Thermal Control of Electronics:
Perspectives and Prospects
Dr. Robert Hannemann
Rohsenow Symposium on Future
Trends in Heat Transfer
Massachusetts Institute of Technology
16 May 2003
Introduction
„
Thermal control of electronics is posing significant challenges
(again)
„
New, or re-invented technologies will be needed
„
Moore’s Law is the iconic driver of electronics progress – but it will
continue until physics gets in the way (cooling!)
„
Research and development is needed as never before
„
Note: what follows is computer-centric, but…
Rohsenow Symposium
5/16/03
2
Perspective on microelectronic heat flux
Heat Flux (W/cm2)
Current Chips
T, K
Hannemann, Bar-Cohen, and Oktay, ca. 1986
Rohsenow Symposium
5/16/03
3
Technology generations
Timeframe
Generation
Representative Product
1945-1955
Historic
Specialty Computers
1955 - 1965
Transistor
Early Mainframe
1965 - 1975
SSI
Mainframe
1975 - 1985
MSI
Minicomputer
1985 - 1990
LSI
Microcomputer
1990 - 2000
VLSI
PC, Notebook, Portables
2000 -
ULSI
Micro-based Server
Trouble
Rohsenow Symposium
Chip Power
Design care
5/16/03
Module Power
Density
Rack Power
Ignore
4
What’s old is new again…
Heroic measures to maintain air
cooling…
Heat pipe concepts…
Two-phase system cooling…
Rohsenow Symposium
5/16/03
5
Design / technology drivers
Performance
Size
Rohsenow Symposium
5/16/03
Cost
Reliability
6
Cost of cooling
Cost of cooling a microprocessor, Intel
Rohsenow Symposium
5/16/03
7
Intel’s inflection
X86 power dissipation
100
90
80
Pentium 4
Chip Power (W)
70
60
50
40
30
Pentium 3
20
10
Pentium
486
0
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Year of Introduction
Rohsenow Symposium
5/16/03
8
Chip power trends
High Performance Chip Power (W)
250
200
150
W
SIA 1993
NTRS 1999
NTRS 2002
100
50
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Year of Release
Rohsenow Symposium
5/16/03
9
Heat flux trends at chip level
Chip Heat Flux
80
70
60
W/cm2
50
SIA 1993
40
NTRS 1999
NTRS 2002
30
20
10
0
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year of Introduction
Rohsenow Symposium
5/16/03
10
A new problem: machine-room heat density
Data from
Amdahl,
Compaq,
Dell, HP,
IBM, SGI,
Sun,
Unisys,
Lucent,
Nortel, and
Cisco
Rohsenow Symposium
5/16/03
11
Key challenges: large systems
„
Solutions for very high chip powers (100 – 200W)
„
High reliability / availability
„
Rack-level cooling
„
Machine-room cooling
„
Cost management
Rohsenow Symposium
5/16/03
12
Key challenges: office systems
„
Cost / performance microprocessors will reach 80 – 100W within 2
years
„
Air cooling must be optimized
„
Acoustics
„
Reduced cost
Rohsenow Symposium
5/16/03
13
Other challenges
„
Telecom systems: central offices already stretching power limits
„
Photonic components provide very serious cooling challenge
„
„
Thermal control at heat fluxes ~ 2 x 103 W/cm2
„
Performance very sensitive to temperature
Harsh environments
„
Automotive
„
Telecom
„
Military
Rohsenow Symposium
5/16/03
14
Research areas
„
Materials
„
„
„
„
Devices
„
„
„
„
Design
„
„
„
Rohsenow Symposium
5/16/03
Optimized fluids for liquid cooling
Materials for package construction / thermal spreading
Phase change materials for transient applications
High static pressure, low acoustic noise blowers
Micro heat pipe structures
Two-phase cooling approaches
MEMS components for liquid / two phase cooling
Small-footprint liquid cooling systems
Package-scale jet impingement / spray cooling
Advanced, integrated thermal design tools
Frame and rack coolers
Equipment room thermal design
15
Conclusion
„
The importance of thermal management in electronics devices and
systems has waxed and waned over the past 50 years
„
Current technologies and applications are once again providing a
serious challenge – perhaps show-stopping – to heat transfer
engineers
„
Breakthroughs are needed in advanced cooling technologies (and
pragmatic design!) at all levels
Rohsenow Symposium
5/16/03
16