Transcript ZZ-HVS: Zig-Zag Horizontal and Vertical Sleep Transistor Sharing to

ZZ-HVS: Zig-Zag Horizontal and Vertical Sleep
Transistor Sharing to Reduce Leakage Power in
On-Chip SRAM Peripheral Circuits
Houman Homayoun Avesta Makhzan and Alex Veidenbaum
Dept. of Computer Science, UC Irvine
[email protected]
Outline







Cache Power Dissipation
Why Cache Peripheral ?
Proposed Circuit Technique to Reduce Leakage in
Cache Peripheral
Circuit Evaluation
Proposed Architecture to Control the Circuit
Results
Conclusion
On-chip Caches and Power

On-chip caches in high-performance
processors are large


more than 60% of chip budget
Dissipate significant portion of power via
leakage

Much of it was in the SRAM cells


Many architectural techniques proposed to
remedy this
Today, there is also significant leakage in the
peripheral circuits of an SRAM (cache)

In part because cell design has been optimized
Pentium M processor die photo
Courtesy of intel.com
Peripherals ?
Addr Input Global Drivers
Bitline
addr0
Global Wordline
addr1
Decoder
Bitline
Local Wordline
addr2
addr3
Predecoder and Global Wordline Drivers
addr
Sense amp
Global Output Drivers





Data Input/Output Driver
Address Input/Output Driver
Row Pre-decoder
Wordline Driver
Row Decoder
Others : sense-amp, bitline pre-charger, memory cells, decoder logic
Why Peripherals ?
100000
10000
1000
6300X
( pw )
100
200X
10
m
em
or
y
ce
ll
IN
V
IN X
V2
X
IN
V3
X
IN
V4
X
IN
V5
X
IN
V6
X
IN
V8
IN X
V1
2
IN X
V1
6
IN X
V2
0
IN X
V2
4
IN X
V3
2X
1

Using minimal sized transistor for area considerations in cells
and larger, faster and accordingly more leaky transistors to
satisfy timing requirements in peripherals.

Using high vt transistors in cells compared with typical threshold
voltage transistors in peripherals
Leakage Power Components of L2 Lache

SRAM peripheral circuits dissipate more than 90%
of the total leakage power
Circuit Techniques Address Leakage in SRAM Cell






Gated-Vdd, Gated-Vss
Voltage Scaling (DVFS)
ABB-MTCMOS
Forward Body Biasing (FBB), RBB
Sleepy Stack
Sleepy Keeper
Target SRAM memory cell
Architectural Techniques

Way Prediction, Way Caching, Phased Access


Drowsy Cache


Evict lines not used for a while, then power them down
Applying DVS, Gated Vdd, Gated Vss to memory cell


Keeps cache lines in low-power state, w/ data retention
Cache Decay


Predict or cache recently access ways, read tag first
Many architectural support to do that.
All target cache SRAM memory cell
Sleep Transistor Stacking Effect

Subthreshold current: inverse exponential function of
threshold voltage
VT  VT 0   (

( 2) F  VSB 
2 F )
Stacking transistor N with slpN:

The source to body voltage (VM ) of
transistor N increases, reduces its
subthreshold leakage current, when
both transistors are off
Drawback : rise time, fall time, wakeup
delay, area, dynamic power, instability
vdd
VC
Vgn
N
CL
VM
Vgslpn
vss
slpN
vss
Source of Subthreshold Leakage
in the Peripheral Circuitry
vdd
vdd
W
6
L
W
3
L
P4
I leakage P3
P2
P1
vdd
vdd
W
 24
L
W
 12
L
addr0
Bitline
Bitline
z
addr1
addr2
addr3
0
1
0
W
1.5
L
W
3
L
N1 I leakage N2
vss

vss
0
1
W
6
L
N3
vss
W
 12
L
N4
vss
The inverter chain has to drive a logic value 0 to the
pass transistors when a memory row is not selected

N1,N3 and P2,P4 are in the off state and are leaking
A Redundant Circuit Approach
Sleep signal
vdd
vdd
vdd
slpP2
slpP1
W
 12
L
slpP3
W
 12
L
W
 12
L
P3
1
0
slpP4
W
 12
L
P2
P1
vdd
P4
0
1
slpN5
N1
slpN1
Sleep signal
VM
W
6
L
vss
N2
slpN2
W
6
L
vss
N3
slpN3 W
L
W
1.5
L
N4
6
vss
slpN4
W
6
L
vss
Sleep signal
Drawback impact on wordline driver output
rise time, fall time and propagation delay
Impact on Rise Time and Fall Time

vdd
The rise time and fall time of the output
of an inverter is proportional to the
Rpeq * CL and Rneq * CL
slpP1
P2
P1
I leakage
1
0

vdd
slpP2
Inserting the sleep transistors increases
both Rneq and Rpeq
0
N1
slpN1
N2
slpN2
I leakage
vss
vss
Increasing in rise time
Impact on performance
Increasing in fall time
Impact on memory functionality
Fall Time Increase Impact
Bitline
Vdd
Wordline pulse
global wordline
driver
Bitline
1
0
M1
M2
local wordline driver
Sense-amp
Wordline pulse
generator
Fall time increase  pass transistor active period increase (read operation)
 The bitline over-discharge, the memory content over-charge during the
read operation.
 Such over-discharge



increases the dynamic power dissipation of bitlines
can cause cell content flip if the over-discharge period is large
The sense amplifier timing circuit and the wordline pulse generator
circuit need to be redesigned!
A Zig-Zag Circuit
Sleep signal
vdd
vdd
slpP4
slpP2
W
 12
L
vdd
1
0
P4
P3
P2
P1
W
 12
L
vdd
0
1
0
slpN5
N1
slpN1
Sleep signal
W
6
L
vss
N3
N2
vss
slpN3
W
6
L
W
1.5
L
N4
vss
vss
vss
Sleep signal

Rpeq for the first and third inverters and Rneq for the
second and fourth inverters doesn’t change.

Fall time of the circuit does not change
A Zig-Zag Share Circuit

To improve leakage reduction and area-efficiency of
the zig-zag scheme, using one set of sleep transistors
shared between multiple stages of inverters


Zig-Zag Horizontal Sharing
Zig-Zag Horizontal and Vertical Sharing
Zig-Zag Horizontal Sharing



Zz-hs less impact on rise time
Both reduce leakage almost the
same
vdd
vdd
Comparing zz-hs with zigzag
scheme, with the same area
overhead
Sleep signal
slpP
P2
P1
1
0
N1
vdd
P3
P4
N3
N4
0
N2
VM
Sleep signal
2 x slpN
vss
R Neq
I share
vss
vss
R N1
I share
R nslp  zz  hs 
vss
R nslp  zz
2
Zig-Zag Horizontal and Vertical Sharing
vdd
Sleep signal
vdd
vdd
Word-line
Driver line K
vdd
vdd
Word-line
Driver line K +1
slpP
P 11
P12
P 13
P14
P 21
P22
P 23
P24
N11
VM
N 12
N 13
N 14
N21
N 22
N 23
N 24
Sleep signal
slpN
vss
vss
vss
vss
vss
Leakage Reduction of Zig-Zag Horizontal
and Vertical Sharing
(a)
(b)
vdd
Vg0
Vg0
N 11
IN11
Vg0
vdd
VM1
I slpN
slpN
vss
Vg0
N 21
IN21
Vg0
vdd
VM1
I slpN
slpN
Vg0
N11
N21
IN 21
IN11
VM2
Vg0
vdd
I slpN
slpN
vss
Increase in virtual ground voltage
increase leakage reduction
vss
VM 1 
VM 2 
n. log
WN 1 1
Wslp N
10
 Vdd  Vg 0
2
n. log
2.WN 1 1
Wslp N
10
 Vdd  Vg 0
2
Circuit Evaluation

Test Experiment




Wordline inverter chain drives 256 one-bit memory cells.
Using Mentor Graphic IC-Station in TSMC 65nm
technology
Use Synopsis Hspice and the supply voltage of 1.08V at
typical corner (250 C)
The empirical results presented are for the





leakage current
rise time and fall time
propagation delay
dynamic power
area
50
4
45
3.9
40
3.8
35
3.7
30
3.6
25
3.5
20
3.4
15
3.3
10
3.2
5
3.1
0
3
e
li n
se
a
b
nt
da
n
u
red
g
za
zig
Leakage Power

hs
zz-
-1W
hs
zz-
-2W
dynamic power
Dynamic power increase of 1.5% to 3.5%
Max leakage reduction of 94%
(uW)
(nW)
Zig-zag Horizontal Sharing: Power Results
Zig-zag Horizontal Sharing: Latency Results
210
(ps)
190
170
150
130
110
90
70
50
li
se
ba
ne
u
red
nd
t
an
g
za
z ig
Propagation Delay


hs
z z-
-1W
Fall Time
-2W
hs
z z-
Rise Time
Both zig-zag and zig-zag share wordline driver fall time is not
affected
zz-hs-2W has the least impact on rise time and propagation
delay
Zig-zag Horizontal Sharing: Area Results
16
14
12
8
(

2
)
10
6
4
2
0
t
ine
an
l
d
e
n
s
u
ba
red

W
W
z ag
g
s -1
s -2
i
h
h
z
zz
zz
Area increase varies significantly from 25% for zz-hs-1W circuit to
115% for the redundant scheme
ZZ-HVS Evaluation : Power Result
1000
(a)
x100
log (nW)
100
x10
x12
10
x2
1
1
2
3
4
5
6
7
8
9
10
number of wordline row
baseline


redundant
zigzag
zz-hs
zz-hvs
Increasing the number of wordline rows share sleep transistors increases the
leakage reduction and reduces the area overhead
Leakage power reduction varies form a 10X to a 100X when 1 to 10 wordline
shares the same sleep transistors

2~10X more leakage reduction, compare to the zig-zag scheme
ZZ-HVS Evaluation : Area Result
160
(b)
140
120
80
(
 2)
100
60
40
20
0
1
2
3
baseline

4
5
6
7
number of wordline row
redundant
zigzag
8
zz-hs
9
10
zz-hvs
zz-hvs has the least impact on area, 4~25%
depends on the number of wordline rows shared
ZZ-HVS Circuit Evaluation: Sleep Transistor Sizing
Leakage power (nW)
Propagation delay (ps)


baseline W(1X) 2W(2X) 3W(3X) 4W(4X)
460
5.11
9.13
12.63
15.7
164
198
180
174
169
Trade-off between the leakage savings and impact
on the wordline driver propagation delay
zz-hvs-3W (3X) show an optimal trade-off
40X reduction in leakage at 5%
increase in propagation delay
Wakeup Latency

To benefit the most from the leakage savings of
stacking sleep transistors



keep the bias voltage of NMOS sleep transistor as low as
possible (and for PMOS as high as possible)
Drawback: impact on the wakeup latency of wordline
drivers
Wakeup latency associated with the zz-hvs-3W circuit
is 1.3ns


4 processor cycles (3.3 GHz)
For large memory, such as 2MB L2 cache the overall
wake up latency can be as high 6 to 10 cycles
Impact on Propagation Delay

The zz-hvs increases the propagation delay of the
peripheral circuit by 5%, when applied to wordline
drivers, input/output drivers, etc

Translate to 5% reduction in maximum operating clock
frequency of the memory in a single pipeline memory

Deep pipelined memories such as L1 and L2 cache hide
negligible increase in peripheral circuit latency
Sleep-Share: ZZ-HVS + Architectural Control

When an L2 cache miss occurs the processor executes
a number of miss-independent instructions and then
ends up stalling

The processor stays idle until the L2 cache miss is
serviced. This may take hundreds of cycle (300 cycles for
our processor architecture)

During such a stall period there is no access to L1 and
L2 caches and they can be put into low-power mode
Detecting Processor Idle Period

The instruction queue and functional units of the
processor monitored after an L2 miss


Instruction queue has not issued any instructions
Functional units have not executed any instructions for
K consecutive cycles (K=10)



The sleep signal is asserted
The sleep signal is de-asserted 10 cycles before the miss
service is completed
Assumption: memory access latency is deterministic.

No performance loss
Simulated Processor Architecture


Parameter
Value
L1 I-cache
L1 D-cache
L2 cache
Fetch, dispatch
Issue
Memory
Reorder buffer
Instruction queue
Register file
Load/store queue
Branch predictor
Arithmetic unit
Complex unit
Pipeline
128KB, 2 cycles
128KB, 2 cycles
2MB, 8 way, 20 cycles
4 wide
4 way out of order
300 cycles
96 entry
32 entry
128 integer and 125 floating point
32 entry
64KB entry g-share
4 integer, 4 floating point units
2 INT, 2 FP multiply/divide units
15 cycles
SimpleScalar 4.0
SPEC2K benchmarks


Compiled with the -O4 flag using the Compaq compiler targeting
the Alpha 21264 processor
fast–forwarded for 3 billion instructions, then fully simulated for 4
billion instructions

using the reference data sets.
L1 and L2 Leakage Power Reduction
60%
50%
40%
30%
20%
10%
vp
r
am
m
p
ap
pl
eq u
ua
k
fa e
ce
re
c
lu
ca
s
m
gr
id
sw
i
w
up m
w
i
av se
er
ag
e
m
p e cf
rl
bm
k
pa
rs
er
gc
c
ga
p
bz
ip
2
0%
L1 Leakage power reduction
L2 Leakage power reduction
Leakage reduction of 30% for the L2
cache and 28% for the L1 cache
Conclusion





Study break down of leakage in L2 cache components,
show peripheral circuit leaking considerably
proposed zig-zag share to reduce leakage in SRAM
memory peripheral circuits
zig-zag share reduces peripheral leakage by up to 40X
with only a small increase in memory area and delay
Propose Sleep-Share to control zig-zag share circuits
in L1 and L2 cache peripherals
Leakage reduction of 30% for the L2 cache and 28% for
the L1 cache