IEEE Solid-States Circuits Magazine - Fall 2020 - 122

semiconductors. By combining a reducVLSI Symposium in a virtual format
tion of the high-resolution ADC circuit
saved roughly 1,000 metric tons of
area that results from technology and
carbon from being released into the
design improvements from 180 to 40 nm
atmosphere.
with the trend of producing these conThe Future of Compute:
verters on 300-mm wafers rather than
How the Data Transformation
200-mm wafers, the energy input to
Is Reshaping VLSI
manufacture these devices is 5% of
Mike Mayberry, Intel, articulated
what it was 20 years ago.
a broad vision during his plenary
Lloyd also painted a vision of
talk, " The Future of Compute. " While
semiconductor technology becomMoore's law scaling will continue, he
ing not on ly mor e ef f icient a nd
showed that tradisustainable to optional process scalerate and produce
ing will have to be
through time but
The annual
augmented b y d e also leading to the
Symposia on
sign co-optimiefficient use and reVLSI Circuits and
z a t i o n a n d novel
placement of more
Technology are
integration. He asunsustainable techregarded as the
serted that materinologies related to
premier midyear
als and technologies
transportation and
gatherings for the
used to produce dein f r a st r uc t u r e. A
advancement of
vices, processes,
theme she emphamicroelectronics
and packaging are
sized throughout
technology and
likely to become
her talk was that it
circuits.
more diverse going
takes far more enforward and that userg y to move the
ing " the right matephysical than the
rial for the right task " would become
digital. Consistent with that theme
the norm. Specifically, the process
was her estimate that, due to the
integration of heterogeneous matereduction in travel, holding the

rials, such as bonding germanium
PMOSs with silicon NMOSs, shows
promise, especially when considering that a record ION versus IOFF
has been demonstrated through
t h is te c h nique. He also outlined
the package integration of heterogeneous technologies to balance many
competing tradeoffs among economics, performance, energy use,
and time to market.
Mayber r y advo c ated for t he
co-development of software and
hardware solutions in the future.
Recently, deep learning architectures have generated significant
adv a n ces i n p e r for m a n ce a n d
accuracy for applications such as
computer vision versus the use
of more conventional computing
approaches. During the next few
decades, he asserted, techniques
such as graph analytics and neuromorphic and quantum computing could hold the key to dramatic
incr e a ses in p er for m a nce a nd
enable new and unique applications,
as shown in Figure 1. However, as
he explained during the interactive but virtual Q&A session, these
potential future technologies could

FIGURE 1: Mike Mayberry's vision of potential hardware/software computing solutions beyond conventional and deep learning paradigms.

122

FA L L 2 0 2 0

IEEE SOLID-STATE CIRCUITS MAGAZINE

IEEE Solid-States Circuits Magazine - Fall 2020

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Fall 2020