L16L17

advertisement
Revisit CMOS Power Dissipation
• Digital inverter:
2
P  CLVDD
 f  I leakVDD  PSC
– Active (dynamic) power
– Leakage power
– Short-circuit power (ignored)
P
N
VDD
CL
Roy & Prasad (2000)
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
1
Leakage vs. Active Power Trends
W. Haensch, IBM J. Res. Dev. 50, 339 (2006)
I leak
 VGS  Vth
W
2

eff COX  m  1VT exp 
Leff
 mVT
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips

 VDS
 1  exp  

 VT



2
Some Observations with Leakage
I leak
 VGS  Vth
W
2

eff COX  m  1VT exp 
Leff
 mVT

 VDS
 1  exp  

 VT



• This is the “usual” (BSIM, Spice) leakage model
• The thermal voltage VT = kBT/q
• This model was derived for 3-dimensional carrier motion,
impinging on a small energy barrier (what about 1-D or
2-D transistors?)
• This model assumes some average “junction
temperature” T but T itself is unsteady during digital
operation! (what about hot phonons?!)
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
3
What About Energy?
• Energy is a better metric when worried about battery life
• So look at energy, not power minimization:
• Critical difference: leakage energy depends on circuit
delay, tp
VDD
1 CLVDD
?
Delay  

1.3
f
I Dsat
(VDD  Vth )
2
Power  fVDD
© 2010 Eric Pop, UIUC
2
Energy  VDD
ECE 598EP: Hot Chips
4
Effects of Lowering VDD
B. Zhai, IEEE Trans. VLSI Sys. 13, 1239 (2005)
• Easy observation: lowering VDD lowers power and
energy… the latter up to a point!
• How low VDD?
• It is theoretically possible to operate circuits near VDD ~
50 mV, deep into the subthreshold regime!
• So… why not do it?
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
5
Energy-Voltage Trade-Off
B. Zhai, IEEE Trans. VLSI Sys. 13, 1239 (2005)
• Remember, delay:
CLVDD
tp 
I ON
• At high VDD  ION = ID,sat
• At low VDD delay too high,
so leakage energy goes up
as well
2
P  CLVDD
 f  I leakVDD
2
E  CLVDD
  IleakVDDt p
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
Optimum
VDD!
6
Principles of Low-Power Design
Roy & Prasad (2000)
• Use the lowest possible supply voltage (VDD)
• Use the smallest geometry, highest frequency devices
BUT operate them at the lowest possible frequency (f)
• Use parallelism and pipelining to lower required
frequency of operation
• Manage power by disconnecting power source when
system is idle (sleep states)
• Design systems to have lowest requirements of
performance for the given user functionality
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
7
Leakage Model: Closer Look
I leak
 VGS  Vth
W
2

eff COX  m  1VT exp 
Leff
 mVT

 VDS
 1  exp  

 VT



• Strongly (exponentially!) temperature dependent!
• Typically people use ΔT = PRTH where
– ΔT is an average “junction temperature”
– P is a time-averaged power dissipation (active + leakage)
• How do we calculate RTH?
• And when is it OK to use it?
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
8
Device Thermal Resistance Data
Single-wall
nanotube SWNT
100000
High thermal resistances:
• SWNT due to small thermal
conductance (very small d ~ 2 nm)
RTH (K/mW)
10000
1000
100
• Others due to low thermal
conductivity, decreasing dimensions,
increased role of interfaces
GST
Phase-change
Memory (PCM)
Silicon-onInsulator FET
SiO2
10
Cu
Cu Via
Power input also matters:
1
0.1
0.01
• SWNT ~ 0.01-0.1 mW
Si
0.1
Bulk FET
L (m)
1
• Others ~ 0.1-1 mW
10
Data: Mautry (1990), Bunyan (1992), Su (1994), Lee (1995), Jenkins (1995), Tenbroek (1996),
Jin (2001), Reyboz (2004), Javey (2004), Seidel (2004), Pop (2004-6), Maune (2006).
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
9
Modeling Device Thermal Response
100000
• Steady-state models
RTH (K/mW)
10000
– Lumped: Mautry (1990), Goodson-Su
(1994-5), Pop (2004), Darwish (2005)
1000
100
SOI FET
10
1
– Finite-Element
0.1
0.01
Bulk FET
0.1
L (m)
1
10
D
L
tSi
W
tBOX
Bulk Si FET
RTH
1
1


2kSi D 4kSi LW
© 2010 Eric Pop, UIUC
SOI FET
RTH
1

2W
1/ 2
 t BOX 


 k BOX kSi tSi 
ECE 598EP: Hot Chips
10
Modeling Device Thermal Response
• Transient Models
– Lumped: Tenbroek (1997), Rinaldi
(2001), Lin (2004)
– Introduce CTH usually with approximate
Green’s functions; heated volume is a
function of time (Joy, 1970)
Instantaneous T rise
T 
E Pt

C  cV
Due to very sharp heating
pulse t ‹‹ V2/3/
– Finite-Element
More general
Simplest (~ bulk Si FET)
P
 r 
T (r , t ) 
erfc 

2 kSi r
 2 t 
t
 r 2 
1
P(t ')
T (r , t ) 
exp 
 dV ' dt '
3/ 2 
3/ 2
8 cV ( ) 0 (t  t ')
 4 (t  t ') 
Temperature evolution anywhere (r,t) due to arbitrary heating
function P(0<t’<t) inside volume V (dV’  V) (Joy 1970)
Temperature evolution of a step-heated point
source into silicon half-plane (Mautry 1990)
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
11
Approaches for Thermal Resistance
Interconnect
• Time scale:
– Transient
– Steady-State
• Geometric complexity:
– Lumped element (shape factors)
Via + Interconnect
– Analytic
– Finite element (Fourier law)
D
L
tSi
W
tBOX
Bulk Si FET
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
SOI FET
12
Shape Factors
Sunderland, ASHRAE (1964), many others
• Heat flux: q = Sk(T1-T0)
• Equivalent thermal resistance
RTH = 1/Sk
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
13
Ex: Heat Loss from Via + Interconnect
Chen, Li, Rosenbaum, Kang, IEEE TCAD ICS 19, 197 (2000)
Cu
Estimating heat loss (thermal
resistance) “looking into” one Cu line:
SiO2
2r
zTOP
Th,eff 
AVia
ACu
AVia   r 2
w
d
hpeff
1
kCu ACu (hpeff )
ACu  wd

z 

 1.86kox log10 1   
 w 

0.66
 w
 
d
0.1
Chen, 2000
zBOT
Typical values
Th ,eff  25  32 K/mW (bot – top)
Si
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
14
Many Shape Factors (Compact Models)
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
15
Thermal-Electrical Cheat Sheet
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
16
Obtaining the Temperature Distribution
• Now we want
temperature
distribution T(x) in 1-D
• Consider power in/out
of a 1-D element
• Simplest case: Si layer
on SiO2/Si substrate
(SOI)
• Or interconnect on
thermally insulating
SiO2
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
17
1-D Interconnect with Heat Generation
L
x
x+dx
d
W
Heat:
Electrical:
tox
SiO2
Si
dT
dx
dV
I  AJ  A  F   A
dx
Q   Ak
Energy balance equation for
1-D element “dx”: pick units of
J/cm3 or W/cm3 (W = J/s)
T0
Energy In (here, Joule heat) = Energy Out (left, right, bottom) + Change in Internal Energy
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
18
Ex: 1D Rectangular Nanowire
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
19
1-Dimensional Heat Equation
unsteady (transient)
T
k
(k T )  Q '''  CV
T
t
SiO2
T
hp
(k T )  Q ''' (T  T0 )  0
A
g
steady, with convection
SiO2
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
20
Interconnect Heat Loss and Crosstalk
© 2010 Eric Pop, UIUC
ECE 598EP: Hot Chips
21
Carbon Nanotube (Cylinder)
L
T
Pt
g
(a)
SiO2
Si
d
A(kT )  p' g (T  T0 )  0
T ( x)  T0 
tOX
p'  I2
tSI
Role of cylindrical heat
spreading (shape factor!)
(b)
g ox 
p' 
cosh( x / LH ) 
1



g  cosh( L / 2 LH ) 
dR
h 1
 I2 2
dx
4q eff
LH 
kA
g
 kox
 8t 
ln  ox 
 d 
900
Role of thermal
contact resistance
Tmax
T (K)
700
500
E. Pop et al.
J. Appl. Phys. 101, 093710 (2007)
© 2010 Eric Pop, UIUC
ΔTC
kA
300
-1
ECE 598EP: Hot Chips
0
X ( m)
1
2
dT
dx

C
TC
R C ,Th
22
Download