Thermal challenges in IC’s: Hot spots — passive cooling ...meptec.org/Resources/04 Honeywell...
Transcript of Thermal challenges in IC’s: Hot spots — passive cooling ...meptec.org/Resources/04 Honeywell...
Thermal challenges in IC’s: Hot spots — passive cooling and beyond
Devesh Mathur, Ph.D.Andy Delano, Ph.D.
Honeywell Electronic MaterialsSpokane, WA, USAFebruary 15, 2007
2 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Overview
• Motivation
• Where are the thermally challenged IC’s?
• Cooling IC’s today
• Cooling IC’s in the future
• How do we get there?
• Conclusions
3 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Microprocessor Applications
IC’s are pervasive in various applications
4 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
All Rely on Chips
10 billion square inches of silicon forecast for 2009
5 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
Putting the Heat Flux Challenge into Perspective
Sunlight on a Tropical Beach
6 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
Putting the Heat Flux Challenge into Perspective
1 Megaton at 1 mile
Sunlight on a Tropical Beach
Shuttle Re-entry
7 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
Putting the Heat Flux Challenge into Perspective
Rocket Nozzle Throat
1 Megaton at 1 mile
Sunlight on a Tropical Beach
Shuttle Re-entry
8 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
0.01
0.1
1
10
100
1000
10000
0 500 1000 1500 2000 2500 3000Temperature (oC)
Hea
t Flu
x (W
/cm
2 )
Putting the Heat Flux Challenge into Perspective
Rocket Nozzle Throat
1 Megaton at 1 mile
Sunlight on a Tropical Beach
Shuttle Re-entry
Tomorrow’s Microprocessors
The thermal challenge in an IC is a truly daunting problem
9 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Most heat leaves through center 15% of the spreader’s surface area
Motivation: CPU Heat Flux is High
IR image of an actual processor
10 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
ITRS Power Predictions
0
500
1000
1500
2000
2500
3000
2002 2004 2006 2008 2010 2012 2014 2016 2018Year
W o
r W/c
m2
Total Power (W) Heat Flux (W/cm^2)
130 nm 90 nm 65 nm 45 nm 32 nm
Motivation: CPU Power is Steadily Increasing
Flux at the core continues to rise despite design innovation
11 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Central Processor Unit(s)
ChipsetWhere are the thermally challenged IC’s?
AMB Memory
Graphics Processor Unit (GPU)Picture of a Workstation
Thermal solutions are required for the CPU and ASIC’s
12 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Memory: AMB IC
• IC: Bare die• Power: 4-8 Watts• Airflow: System dependent,
usually compromised- Sometimes cards are
perpendicular to flow- In servers and workstations,
many cards are packed densely together
• Thermal Solution: TIM and a heat spreader
AMB IC is HOT
Up to16 Memory Cards
Memory cards require thermal solutions and system level attention
13 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
The Most Critical Heat Pathways
Heat Spreader
Thermal Interface Material
The highest heat fluxes travel through 0.20 C/W of the total budget
Thermal Interface Material
Typical Thermal Budget
TIM 1, 0.050
Heat Spreader, 0.100
TIM2, 0.050
Heat Sink Base, 0.100
Heat Sink Fins, 0.100
Pre Heat, 0.100
0.00
0.10
0.20
0.30
0.40
0.50
0.60
θ (˚
C/W
)
{
14 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Classical Packaging Limits
• Classical Package consists of:- TIM1- Heat Spreader- TIM2
• Typical limitations- TIM1
Heat Spreader FlatnessHeat Spreader Surface
- Heat SpreaderSpreading ResistanceCTE mismatch
- TIM2Heat Spreader FlatnessHeat Spreader Surface
θ junction to heat sinkwhere we are vs. best in class
0.20
0.06
0
0.05
0.1
0.15
0.2
0.25
Typical Modeled Best in Class
˚C/W
Room to improve classical package with improved technology
15 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Materials Roadmap for Critical Heat Pathways
Heat Spreader
Thermal Interface Material
HEM produces and develops products for critical heat pathwaysThe products control at least 0.20 C/W of the total budget
Thermal Interface Material
-2004------------------------ 2006---------------------------2008+
16 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Memory Module Heat Sink: Thermal Modeling
Actual Memory Heat Sink
3D CAD Model for Mechanical FEA Analysis
Is TIM1 important to the AMB?
17 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
FEA results predict >10 ˚C reduction on AMB chip, allowing superior performance
Honeywell PCM45 Competitor’s MaterialTIM 1 Performance: Memory Module
18 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Honeywell’s TIM’s*
HEM representative products cover a wide performance range to help address thermal challenges
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0 1 2 3 4 5 6 7 8 9 10
Bondline Thickness (BLT), mil
Ther
mal
Impe
denc
e (C
-cm
2/W
)
PCM45F- PadPCM45F-SP PCM- Next Generation
APS Indium- Combo
(Polymer Filled) (Polymer Filled)
(Advanced Polymer Solder) (Pb Free- Metallic)
*Actual Data may vary in application. ASTM D-5470
19 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Collaboration is key to Innovation
IndustryAcademia
People
External Forces Competition
Rapid and fundamental innovation requires increased collaboration
20 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
• Products - Optimized particle TIMS- CNT TIMS- Advanced Heat Spreaders
Collaboration: example CTRC• Path
- Attend bi-annual CTRC meetings to review work
- Determine and support relevant technologies
- Productize
• Purpose: Long term research and development in the area of high-performance heat removal from compact spaces
- NSF Industry/ University cooperative research center at Purdue University
• Progress- Member since 2002- Hired CTRC graduate
Boiling Research
Particulate TIM ModelsSource: CTRC, Purdue. Used with permission.
Industry issues worked on fundamentally to address challenges
21 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Nextreme’s eTEC for Hotspot Cooling
Thermoelectrics: example Nextreme Active Cooling
• Thin film thermoelectric device operates between the heat spreader and IC
• Hot spots are addressed with refrigeration- Heat flux into heat spreader
increases, however, hot spot temperature is reduced
- Some additional power is required
• Challenges- Improving TE materials- Understanding Reliability
Source: Nextreme, used with permission.
With improved materials and reliability, thermoelectric devices could revolutionize hot spot cooling
22 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Improving the Heat Spreader Improves the performance of everything downstream
• Standard Heat Spreader• Tmax = 81.2 ˚C
• Enhanced Heat Spreader• Tmax = 63.4 ˚C
New: Honeywell’s Active Heat Spreader
Significant (17.8°C) reduction over current technology
23 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Other Innovations
• Power limitations for chip architects
• Software and threading innovation
• Data center power balancing
• Discrete device computing
• Multi die and multi core
• Feedback loops, throttling, and thermal shutdown
Smart chip & system design can also address thermal challenges
24 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Thermal Management at the IC Level
• This technique is currently used in:- mobile computers- CPU’s- AMB’s (DIMMs)
Power management can improve thermals, but will reduce performance
IC T
empe
ratu
re
Shutdown
Power rationing
Fan speed control
Permanent thermal damage
!!Computing errors!!
Reduced performance and reliability
In-spec performance and reliability
Low dB acoustics
25 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Summary
• Thermally Challenged IC’s found throughout modern computers- CPU- ASIC’s
GPUMemoryChipsets
• Overall Heat Flux and Hot Spots are pushing the limits of today’s products
• Some of today’s cutting-edge products address the thermal challenge today
• Next generation products offer hope. Collaboration is key.
Innovation in thermal management will enable better performance
26 Devesh Mathur & Andy Delano HONEYWELL – CONFIDENTIAL 2/15/2007 MEPTEC
Summary
• Thermally Challenged IC’s found throughout modern computers- CPU- ASIC’s
GPUMemoryChipsets
• Overall Heat Flux and Hot Spots are pushing the limits of today’s products
• Some of today’s cutting-edge products address the thermal challenge today
• Next generation products offer hope. Collaboration is key.
Innovation in thermal management will enable better performance