Transcript
8/3/2019 Moore's law - SoC
1/35
SoC DESIGN
Motivation for SoC Design
Review of Moores Law
8/3/2019 Moore's law - SoC
2/35
Outline
History of Transistors and circuits
The Integrated circuit manufacturing
process
Moore Law is announced
Benefits of ICs
Extrapolating Moores Law to itsconclusion
Technological advances
Moores Law version 2?
8/3/2019 Moore's law - SoC
3/35
Discrete Transistors and Circuits
The transistor succeeded the valve in the late 1940s
Electronic engineers began to design complex circuits usingdiscrete componentstransistors, resistors, capacitors
Performance and other problems were noticed due to thenumber of separate components
Circuits were unreliable and heavy
High power consumptionlong time to assemble
Expensive to produce
8/3/2019 Moore's law - SoC
4/35
The SolutionIntegrated Circuits
Build entire circuit on a wafer of silicon
Use masking and spraying techniques in manufacture
Pure silicon wafers made from large crystals of silicon
Areas of silicon doped with suitable elements e.g. Be
Conductive tracks made from aluminium
Use this technique to produce other components e.g. capacitorsand resistors on the same wafer
8/3/2019 Moore's law - SoC
5/35
Problems solved
Inter-device distances reducedfaster circuits
Lightweight circuitssuitable for space travel
Cheaper assembly costafter recovery of R&D costs
Identical circuit propertiesbetter matching
Less power requiredless heat dissipated
Smaller circuitssmaller devices could be built
8/3/2019 Moore's law - SoC
6/35
INTRODUCTION
8/3/2019 Moore's law - SoC
7/35
Birth of Moores Law
The 19 April 1965 issue of Electronics magazine, marking theMcGraw-Hill publication's 35th anniversary,
It contained an article with the title "Cramming morecomponents onto integrated circuits."
Its author, Gordon E. Moore, director, Research andDevelopment Laboratories, Fairchild Semiconductor, had beenasked to predict what would happen over the next 10 years inthe semiconductor components industry.
His article speculated that by 1975 it would be possible tocram as many as 65 000 components onto a single silicon chipabout 6 millimetres square.
8/3/2019 Moore's law - SoC
8/35
Gordon Moore - Observations
Gordon Moore worked for Fairchild Semiconductors
He noticed a trend in IC manufacture
Every 2 years the number of components on an area of silicondoubled*
He published this work in 1965known as Moores Law
His predictions were for 10 years into the future
His work predicted personal computers and fast
telecommunication networks
* Sources vary regarding time period
8/3/2019 Moore's law - SoC
9/35
Moore's Law
Defined by Dr. Gordon Moore during the
sixties.
Predicts an exponential increase incomponent density over time, with a
doubling time of 18 months.
Applicable to microprocessors, DRAMs ,
DSPs and other microelectronics.
Monotonic increase in density observed
since the 1960s.
http://www.intel.com/pressroom/kits/bios/moore.htmhttp://www.intel.com/pressroom/kits/bios/moore.htm8/3/2019 Moore's law - SoC
10/35
COST AND CURVES
8/3/2019 Moore's law - SoC
11/35
# Components / Integrated function
8/3/2019 Moore's law - SoC
12/35
Moore's Law and Performance
The performance of computers isdetermined by architecture and clock speed.
Clock speed doubles over a 3 year period
due to the scaling laws on chip. Processors using identical or similar
architectures gain performance directly as afunction of Moore's Law.
Improvements in internal architecture canyield better gains than predicted by Moore'sLaw.
8/3/2019 Moore's law - SoC
13/35
Moores Law - Density
8/3/2019 Moore's law - SoC
14/35
Moores Law - Clock Speed
8/3/2019 Moore's law - SoC
15/35
Moores Law (Technologists)
Parameters
16 transistor/chip circa 1964
59% growth/year
36 years (2000) and counting
1styears 16 ??? 3rdyears 64 ???
15thyears 16,000 ???
24th
years 100,000 ??? 36thyears 300,000,000 ???
Was useful & then got more than 1,000,000 times
better!
8/3/2019 Moore's law - SoC
16/35
Moores Law Data (Technologists)
8/3/2019 Moore's law - SoC
17/35
Other Moores Laws
Other technologies improving rapidly
Magnetic disk capacity
DRAM capacity
Fiber-optic network bandwidth
Other aspects improving slowly Delay to memory
Delay to disk
Delay across networks
Computer Implementors Challenge Design with dissimilarly expanding resources
To Double computer performance every two years
A.k.a., (Popular) Moores Law
8/3/2019 Moore's law - SoC
18/35
Cost Side of Moores Law
About every two years: same computing at half cost
Long-term effect:
1940s Prototypes for calculating ballistic trajectories
1950s Early mainframes for large banks
1960s Mainframes flourish in many large businesses
1970s Minicomputers for business, science, & engineering
Early 1980s PCs for word processing & spreadsheets
Late 1980s PCs for desktop publishing
1990s PCs for games, multimedia, e-mail, & web
Jim Gray: In ten years you can buy a computer for the cost of its salestax today (assuming 3% or more)
8/3/2019 Moore's law - SoC
19/35
Graph of Moores Law
8/3/2019 Moore's law - SoC
20/35
Example
8/3/2019 Moore's law - SoC
21/35
Market
8/3/2019 Moore's law - SoC
22/35
IC Technologies
Small Scale Integration (SSI) combined around 10 discretecomponents onto 5mm square of silicon substrate.
SSI led to Medium Scale Integration (MSI), then Large ScaleIntegration (LSI) with many thousands of components in thesame area of silicon.
Very Large Scale Integration (VLSI) provided the means toimplement around 1 million components per chip.
Current technology produces silicon wafers with around 50million components per chip. The Pentium 4 has around 55million components on the wafer (2003).
8/3/2019 Moore's law - SoC
23/35
IC Technology
8/3/2019 Moore's law - SoC
24/35
The Next Step
INTEL have announced that they have the technology
to produce microprocessors containing more than 400
million transistors, running at 10 gigahertz andoperating at less than one volt, in the next five to ten
years.
This is in line with Moores law
8/3/2019 Moore's law - SoC
25/35
Shrinking the Size of a Component
How small can a component become?
What limits the size of a device?
What do we make the devices from?
Do quantum effects have an influence here?
If there is a limit, what happens to Moores Law?
8/3/2019 Moore's law - SoC
26/35
The Current Limitations
Circuits cannot be reduced beyond atomic size
Quantum effects reduce the reliability as size decreases
Lithographic techniques become more complex as the size of
components becomes smaller than the wavelength of light
Speed of electrical signals is finite
This suggests that Moores Law will finally end
8/3/2019 Moore's law - SoC
27/35
Why does the law exist?
Some of the factors that contribute to Moores Law:
Manufacturers wishing to keep up with the law
Competition between manufacturersSuccessive technologies providing better design tools
Customer demand for better products
Mans constant struggle to advance knowledge
There may be other factors too
8/3/2019 Moore's law - SoC
28/35
Future of Moores Law
Short-Term (1-5 years)
Will operate (due to prototypes in lab)
Fabrication cost will go up rapidly
Medium-Term (5-15 years) Exponential growth rate will likely slow
Trillion-dollar industry is motivated
Long-Term (>15 years)
May need new technology (chemical or quantum)
We can do better (e.g., human brain)
I would not close the patent office
8/3/2019 Moore's law - SoC
29/35
Future of Harnessing Moores Law
Thread-Level Parallelism Multiple processors cooperating (exists today)
More common in future with multiple processors per chip
Parallelism in Internet? The Grid.
System on a Chip Processor, memory, and I/O on one chip
Cost-performance leap like microprocessor?
(e.g., accelerometer at right)
Communication World-wide web & wireless cell phone fuse!
Other properties: robust & easy to design & use
8/3/2019 Moore's law - SoC
30/35
Lateral Thinking
To improve the performance of devices, new technologies arein development:
Quantum storage (quantum data registers - a faster, more
efficient way to store and retrieve data than the binary system
we use today)
Light operated transistors
Electro-optical polymers and more are showing newtechniques for achieving the ever higher performance
demanded by industry and consumers
http://www.umich.edu/~focuspfc/http://www.evidenttech.com/applications/optical_transistor.phphttp://www.evidenttech.com/applications/optical_transistor.phphttp://www.umich.edu/~focuspfc/8/3/2019 Moore's law - SoC
31/35
The Future of ICs
Moore acknowledged that his "law" won't hold forever. Heasserted that the right technological approaches can delay
"forever", extending the longevity of his original prediction.
Intel are working on new ideas such as SiGe and strainedsilicon to delay the end of Moores Law
Designing transistors that switch at speeds around THz (can
switch on and off a trillion times per second)
The advances continue!
8/3/2019 Moore's law - SoC
32/35
The End of the Line?
It is obvious that technology will improve
We may meet the lower size limit of a transistor
Therefore the abilities of the transistor itself will
have to improve instead
Faster switching, lower power designs etc.
ICs still improve
8/3/2019 Moore's law - SoC
33/35
Moores Law version 2?
After his law is no longer validwhat can we use to measuretrends?Component density?
Noit would be fairly constant
Performance?Yesbut which metric?
Switching rate?Individual or bulk?
Rise time?
Access time or read/ write timeOther measurable attributes
8/3/2019 Moore's law - SoC
34/35
Moore version 2s metric(s)
Technological advances will continue as long as there isdemand for digital devices
It is immaterial whether the component density limit is
reachedAnother metric will have to be chosen to allow the ICevolution to be mapped and to allow valid predictions to bemade
Which metricthis is extremely complex to choose
8/3/2019 Moore's law - SoC
35/35
Conclusion
Moores law will eventually reach its inevitable conclusion
Technology will continue to advance
ICs with improved properties will be manufactured
Another metric will need to be chosen to allow the future trends
to be mapped and predicted
The complexity of current IC design means this choice will be
difficult
top related