From Bits to Billions Enabling the Digital Revolution

From Bits to Billions Enabling the Digital Revolution - or How to retire in less than a decade

Digital Revolution • We live in a time where a great deal of the information or data we receive has been created, stored, transmitted, encrypted or manipulated digitally. . . • Why? • computation, especially compression, is relatively easy on digital data using a computer • MP 3 music, JPG pictures, MPG et al movies • voice calls, maps, GPS, networking, storage • What is doing that computation? • in the vast majority of cases. . . an ARM processor 2

ARM • A small, but powerful, microprocessor • able to run for many hours off a battery • • Practically every cellphone the majority of digital cameras and MP 3 players and GPS units and printers and disk drives and networking gear • and Blue. Tooth, and Smart Cards, and Cars &. . . 3

Papal Inauguration 2005 4

Papal Inauguration 2013 5

Dave Jaggar • Graduated from Canterbury in 1991 • Year 11 Applied Maths teacher said I'd never be an engineer • BSc & MSc in Computer Science • Master's Thesis studied the ARM processor before it was well known • Hint: and I studied a lots of old 'forgotten' stuff. . . • MSc Thesis marked by Professor Doran here in Auckland • I did all the Computer Science papers Canterbury offered • Hint : a fairly broad background • ARM employee number 17 • Senior Software Engineer in 1991 and half of 1992 • rewrote the ARM simulator from my thesis • Head of Technical Marketing & Architecture Committee • Architected most ARM products from 1992 to 2000 • Head of ARM Austin 1996 to 2000 • High performance CPU team • Retired in 2000 (mostly. . . ) • Hint: get in at (or near) the ground floor of a startup 6

Gearing • Before technology is successful, its often fairly dull • • • Google. . . just another search engine Facebook. . . online register of Harvard students faces You. Tube, Twitter, Instagram, Tumblr, Amazon, EBay. . . Apple App store • 75 billion downloads. . . revenue soon to exceed i. Tunes Bottled Water • US$195 B in 2018 Hint: the next big thing probably sounds pretty dull today • ARM's Business Model • design a processor once • license it hundreds of times • for a few million dollars each time • licensees sell hundreds of millions of processors • and pay ARM a small royalty • on average everyone on the planet has bought about 7 ARM processors • ARM's Market Cap is somewhere over US$20 billion • Lesson: a small number times the world population is a large number 7

Shipments 8

ARM Startup • ARM started with 12 engineers from Acorn in 1991 • • Added about 10 staff per year for 3 years • • • as our major outgoing was in salaries, we stayed quite small borrowed legal, finance, marketing people from elsewhere. . . Spending any money needed justification • • • mostly Cambridge graduates, smart guys, but relatively "narrow" and UK£ 1. 5 millon (US$2. 5 million) from Apple wanted ARM 6 for its Newton project spending £ 10 K needed one A 4 page of justification (approx 600 words) • so spending £ 100 needed 6 words of justification ☺ limited administration overhead (one office manager) no company credit cards limited company perks economy class air fares for everyone Hint : Cash is king for a startup, so run lean and mean until your income is sure and steady, and then, Lesson: keep running lean and mean as it's part of your culture. . . 9

ARM Barbecue 1991 10

ARM Company Motto • Robin Saxby • • • Invented the ARM 'Partnership' Business Model Completely foresaw, and the created, ARM domination "We'll Never Make or Sell Chips" "Solutions not Problems" "Every Problem is An Opportunity" • Work Hard But Have Fun • I worked very hard (60 hour weeks) • I had great fun most of the time • Hint: "The Harder I Worked, the Luckier I Became" Hint : Do What You Enjoy (that old chestnut) - And if you don't enjoy it, get on with it anyway, just to get it done - Lesson: If you aren't making mistakes, you're probably standing still 11

ARM Partners 12

The ARM 7 Processor (1992) • The two original architects of the ARM processor didn't join the startup, and after a death in the company, I was the only person in the company that knew how the processor worked. . . but really I had no clue. . . • I was completely thrown in the deep-end • to this day I remember how scared I was, I really had no clue • but, every problem is an opportunity • We needed a follow-on processor to the ARM 6 • Before my input, ARM was just another chip for computers • Opportunity: redesign it to replace high volume, low cost, embedded processors • as dull as bottled water. . . 13

ARM 7 and ARM 7 DMI • First, a quick spin to fix some basic problems with ARM 6 • Made it much faster with one simple change • • This was the opportunity to give it a new name ("phew") It wasn't fully backwards compatible with ARM 6 Huge uproar over 8 instructions from one of the original ARM architects Hint: Watch out for politics (Lesson: I didn't, and I'm still useless at it) • Made it run at 3. 3 volts (down from 5 volts) • Made it remember where it was up to after a reset • Filed some ("narrow") patents, quite specific, but easily defended • Then, two optional parts • Added a faster multiplier • essential for a lot of Digital Signal Processing • Added simple but powerful debug hardware • the chip wasn't being used in a computer with a keyboard & screen • so to debug code extra hardware was needed to see what was going on • Filed some decent patents that covered these parts of the chip well • Started showing it to customers • the response was good, but after they had evaluated it, they said "ARM programs are too big, you're blowing our memory budget" 14

Nippon Investment & Finance ('93) • Nippon Investment and Finance (part of Daiwa Securities) invested UK£ 650, 000 in ARM • ARM did not need (or ever use) the money • ARM needed exposure in Asia • I knew (know) nothing about this stuff • Over dinner I asked one question "When will you sell your ARM shares? " • Answer "When the price hits the target we've set" • Hint: Share prices go up, until they go down, so decide upfront what profit you expect, otherwise you get sucked into greed 15

ARM 7 TDMI (1994) • Thumb, a second instruction set • More compact than the original one • designed halve the cost of the system memory • smaller code than the chips it was replacing • room for new features in a new product with no increase in cost • ARM instruction set still available for maximum performance at high cost • Smart politically. . . relatively small change to the overall chip • • Thumb, just a wee thing on the end of your ARM Extremely well accepted by end customers • Motorola, Intel, MIPS etc too big, too slow to react • I had no idea how slowly big companies moved • It "rewrote the rulebook" on what embedded processing was • Two key ARM patents ("broad" ones at last) • Hint: Protect Your Intellectual Property in a variety of "breadths" • Lesson: many patent lawyers don't get this point until it's too late. . . 16

ARM changes gear (1994 -1995) • I strap an airplane to my backside for a year • someone worked out my average speed was more than 300 km/h over a whole year … • • Nokia, Ericsson, Alcatel, Qualcomm, Motorola(!) Apple, Microsoft, Seagate, Maxtor, Cisco TI, IBM, Oracle, Bell Labs Epson, HP, Brother Nintendo, Sega, Yamaha, Sony Ford, GM, Toyota, Honda, Mercedes, Siemens Kodak, Nikon, Panasonic, JVC, Fujitsu NKK Steel, Eurofighter, and the NSA (I think) 17

Much Success 18

Strong. ARM ('94 & '95) • Digital Equipment Corp (DEC). . . • were the 3 rd or 4 th biggest computer company in the world • one team did Prism, Alpha & PPC (Somerset) • Prism cancelled because it wasn't part of DEC's mainstream VAX product line • So the engineers called the next CPU "Extended VAX" • fooling management that it was VAX like • Hint: it wasn't. . . • Launched as DEC Alpha • 200 MHz when every on else was around 66 MHz • But too late to save DEC • Probably the best design team on the planet • and a pretty small team. . . • and they wanted to build an ARM • Hint: there's always someone smarter than you 19

The ARM Architecture • For DEC I wrote the ARM Architecture Reference Manual so they actually knew what to build • ARM, literally wrote the Rule Book • A couple of months on site in Austin, Texas in late 1994 • Got to know the team very well, sat in on design meetings, learnt a whole lot, and loved Austin • I didn't downplay Thumb, but I didn't talk it up either • old ARM was not well protected so they could do it anyway • They knew nothing about embedded design • PC/Workstation backgrounds • A little like Acorn, MIPS, SPARC, most of Intel etc • Lesson: "You've got two ears and one mouth, try to use them in that proportion" 20

Then Back In Cambridge (1995) • ARM 8 was a complex resource hungry processor for Acorn • Acorn on the design team, paying the bills, 20 engineers from ARM • Not targeted at embedded (no Thumb, no Debug) • I quietly started new processor ARM 8 E • ARM 8 family member in name only • About as much about ARM 8 as Alpha was about VAX • i. e. none … just hiding it from the Acorn Board. . . Lesson from VAX/Alpha • A direct rip off of the Strong. ARM datapath • fast. . . • And a small team like Strong. ARM • hopefully fast to market too (Hint: It wasn't Lesson: Don't take you're eye off the ball) • Added Thumb and Debug • performance for the embedded market • ARM 8 E was launched as ARM 9 TDMI 21

ARM 7 TDMI and ARM 9 TDMI • ARM 7 x 0 • a single combined instruction and data cache • or no cache at all, maybe just on chip SRAM • • 3 stage pipeline (f, d, e) very small die size very low power maybe 10 times faster than an 8 bitter • ARM 9 x 0 • • • separate instruction and data caches (or memories) single off chip bus 5 stage pipeline (f, d, r, e, w) about double the performance of ARM 7 x 0 and about double the power and die size • Together responsible for 90% to 95% of ARMs shipped to date 22

ARM 10/Strong. ARM 2 (1996) • Joint Development between ARM and Digital • Half the costs, learn from each other • Bring a few engineers over from the UK • Digital had a good idea of how to go much faster • experience with floating point and better for Unix • ARM knew all about embedded • Digital needed Thumb IP • Cambridge is cold in winter • Austin isn't • Margaritas taste better than warm beer • My eldest child was born • I wanted to be home more • Robin's only regret • ARM was getting bigger, slower, political. . . 23

Austin Design Centre ('97 & '98) • Sometimes things just go to custard • Digital Equipment HQ sues Intel • Intel Buy Digital (cheaper than a law suit) • Strong. ARM design team quit to do a startup • but they didn't know about embedded … • Chip design tool supplier bought and scrapped • changing chip design tools is a big deal. . . • • • ARM 9 team were having huge problems with OS bugs Big problems = big opportunity Start an ARM design centre in Austin • teach ourselves high performance, floating point, running Unix • New chip, new team, new tools, new country • Hiring frenzy • plus infrastructure, buildings, admin, timezones • Startup atmosphere, I just replicated early ARM • borrowed legal, finance, marketing people from elsewhere. . . • High Performance chips that Cambridge could proliferate 24

How to avoid ARM 9 OS bugs • I wrote an instruction set simulator for my thesis • ARMulator could boot an OS in about 1 minute • System was actually 'usable' • type on keyboard, tap on screen etc • I used it to model ARM 10, and record what the correct path was through the OS • played the real chip models against that path • fixed every difference • Lesson: Broad background from University really helped 25

Broad not Bored • Cambridge graduate. . . 26

Initial Public Offering (IPO) • In 1998 ARM listed its shares on the London and NASDAQ Stock Exchanges • ARM employees had been given share options every year • I had no idea what they were worth • As it turned out, nor did anyone else • Lesson: NIF Investment in 1993 • Know your sell price and stick to it • But how to know what was a good sell price 27

The Four M Plan • There were 4 levels of wealth that were important to me • Split my share options into four equal sized chucks and sell when each of the first three quarters hit it's price goal. See what happens with the final quarter • Mortgage. . . own the family home in Texas mortgage free • about US$500 K before 40% tax • IPO set a share price, so we hit that immediately • Mac. Donalds. . . put enough money in the bank that flipping hamburgers at Mac. Donalds will support the family • about US$1 million after tax • Share price hit that level about 6 months later. . . • no More work. . . retire early if desired • about US$3 million after tax • Share price reached that about 3 months after we sold the last lot. . . • Monaco plan. . . I can retire to Monaco. . . • try and pick the top of the market (gamble one quarter of my stock) 28

Year 2000 • ARM 10200 first silicon fully functional • Revision 1 well under control with a few enhancements • Design Flow was actually flowing … • ARM 11 specified and understood (Cambridge) • Austin team (50 people) were actually a team • a layer of good leaders beneath me • I needed a rest • Share price approaching UK£ 10 (Market Cap >U$20 b) 29

Retirement • I bailed back to NZ in early June 2000, just under 9 years at ARM • technically on sabbatical • thinking about how to do a CPU that was completely new • third child born in July • eldest turning 4 … schooling • Robin Saxby had a regret. . . • then 3 months later. com tech stock crash • 60% wiped off NASDAQ over a few months • I sold up my final quarter of shares at the absolute top of the market • Lesson: Luck. . . but why did I have so many shares in the first place. . . hard work • Lesson: Shares go up, and then down (often quickly), and then take 10 or 15 years to recover. . . so don't be greedy 30

2001 to 2014 • Built a home on the outskirts of Christchurch • 2 years of planning and 2 years of building • Hint: house architecture is much easier than chip architecture. . . • Children • I'm just a Dad • Travel • Show my children some of the world • A few lectures • Performed a wedding ceremony • For Robin Saxby's daughter 31

The World is Changing Faster • So much opportunity to be creative now • Combine a 3 D printer and a Raspberry. Pi • limitless possibilities • Consumer devices like phones, games, media players • Robotics • Home Automation • Kick. Starter. . . • My light switch • • • Pi with BT & Wi. Fi Touchscreen CBus IF Plastic XBMC • Hint: What could be more dull than a light switch. . . 32

Biggest Lesson • A Cambridge University Degree isn't any better than a Canterbury University Degree • and in some ways it's worse • if you think you know everything you're not going to be any good at inventing what will come next • Often what comes next will be a combination of many things, things that no one yet knows • a broad education is more valuable than a narrow one • because you can stand tall upon a broad foundation • ARM was inventing new stuff, so it couldn't be taught • not just know how to know, or know how to be taught • you need to know how to learn, independently • Mistakes are good things, admit them, embrace them • they mean you are learning, so why not just try it and see 33