|
LAVA: Leakage Avoidance and Analysis
LAVA provides a comprehensive environment
for modeling and optimization of chip leakage. Leakage management
already is -- and will continue to be -- one of the most critical
needs for chip design.
Leakage is power consumed
by the circuit when it is off, and this problem has emerged as
the most critical design challenge for current and future integrated
circuits because it limits the frequency, yield and power of most
of IBM’s leading designs. Traditionally, Complementary Metal-Oxide
Semiconductor (CMOS) circuits have had little or no leakage power,
but due to continued scaling of device dimensions, leakage power
is now increasing exponentially with each technology generation
(see Figure 1). It is expected to be the dominant part of total
power by 65nm process technology node. To accurately analyze and
reduce leakage power, IBM Research has initiated a comprehensive
program -- Leakage Avoidance and Analysis (LAVA).

Figure 1. Leakage power (sub-threshold and
gate) is increasing at an exponential rate and is growing much
faster than dynamic (or switching) power.
Leakage is dependent on several design variables
(see Figure 2) and has two key components: sub-threshold and gate
leakage. Sub-threshold leakage is dependent on process variations,
supply voltage and temperature, while gate leakage is dependent
on gate oxide thickness. Both gate and sub-threshold leakage are
dependent on circuit topologies and logic of the circuit. LAVA
accurately analyzes leakage by modeling within-chip and chip-to-chip
process variations, probability-driven static leakage estimation,
and temperature and power supply co-analysis. LAVA is organized
into three main components: LAVA-Chip for chip-level leakage estimation
and planning; LAVA-macro for unit and macro-level analysis and
optimization; and LAVA-mod for leakage characterization and modeling
of standard cells and design building blocks.

Figure 2. Leakage is dependent on circuit topology, operating
environment, input patterns and process variations.
Figure 3. Overview of the different components
of Leakage Avoidance and Analysis (LAVA).
Reducing leakage
has emerged as a key objective in current and future chip designs.
Several leakage minimization techniques have been developed in
LAVA to reduce the active leakage of the circuit. For instance,
LAVA performs simultaneous Vt-assignment and sizing to reduce
both the gate and sub-threshold leakage. Leakage can also be reduced
by assigning the circuit to a low leakage state by controlling
the input and latch outputs. While leakage varies exponentially
across the process spread, adaptive circuit design is critical
in managing leakage in sub-90nm design processes. And, adaptive
control of power supply voltage and threshold voltage significantly
reduce leakage, while also improving circuit performance and parametric
yield.

Figure 4. Leakage analysis for an ASIC chip
in LAVA, considering power supply and temperature variations across
the chip.

Figure 5. Sample screen shots of the graphical
interface to LAVA tools.
Anirudh Devgan, D. Blaauw,
F. Najm and S. Narendra, “Leakage
Trends, Estimation and Avoidance,’ Tutorial, IEEE International
Conference on Computer Aided Design (ICCAD), Nov 2003.
R. Rao, Anirudh Devgan, D. Blaauw,
D. Sylvester, “Paramateric Yield
Estimation considering Leakage Variability,” ACM/IEEE
Design Automation Conference, June 2004.
H. Su, F. Liu, Anirudh Devgan, E.
Acar, S. Nassif, “Full-chip Leakage
Estimation considering Power Supply and Temperature Variations,”
In Proceedings of IEEE International Symposium on Low Power Electronics
and Design, 2003.
E. Acar, Anirudh Devgan, S. Nassif,
H. Su, F. Liu, “Leakage
and Leakage Sensitivity Computation for Combinational Circuits,”
In Proceedings of IEEE International Symposium on Low Power Electronics
and Design, 2003.
R. Rao, J. Burns, Anirudh Devgan,
R. Brown, “Efficient Techniques
for Gate Leakage Computation,” In Proceedings of IEEE
International Symposium on Low Power Electronics and Design, 2003.
Anirudh Devgan and Chandra Kashyap,
“Block-based Static Timing Analysis
with Uncertainty,” In proceedings of IEEE/ACM International
Conference on Computer Aided Design, 2003.
Copyright © (2003) by IEEE. Permission
to make digital or hard copies of part or all of this work for personal
or classroom use is granted without fee provided that copies are
not made or distributed for profit. To copy otherwise, to republish,
to post on servers, or to redistribute to lists, requires prior
specific permission and/or a fee.
Anirudh Devgan received the
IEEE William
J McCalla award in 2003.
IBM Microelectronics Division One
Team award in 2001.
IBM Corporate award in 2000.
IBM Outstanding Research Accomplishment
and IBM Outstanding Innovation awards in 1999.
|
 |
 |
|
Anirudh
Devgan
Researcher
|
What is the most exciting potential
future use for the work you're doing?
Leakage and power are the
most critical problems that the semiconductor industry is
currently facing. If one can effectively control leakage
and total power of our chips, we can truly extend the semiconductor
roadmap and Moore's Law for several process technology generations.
What is the most interesting part
of your research?
Design automation is a combination of computer
science (in term of new, fundamental algorithms) and engineering
physics (in terms of new process technology effects). I
really enjoy tackling exciting, challenging and relevant
problems posed by this merger.
What inspired you to go into this
field?
The most interesting part of research for me is
to identify critical new problems, to find novel and appropriate
solutions and to implement them to be useful in a practical
setting. With LAVA, the leakage problem is so vast that
it requires knowledge of various different fields. It has
been fun interacting with various parts of IBM (process
technology, characterization, design, design automation,
test, etc).
What is your favorite invention
of all time?
Written language.
|
| Research Team |
 |
 |
|
Emrah Acar |
Anirudh Devgan
|
Sani Nassif |
|
|