 2010Year
0.001
0.01
0.1
1
10
100
1000 P
rice/
MB
yte,
Dol
lars
HDD DRAM Flash Paper/FilmAverage Price of Storage
IBM 18.2GB Ultrastar
IBM Deskstar 37GB
Toshiba 6.4GB
IBM Deskstar4
IBM Deskstar3
IBM 16.8GB Deskstar
IBM 8.1GB Travelstar
Seagate 8.6GB
Quant 4.5GB
64MB
IBM 9.1GB Ultrastar
96 MB Flash Camera Mem.
64MB Flash4MB Flash
16MB Flash1MB Flash
512KB Flash256KB Flash
128KB Flash
8KB
32KB 64KB
128KB
512KB 1MB 2MB
4MB
IBM6150
Wren II Seagate ST125
Maxt170IBM0615
IBM0663
Seagate B'cuda4
Seagate ST500
oem
prc2
000a
a.pr
z
128MB Flash
64MB
Ed Grochowski at Almaden
128MB Flash
IBM 25GB Travelstar
IBM 340 MB Microdrive
IBM Deskstar 25GB
IBM Deskstar 75GXP
IBM 1 GB Microdrive
1" HDD ProjectionDataQuest 2000
Flash ProjectionDataQuest 2000
Range of Paper/Film
3.5 " HDD 2.5 " HDD
1 " HDD
Flash
DRAM
http://www.almaden.ibm.com/sst/
Disk drive data rates – IBM
90 95 100 105Availability Year
1
10
100
Max
imum
Inte
rnal
Dat
a R
ate,
MB
ytes
/s
90 95 2000 2005
Magnetic Hard Disk Drive Internal Data Rate IBM Products
2.5 inch Mobile Products
3.5 inch Server Products
40 % CGRUltrastar 18XPUltrastar 9ZX
Travelstar 10GTTravelstar 14GSUltrastar 2XP
Travelstar 6GTTravelstar 8GS
Travelstar 3GNTravelstar 5GS
Travelstar 4GTTravelstar VP
Travelstar 3XP
Travelstar 2XP
Ultrastar XP
Travelstar 2LPTravelstar XP
Travelstar Travelstar LP
Wakasa
SpitfireAllicat
Sawmill
Corsair 1,2
Ultrastar 18ZXUltrastar 36XP
1 GHz
100 MHz
10 MHz
Data Channel Performance
Ultrastar 36ZX
Ed Grochowski at Almaden
drtr
d200
0va.
prz
Travelstar 25GS
Ultrastar 36 LZX
Data rate = Linear x RPM x Disk Density Diam.
.
Deskstar 75GXP
Travelstar 32GH
http://www.almaden.ibm.com/sst/
Disk storage performance
1990 1995 2000 2005 2010Availability Year
1
10
100Ti
me,
mill
isec
onds
IBM HDD Access/Seek Time-Performance Increase
Accessing
Seeking Rotating
Ultrastar 36ZX
Ultrastar 36XPUltrastar 18ZX
Ultrastar 18XPUltrastar 2XPUltrastar XP
Ultrastar 9ZX
IBM Advanced Technology
seek time ~(inertia/power)1/3 x (data band)2/3
rotational time ~ (RPM) -1
(latency)
SEEK
1999
A.P
RZ
Ultrastar 36LZX
http://www.almaden.ibm.com/sst/
Visualization
Display Walls
Display Walls
Virtual DisplaysPortable Virtual Projection Display, BSAC, UCB
Liquid CrystalDisplay
Virtual Retina Display
http://www.mvis.com/techprod.htm
E-Ink – Electronic Paper
http://www.eink.com/technology/index.htm
Gyricon – Electronic Paper
Technology on the horizon
Smart Dust - UCBRF Mote
Sensor Laser Mote with CCD
Laser MoteRF Mini Mote I
RF Mini Mote II
IrDA Mote
http://robotics.eecs.berkeley.edu/~pister/SmartDust/
MEMS – Pressure Sensor
http://www.darpa.mil/mto/mems/presentations/memsatdarpa3.pdf
MEMS - Sonoelectronics
http://www.darpa.mil/mto/mems/presentations/memsatdarpa3.pdf
As Our Bodies Move On-Line• New Sensors—Israeli Video Pill
– Battery, Light, & Video Camera– Images Stored on Hip Mounted Device
• Next Step—Putting You On-Line!– Key Metabolic and Physical Variables– Wireless Internet Transmission
• Post-Genomic Individualized Medicine– Combine Your Genetic Code & Imaging,
with Your Body’s Data Flow – Use Powerful Data Mining Techniques
www.givenimaging.com
MEMS - Biosensors
http://www.darpa.mil/mto/mems/presentations/memsatdarpa3.pdf
MEMS - Picosatellites
http://www.darpa.mil/mto/mems/presentations/memsatdarpa3.pdf
MEMS - Picosatellites
http://www.darpa.mil/mto/mems/presentations/memsatdarpa3.pdf
Carbon nanotubes - IBM
http://www.research.com/resources/news/20010425_Carbon_Nantubes.shtml
Molecular Computing
Mark A. Reed and James M. Tour, Scientific American, June 2000
Molecular computing
A micron resolutioncircuit pattern
Polymer transistorson plastic substrate
Philips ResearchEindhoven
http://www.research.philips.com/pressmedia/pictures/polelec.html
An inorganic semi-conductor droplet processed at plastic compatible temperatures
Polymertransistorson plastic substrate
http://www.media.mit.edu/nanomedia/projects.htmlhttp://www.bell-labs.com/org/physicalsciences/projects/plastic/plastic.html
Molecular Transistor
Mark A. Reed and James M. Tour, Scientific American, June 2000
Polymer Radio Frequency Identification Transponder
http://www.research.philips.com/pressmedia/pictures/polelec.html
Bell Labs – MEMS Switch
http://www.bell-labs.com/org/physicalsciences/projects/mems/mems.html
Bell Labs – MEMS Switch
http://www.bell-labs.com/org/physicalsciences/projects/mems/mems.html
Holographic Storage - InPhase
http://www.inphase-technologies.com/
Portable Devices
Applications
E-Science: Data Gathering, Analysis, Simulation, and
Collaboration
LHC
CMS
Simulated Higgs Decay
Applications Drivers [circa 2001]• Today: data intensive science
communities– High energy physics– Astronomy– Earth science
• Emerging: human intensive science communities– Engineering– Biology– Nanoscience– Exploration
Tele-Science
Whole Internet Simulation
• ~106 – 108 routers• ~109 – 1011 devices
• Modeling of routing and dynamic resource management
• Determining the response of the internet [and other networks] to accidental and intentional disruptions
• Creating an automated market for bandwidth, network CPU cycles, lambdas and network storage, etc
Electron microscopy
500 Å
400 kV electron image of herpesvirus capsids
Electron microscopy8.5 Å Structure of the HSV-1 Capsid
• 12 Pentons (VP5)• 150 Hexons
(VP5, VP26)• 320 Triplexes
(1 VP19C, 2VP23)
Diameter 1250Å
Total no. proteins 2820Capsid shell 192 MDaAsymmetric unit 3.2 MDa
Biomedical imaging
MRI Nuclear Medicine
Optical TomographyX-ray ComputedTomography
Biomedical simulation
U.S. Department of EnergyPacific Northwest National Laboratory
11/25/00 27
Virtual Lung from PNNL's Virtual Biology Center
NWGrid & NWPhysdesigned to simulate coupledfluid dynamics and continuummechanics in complexgeometries using 3-D, hybrid,adaptive, unstructured grids.
NWGrid - grid generation &setup toolbox
NWPhys - collection ofcomputational physics solvers
Harold Trease, PNNL
In 1988 … NEXRAD Was Becoming a Reality
s fn
The Impact of Horizontal Resolution
Radar CAPS 6-hour LocalForecast
s fn
Wildfire simulation
Scheduling of large systems
Continental Airlines
Simulation of complex mechanical systems
Analysis of large data collections
les robertson - cern-it-09-00 10 last update 11/25/00 5:15CERN
The LHC DetectorsCMS
ATLAS
LHCbRaw recording rate 0.1 – 1 GB/sec
3.5 PetaBytes / year~108 events/year
Grid – Prediction Started at the Large Scale…
This ThunderstormFalls Through the Cracks
736 kmgridspacing
19502000
A Dynamic Environment
10 km
3 km1 km
20 km CONUS Ensembles
10 km
3 km
3 km
3 km
3 km
10 km
20 km CONUS Ensemblest=0
t=6 hrs
Advanced Regional Prediction System (ARPS)
ARPS Data Analysis System (ADAS)
ARPS Numerical Model– Multi-scale non-hydrostatic prediction model with comprehensive physics
– Plots and images – Animations – Diagnostics and statistics – Forecast evaluation
– Ingest – Quality control – Objective analysis – Archival
Single-Doppler Velocity Retrieval (SDVR)
4-D Vari ational
Data Assimilation
Variational Vel -ocity Adjustment
& Thermo-dynamic Retrieval
ARPS Data Assimilation System (ARPSDAS)
ARPSP LT and ARPSV IEW
Inco
min
g da
ta
Oklahoma MesonetWSR-88D Wideband
ASOS/AWOS
SAO
ACARSCLASS
Mobile Mesonet
Profilers
Rawinsondes
Satellite
Lateral boundary conditions from large-scale models
Gridded first guessData Acquisition
& AnalysisData Acquisition
& Analysis
Forecast GenerationForecast Generation
Parameter Retrieval and 4DDAParameter Retrieval and 4DDA
Product Generation and Data Support System
Product Generation and Data Support System
6 pm 7 pm 8 pmFo
reca
stR
adar
2 hr 3 hr 4 hr
Weather Service Radar Sites
Observations: A Dynamic EnvironmentObservations: A Dynamic EnvironmentNEXRAD Data via Internet/AbileneNEXRAD Data via Internet/Abilene
As Weather Develops, QOS Must AdaptAs Weather Develops, QOS Must Adapt
Evolution of Computer Power/Costs