Hi, thanks for tuning into Singularity
Prosperity. This video is the second in a
multi-part series discussing computing.
In this video, we'll be discussing modern
computing - more specifically, Moore's Law
with the exponential growth of
technology due to our ability to pack
more and more transistors into
integrated circuits and the potential
death of Moore's Law! In the previous
video we discussed the evolution of the
field of computing, from the pre-computer
era to vacuum tubes, transistors and
finally the integrated circuit. I highly
recommend you check it out for some more
background context into computing. One of
the largest breakthroughs in electronics
and computing was the integrated circuit,
a way to put many transistors into a
single chip instead of individually
wiring them together. After Gordon Moore,
one of the founders of Intel saw the
doubling of transistors on integrated
circuits, he extrapolated the data and
made one of the greatest predictions in
human history:
"The number of transistors and resistors
on a chip would double every 24 months",
in other words, computing power would
double every 24 months at low cost.
Integrated circuits are used in
practically every device that requires a
digital logic operation to be done, these
operations can consist of converting
analog signals to digital, amplifiers,
computation oriented - the list can go on
and on. As our world becomes more
digitized, the amount of integrated
circuits will only continue to increase.
In fact, every year since the birth of
the IC in 1958,
more and more have been produced year in
and year out, with for the first time
ever in 2018, more than 1 trillion to be
produced in just that year alone! The
most astonishing part about this fact is
that this number is only set to grow as
sensors and computers become ever more
ubiquitous and affordable. Looking at
just one field in the broad scope of
technology, the Internet of Things,
connected devices are expected to nearly
double from this year to 2020, reaching
over 50 billion, with each one of those
devices having tens, hundreds or even
more ICs within them. Almost everything
is on its way or has an IC in it from:
airplanes, cars, speakers, blu-ray players,
toys, door locks, lights and countless
other technologies. The real power of ICs
however and what has really shaped our
world and fueled the growth of Moore's Law
is a use of ICs for computing. When we
think of computers, the first component
that often comes to mind is the
microprocessor. A specialized integrated
circuit that is made for
computing. Microprocessors used to be just
one IC, but as computers evolved and
more complexity in design was needed,
the central processing unit emerged. A
CPU simply put is the part of the computer
that executes instructions. It can be
implemented using a single IC, multiple
ICs, individually wired transistors or a
room full of vacuum tubes and relays. A
microprocessor is just a single chip
implementation of the CPU, which is why
the terms are often used synonymously.
Other examples of where integrated
circuits in modern computers are used
is the RAM, DRAM, hard drives, solid-state
drives, GPUs, the motherboard which is
essentially many ICs with many functions
and more - essentially there are tens of
hundreds in typical computers, each with
the specific tasks. As the number of
transistors on integrated circuits has
increased, has led to the ability for the
production of components with more storage,
speed, memory, etc than ever before at
increasingly affordable prices. In 1971,
the first commercial microprocessor, the
Intel 4004 had a transistor count of
2,300, 8 years later in 1979 the
Intel 8088 had 29000, 10
years later the Intel 800486 and nearly
1.2 million. Then in 1999 the Pentium 3
had 9.5 million and following that in
2000 the Pentium 4 had 42 million. Since
the 2000s, the transistors on chips have
been increasing at an increasingly fast
rate. While this applies to
microprocessors, similar trends have been
followed in all integrated circuit
applications. For example, as seen in the
price of memory and storage over the
years per gigabyte as the number of
transistors has increased:
[Music]
The majority of people nowadays have a
computer, whether it be a desktop, laptop
or more commonly a smartphone. While the
latest commercial desktop and laptop
processors are using 14 nanometer
transistor sizes, as of this year, the
mobile industry has pushed forward with
10 nanometers. The Samsung S8 with its
Exynos 8895, Qualcomms Snapdragon 835
and Apple's iPhone X with its A11
Bionic chip, all feature a 10
nanometer transistor process. To put that
number in perspective, 10 nanometers is
one billionth of a meter: Point Zero Zero
Zero Zero Zero Zero Zero Zero One meters (.000000001m),
20 Silicon atoms wide, you can fit about
10,000 along the width of an average
human hair! When holding your new phone,
you're essentially holding 3 to 4
billion transistors! As seen here,
transistor sizes have been decreasing
exceedingly fast since 1971, but since
the 14 to 16 nanometer range, things have
slowed down quite a bit. One of the
primary reasons this was due to was
quantum effects. One of these effects is
quantum tunneling, this caused because
the distance between the source and
drain of the transistor is so small that
electrons jump across the barrier. So
instead of staying in their intended
logic gate, the electrons end up
continuously flowing from one gate to
the next, essentially making it
impossible for the transistor to have an
off state. Here is a conventional planar
CMOS transistor. On top of a silicon
substrate are two electrical terminals,
the source and the drain, separated by an
electrically controlled gate. When
voltage is applied to the gate a
conductive channel is formed and
electrons flow from the source to the
drain.
When voltage is removed the current
should completely cease, however, in modern
transistors substantial leakage flows
even when the gate is turned off.
Unfortunately, this leakage current
increases with every generation of
transistors and represents a growing
proportion of power consumption. To solve
this, a radical redesign of transistor
has taken the industry by storm, the
FinFET. The FIN-shaped Field Effect
Transistor, essentially takes a typical
2D planar transistor and reorients the
gate vertically to make it 3D. This allows
more gate control since now the gate of
the transistor covers the top and sides,
which therefore reduces the leakage induced
by quantum tunneling. Each FinFET has
three fins, with the fins being the
source and drain of the transistor going
through the gate. The FinFET also allows
for less heat generation and power
consumption, since one gate can
essentially control three nodes, which
correlates the longer battery life spans.
FinFETs allowed the scaling of
transistors from 16 nanometers to 10
nanometers, as exemplified by the mobile
processors mentioned earlier and now
other major semiconductor manufacturers,
such as Intel are releasing their 10
nanometer desktop and laptop lines in
2018: with Cannon and Ice Lake. As a side
note, Intel has been using FinFETs since
their 22 nanometer Ivy Bridge
architecture, referring to them as
Tri-gate Transistors. FinFETs should
allow scaling down to 7 nanometers with
minimal leakage, with IBM successfully
demonstrating a 7 nanometer node in 2015
and release expected by the 2019 to 2020
range. Unfortunately when scaling lower
than 7 nanometers, quantum tunneling once
again rears its head. The further
miniaturization of transistors will open
up the Internet of Things for everybody,
with the ability to embed sensors into
nearly anything. What's more exciting to
me personally is the applications of
this miniaturization on microcontrollers
like the Raspberry Pi, which is now more
powerful than some early to mid-2000
mid-2000 level computers. All the
technological leaps and bounds due to
the miniaturization of the transistor
have been amazing, but harbor one
question: When will the shrinking a
transistor stop, and Moore's Law end?
At its current definition, the
acceleration of Moore's Law cannot continue
forever.
Currently, Moore's Law is a physical law,
it is linked to the size of a silicon
atom, therefore we will hit a minimum
value for the size of a transistor. As
explained previously, FinFETs have
allowed scaling of transistors down to
7 nanometers with 5 nanometers
still being a theoretical possibility,
but that may not work due to electron
leakage induced by quantum tunneling.
Recently, as of June 2017, IBM announced
they had scaled down the transistor to 5
nanometers by reorienting the transistor
once again, and were able to fit 30
billion transistors on a chip the size
of a fingernail! From the 2D planar to 3D
FinFET and now to the GAAFET. The Gate-
All-Around Field Effect Transistor, which
is sort of a 2D-3D mix, and relies
heavily on FinFET design methodologies.
The GAAFET essentially adds another
transistor compared to FinFETs. Instead
of the fin of the source and drain being
aligned vertically, GAAFETs align them
horizontally using silicon nanosheets,
as seen in this photo. Due to these added
transistors, devices will become more
power and heat efficient once again, up
to 40% faster and 75% more efficient,
compared to 10 nanometer processors that
are just coming to market today. Using silicon
nanosheets that will enable 5 nanometer
technology. To do this, we developed an
entirely new architecture. Today's chips
use what is known as a FinFet
architecture and we even use it in our
state of the art 7 nanometer chips. But to go
beyond 7 nanometers and build new 5
nanometer technology, we use stacks of silicon
nanosheets. Here we can see the
difference between today's FinFET
architecture and the stacked nanosheets.
Instead of three fins side-by-side in
which the current flows along the side
of the fin, the silicon nanosheets are
layered on top of one another and the
current flows along the direction of the
sheet. It's clear today 5 nanometer chips are
possible, it's going to happen! The GAAFet
is expected to start rolling out to
market around 2021 to 2023, and this
technology is also theorized to allow
scaling down to 3 nanometers, which is
now in its research phase and may come
to market around the 2024 to 2026 range.
At 1 nanometer we reach the smallest a
transistor can go if we rely on
silicon, at 1 nanometer the source
and drain are just 2 silicon atoms across! It
is unknown if we can reach this
milestone at a commercial level, but some
tests do show promise through the use of
carbon nanotubes and other design
methodologies. Moore's Law based on the
miniaturization the silicon transistor
will die around the mid to late-2020s.....
Now all those people, articles, videos
(even this video) use Moore's Law being dead
as a provocative title. Beyond the
miniaturization of the transistor, there
are various other aspects of computing
to perfect before Moore's Law even comes
close to ending. In fact, it may never end
until we reach Planck level technology
at 10^-35 meters in size.
So, what we will see in the next few years
is an uncoupling of Moore's Law from
transistor density and more towards raw
computing performance, through multiple
design methodologies. Looking at the
progression of Moore's Law in terms of
computing performance, over the past 120
years, from Babbage's analytical engine,
we can see that the last seven data
points are given by GPU performance not
CPU. With some of them NVIDIAs latest
cards having over 8 billion transistors,
with the Titan X, their latest card,
having 12 billion and still using 16
nanometer FinFET architecture! In the
next video in this computing series,
we'll expand further on GPUs as
computing alternatives, as well as other
ways to maximize classical computing
architecture from new materials, FPGAs,
additional cores and more! In videos
afterwards, we'll discuss various new
branches of computing that are currently
being developed such as: parallel
computing, bio and quantum computers,
optical computing, and neuromorphic
computing!
At this point the video has come to a
conclusion. I'd like to thank you for
taking the time to watch it, if you
enjoyed it please leave a thumbs up and
if you want me to elaborate on any of
the topics discussed or have any topic
suggestions please leave them in the
comments below.
Consider subscribing to my channel for
more content, follow my Medium
publication for accompanying blogs and
like my Facebook page for more
bite-sized chunks of content. This has been
Ankur, you've been watching
Singularity Prosperity and I'll see you
again soon!
[Music]
For more infomation >> CHALLENGE BULDAK KOREA 100 CABE HARGA MURAH BGT?? - Duration: 13:25. 
For more infomation >> 8BOTSBOYZ - มันไม่ดี (Official MV) - Duration: 4:04. 
For more infomation >> La mia OPINIONE su iPhone X - Duration: 3:51. 
For more infomation >> [Radio] RAP 1 - Những Bài Rap Buồn Hay Nhất Dành Cho Con Gái 2017 (Nhạc Rap Tuyển Chọn) - Duration: 1:09:59. 
For more infomation >> Xem Chứng Minh Thư Biết Vận Mệnh Cuộc Đời SƯỚNG KHỔ- GIÀU NGHÈO Không Cần Đi Xem Bói - Duration: 14:45. 
For more infomation >> (con SUBTÍTULOS) SERIES para aprender ESPAÑOL || PRÁCTICA de escucha || María Español - Duration: 23:53. 
For more infomation >> عــــــــاجل: الكشف عن مفاجأة مغربية إماراتية بسبب القمر الاصطناعي "محمد السادس" - Duration: 1:43. 
For more infomation >> Turismo - GTA V Cinematic | by KNWHTCB - Duration: 4:16. 
Không có nhận xét nào:
Đăng nhận xét