Looking back and forwards: the “Execution Dilemma” and the “Importance of Radical Innovation” Domenico Rossi ST Microelectronics CAS Workshop 2017 Pavia – March 20 th – 21 st Almo Collegio Borromeo
The semiconductor industry: why innovation counts A Young Industry: 1 E+3 • 1947: 1 st Transistor Func4ons/Chip [ Giga (10^9) – bits, transistors • Early 60’s: 1 st ICs 1 E+2 • 1968: 1 st Self Aligned CMOS 1 E+1 • Since then: Moore’s Law: 1 E+0 • You know what to MPU Transistors/Chip do High-Performance • not easy 1 E-1 MPU Transistors/Chip Cost-Performance but Consolidating 1 E-2 2000 2005 2010 2015 2020 2025 * Source: 2014 ITRS 2.0 Average Industry Moore’s Law year 2X Functions/Chip per 2 Years
What about execution? The Zetta Byte Era
The Zettabyte Era 1/2 4 • Annual global IP traffic has passed the zettabyte ([ZB]; 1000 exabytes [EB]) threshold by the end of 2016 • It will grow at a compound annual growth rate (CAGR) of 22 percent from 2015 to 2020. • Monthly IP traffic will reach 25 GB per capita by 2020, up from 10 GB per capita in 2015 . • Smartphone traffic will exceed PC traffic by 2020 . • In 2020, PCs will 29% of the traffic (it was 53% IN 2015). • Smartphones will account for 30 percent of total IP traffic in 2020, up from 8 percent in 2015. • The number of devices connected to IP networks will be more than three times the global population by 2020 . • There will be 26.3 billion networked devices in 2020, up from 16.3 billion in 2015.
The Zettabyte Era 2/2 5 • Broadband speeds will nearly double by 2020 . • By 2020, global fixed broadband speeds will double reaching 47.7 Mbps, up from 24.7 Mbps in 2015. • Internet video TV grew 50% in 2015 . • This traffic will continue to grow at a rapid pace, increasing 3.6-fold by 2020. • Consumer video-on-demand (VoD) traffic will nearly double by 2020 . • The amount of VoD traffic in 2020 will be equivalent to 7.2 billion DVDs per month. • Internet gaming traffic will grow sevenfold from 2015 to 2020, a CAGR of 46 percent. • Globally, Internet gaming traffic will be 4 percent of consumer Internet traffic in 2020, up from 2 percent in 2015.
Market Requirements • It’s no secret that data centers, consume a huge amount of energy • In US: 70 billion KWh of electricity in 2014 = 2% of the country’s total energy consumption equivalent to the energy consumed in 6.4 million average American homes that year. • It is not known how extraordinary the ability of data center industry’s ability to improve energy was. • While the demand for data center capacity in the US grew tremendously the total data center energy consumption grew only slightly. • Staying at the efficiency levels of 2010, data centers would have consumed twice this level of energy according to US Department of Energy. • Researchers expect total US data center energy consumption to grow only by 4 percent between now and 2020 – reaching about 73 billion KWh.
Technology Adoption • Servers have gotten a lot more powerful and efficient. • Semiconductor industry played an important role to make it happen Technology adoption by date of ASIC Award: 130nm Courtesy of Cisco. 90nm 65nm 55nm 45/40nm 32/28nm 22/20nm 16/14nm 10/7nm 2015 2016 2017 2010 2012 2013 2014 2007 2009 2011 2006 2008
Speed / Power / Area Evolution SPEED + 25% 16FFC N07 POWER - 30% mm2 AREA * - 48% Courtesy of TSMC
ASICs for Networking Die 400 – 450 mm2 Power 150 – 175 W FcBGA 50 – 55 mm, 12 – 14 layers 2500 – 3000 2008 2016 Same di size, Same Architecure, Same Enabling IP, Same Package Gate count x 10, Memory Size X10, Core Speed x 2, High Speed links x10,
Today Complexity …… Logic 700M gates Memory 2 GigaBits Serdes 150-200@56GB Memory I/F @2GBbps (HBM)
ASIC: the execution … paradigm Libraries, IPs CAD Tools COMPLEX ASIC EXECUTION: 12 Months Tape 1 st Netlist 2 nd Netlist Final Netlist out 12 months for ASIC execution, regardless the technology and complexity
Charlie Chaplin in 1936 “Modern Times”: Little Tramp working on the giant machine.
the production line … Implementation Synthesis Sign-Off DFT 12 months? Yes we can
What about performances? Challenge: how to extract more value from a new technology that costs many $B? Frequency (GHz) Margins Lost performances Sign-off with increased margins Year Technology Node
The Productivity Vs.Performances 15 Challenge • EDA industry has not tackled this productivity to performnca gap very efficiently. • While traditional issues existed regarding chip design, centered on speed and the chip area are quite well addresse, newer concerns related to power, signal integrity, yield, and DFM are still to be addressed. • EDA will have to evolve to aid true system level definitions, to help architect hardware and software, based on performance, power and other functional needs. • The software community has to be seamlessly integrated into the EDA fold with lesser pain. • IP reusability and integration (especially if from multi-vendors)
… ..and what about innovation?
The XV century “Old World” • In the XV century the geographical knowledge of the world was fairly limited and based on the following assumptions • Only three large continents (i.e., Europe, Africa, and Asia) • Only one large sea (i.e., Ocean Sea) • The most difficult challenge in the XV century was to reach the fabulous and wealthy markets of India and Cathay (i.e., China) by sailing West across the Ocean Sea (i.e., Atlantic Ocean)
Columbus: an innovator in the XV century • Christopher had a clear Target Reaching East Indies by sailing Westward • a Budget Spain’s King Ferdinand and Queen Isabelle financed the journey Cristoforo Colombo • the Means to succeed (1451-1506) was an Italian The Famous Columbus Caravels explorer, navigator, colonizer, (Nina, Pinta, Santa Maria) and citizen of the Republic of Genoa
Columbus: an innovator in the XV century • Between 1492 and 1504 Columbus carries out four expeditions, but he never reached the East Indies • During his 2 nd voyage after shoring in Cuba Columbus was convinced that eventually he had reached Cathay • He obstinately refused to recognize the new reality emerging from his voyages • Cuba was not Cathay nevertheless..
Columbus: an innovator in the XV century • He proved he was able to execute (2000 miles sailed) • the ROI for Spain was good • Spain enjoyed European Supremacy overseas until the late sixteenth century
GPS Military Automotive Portable Navi / Mobile Late 90’s 1980/90 early 2000 1.4B Smart Phones in 2015 88 M cars in 2015
Columbus: an innovator in the XV century • Columbus failed , but he succeeded • Sometimes You “just” fail • Sometimes You do not fail , but You do not succeed • When the first working laser was reported in 1960, it was described as "a solution seeking problems" • Very often You fail , “before” You succeed
MEMS Automotive µ actuators HDD Gaming / Mobile 1980/90 Late 90’s no massive early 2000 deployment 88 M cars in 2015 1.4B Smart Phones in 2015
From Columbus to Bayes • Bringing New Technologies up, Opening New Markets, Launching New Products, like Discovering a New Continent, could be defined in the decision theory, “Uncertain Events” • “Uncertain” because of the unpredictability of events which could influence the outcome of a given decision, but “Observable” • We have prior beliefs that affect how we interpret new evidence: Cuba was not Cathay! • We conflicts between the original theoretical project (i.e., the prior) and an emergent and unexpected new reality (i.e., the posterior)
The Bayes’ Theorem Our beliefs (i.e., prior) should converge towards the truth (i.e., posterior) as we are presented with more evidence over time The stake holders converge towards the truth and make the right decision only if Thomas Bayes ideas are debated, experience is (1701 - 1761) was an achieved, and evidence is uncovered English statistician, philosopher, and Presbyterian minister
Open Innovation 2.0: a new way of doing innovation Government/Public Source: Open innova7on 2.0 yearbook 2016 Academia European Commission Industry Ci8zens • The traditional paradigm of Open Innovation 1.0 based on a tight collaboration between the Company and a Value Network of Universities, Customers, and Suppliers is no longer sufficient • Therefore new key actors such as Government, Public Authorities, and Citizens must be included in a new collaborative ecosystem (Open Innovation 2.0)
The path to Autonomous Driving 27 IoT and Autonomous Driving: two examples of Open Innovation 2.0
The path toward Autonomous Driving: the Governments 2014 2016 2018 2020 PED LKA LDW AEB LDW AEB PED LDW AEB PED LKA LDW AEB PED Potential LDW AEB PED LDW = Lane Departure Warning AEB = Autonomous Emergency Braking AEB PED LKA LDW PED = Pedestrian Detection Source : Regional NCAP agencies LKA = Lane Keeping Assist From Passive to Active Safety to Autonomous Driving
The path toward Autonomous Driving: the Customers C Car Makers System Makers
Recommend
More recommend
