## Von Neumann Architecture – Computerphile

>>Sean: Professor Brailsford – a lot of people

talk about von Neumann architecture and we’ve talked about Babbage; we’ve talked

about Turing. Who was von Neumann ?!>>DFB: We’ve done a lot about Turing; we’ve

done a fair bit about Babbage, in the generation earlier. In fact a lot of

people, I guess, in the English-speaking world would regard Turing, in some way, as

being the Father of Computing. He came up with this very important result, in the

mid-thirties, about what was computable, and as we now know. to his credit. he

wasn’t at all afraid to burn himself with a soldering iron and try to create

hardware. Which he did at Bletchley Park during the war. So, yes, if he’s the Father,

if Babbage was the Grandfather and if Ada, Countess of Lovelace, was the Great-Aunt, then who on earth was John von Neumann?

And why is he mentioned alongside Turing? Well right at the outset let’s say John

von Neumann was the impossibly talented, impossibly charismatic, very wealthy,

Uncle to computing. It was he that, in the mid-1940s, in a way, made it all happen by

the force of his own personality. And kept it, not just in an enclosed

community, but encouraged all those who wanted to build general-purpose

computers to come along to this Summer School and do it. But that really, I guess,

is in the future. It’s where we’ve got to get to in the von Neumann story. But, yes,

you’re quite right Sean, to say that one of the first phrases that almost any

computer scientist hears about is The von Neumann Architecture for computers,

which to a large extent we still follow even now – all of, what 60-70 years later.

So, we’ve mentioned EDSAC before, we’ll be coming back to mention this very

important early computer again. But it is a von Neumann machine and all it’s

saying is [that] it’s very simple to build a computer. You need a Store, or Memory as

it’s more commonly called nowadays, to hold your instructions and your data.

You need a control unit, often called a CPU now. And

you need an arithmetic capability – the ALU, the arithmetic and logic unit. Again

many of you will know in modern chips those two are often combined into what

we just call “the CPU chip” nowadays. And you need input devices of various sorts

leading back to people’s Teletypes [with] input/output devices for backup storage,

disk and so on. You need to be able to do input and output. So there it is. It’s

just one, two, three, four, five boxes – that is the von Neumann architecture. And it’s

very, very similar today. There was a big debate at the time about that Store that

Memory. Shouldn’t you, for safety’s sake, put the instructions of your program in

a different sort of memory to your data? Wouldn’t it be safer to do that and

better in some ways? On the other hand, clearly, if you’ve got a good memory

technology, that works, the temptation might be just to put them [instructions and data]

in separate areas of that same technology and try and take some sort of precaution about

them not interfering with each other. In EDSAC the only way to get into a

subroutine and get back out of it again was to over-write part of your program

instructions! Let’s just return back to this incredible character John von

Neumann. How does he fit in alongside Alan Turing? Well, like I said, he’s the

older, impossibly talented, Uncle. Did Turing and von Neumann know each other?

Oh yes they did! They were both, basically, trained as mathematicians. Von Neumann –

it’s hard to know where to begin and where to end. You can’t exaggerate enough

about how good he was. He was Hungarian and his

Hungarian name – where they give surnames first – was, I think, Neumann Janos.

His father was very, very wealthy and when one was quite young the family

became ennobled in Hungary. [They] became basically at the level of Baron –

hereditary Baron I think over here. Janos was very talented; he was

a childhood prodigy. He could divide one eight-digit number by another eight

digit-number in a fraction of a second when he was aged about 6. He loved

history; he was a multi-talented polymath he easily came top of the class he

effortlessly took in detail. And that’s the first thing that all of his

mathematics contemporaries said about him was his sheer speed of picking up

new ideas and seeing the ramifications of them. So he was notorious even as a

teenager, and as a maths undergrad. He did his early education I think up to PhD

level in Budapest. He almost naturally ended up at a place we mentioned before

in connection with Godel and Hilbert. I’m talking, of course, about Gottingen

University in Germany. So Neumann Janos makes the journey, via a PhD, to becoming

effectively the research assistant to the superstar David Hilbert at

Gottingen. But because his family had been ennobled he’s not Janos Neumann

any more – Neumann Janos sorry – he’s Johann von Neumann

– impossibly talented. Hilbert his supervisor, at a seminar given by Johnny, John, Johann (!) asked who his tailor was.

Because his impossibly smart pinstripe suit was just a complete

knockout. So he was a legend almost the moment he got there and did some

fabulously important work there. It was obvious that for somebody of his talents

he was going to get a full Professorship very quickly indeed. I think he became

impatient, waiting for it to happen at any German university. So, in the late thirties

– ’37-’38? Somewhere around there anyway – Anyway he got an offer from Princeton, in

New Jersey. And that was, I don’t know, very timely. It all fitted in together

very well. As part of his tours of Europe, giving seminars, and on his way to

Princeton, I think he met Turing in the mid-thirties in Cambridge. Because he

gave seminars there and I think a lot of mutual respect grew

up. I mean, obviously, Turing being in awe of von Neumann wouldn’t be so exceptional.

But after that 1936 paper of Turing’s, about decidability, following on from

Godel and all that, von Neumann rated Turing – there was no question about that.

This was evidenced by the fact, you will recall – those of you who have seen my previous

videos – that Turing also took a sabbatical and worked with Alonzo Church at Princeton. Well, von Neumann was there by that time.

Von Neumann was such a superstar, they not only made him [full] Professor of Mathematics at

Princeton, at an absurdly early age – probably about 30, something like that.

But some of you will recall, right next door to Princeton – about a mile and a

half across the meadows – is the Institute for Advanced Study which had been

endowed in the early 1930s by a multi-millionaire. And this really was

the ultimate Club to be invited to join. You’ve got to be of the quality of

Einstein, who accepted the invitation. Hermann Weyl one of the founders of

quantum mechanics; Godel – we know about Godel. Godel was invited to just come

to IAS – Institute for Advanced Study at Princeton – stay as long as you like. Yes,

you’re a Professor; everything found: food accommodation; the lot. All we want is to

have the greatest minds here. Von Neumann was offered a professorship in that

community, I think at age 35 – maybe slightly younger. Unbelievably young. He’d

hardly been at Princeton a year or two, as an ordinary mathematics professor.

Everybody thought “This is a truly phenomenal person”. He reminded many

people of absolute superstars like Newton, Gauss, Euler, Einstein, Hilbert

himself. Even at early to mid- thirties they could see that potential

in him, there. So, yes, Turing visits Princeton in 1938 – worked with Alonzo

Church – but also of course had frequent interactions at seminars with von Neumann

And then came the big question, if you recall, for Alan Turing: “Should I

return to England and do my patriotic duty?” According to Andrew Hodges’

definitive biography of Turing, Turing’s father was all for Turing

staying in Princeton, you know: “Keep out of the war; get a prestigious mathematics

job”. And that was underlined by the fact that von Neumann offered Turing a job. He

basically said: “Turing, would you like to be my research assistant at the

Institute for Advanced Study?” Now Turing could see, straight away, that would

just make you as a mathematician. You were invited, by the great von Neumann,

to be his research assistant, at IAS! Only problem was, I think, first of all, I think

Alan Turing did feel a certain patriotism in wanting to come home and

do his bit. There was also the worry that at that time von Neumann had not

properly got into computing – he’d not turned his considerable talents to

considering it – and for the research assistantship he wanted Alan Turing to

do quantum mechanics, another of von Neumann’s great loves. And I don’t think

Turing was keen on that because he knew from experience, at Cambridge, where

he’d tried doing mathematical physics, it really wasn’t his arena at all. So he

politely declined with great thanks, came back to England and the rest, if you like,

is more or less history. Now there’s “Johnny”, as he’d now become. Johann von

Neumann speaker of four, five languages including Italian and English, once he’d

transferred from Gottingen to Princeton, wanted to become the all-American genius.

So he was … on more formal occasions he was just John von Neumann but to his

friends he was “Johnny”. You can’t exaggerate enough! I mean, his wife said: “He

can count everything but calories”. He was fond of food and drink; the champagne

parties! the glitz! the glamour! the girlfriends! Good old Johnny – he

absolutely was the antithesis of the shy mathematician. He was all-

encompassing and everybody who met him was just stunned by how he could see

his way through problems in no time flat. And just do

impossible things. And so he was there, in a very ,very nice position, Institute for

Advanced Study, even before the Americans joined the war. But he stayed

there throughout the war. But being who he was he was endlessly in demand to be

a consultant and, most famously, along with people like J Robert Oppenheimer he

was one of the consultants employed on the Manhattan Project – the atomic bomb and

the hydrogen bomb. But he was used by the Army, the US Navy, the Air Force

– everybody wanted Johnny as their consultant. And this even included, as the

war developed, the fact that one of the earliest computers which we have

mentioned in a previous video was University of Pennsylvania, Moore School

of Engineering. They helped develop the ENIAC which, if you remember, was a vacuum

tube computer running on decimal arithmetic and initially devoted to

gunnery trajectory calculations. It was a bit late in 1946 to be in direct

strategic use in gunnery, Actually Johnny, who by this time *of course* was a

consultant to the Moore School, devised a way, I think, to turn ENIAC into being a

general purpose computer, although it wasn’t a very efficient one, and I

believe he used it for some calculations relevant to the atomic bomb and all that

kind of thing. The natural question arising with everybody at the end of the

war went like this: “We all know – or even though it’s top-secret at Bletchley Park

we have heard gentle rumours – that computers are being developed all over

the place and you’ve always got to say: Are they special-purpose? Are they

general-purpose? Are they binary? Are they decimal? Are they fully

electronic? Or are they electromechanical? And literally there must have been at

least a dozen machines around that satisfied some of these criteria. If you

ask Germans about “Who’s the father of computing?”, they’ll say “Konrad Zuse”, He

developed electromechanical machines that were Turing-complete and

calculated things. But they never got beyond electromechanical. You get on to

electronic ones – you get Atanasoff and Berry’s electronic, valve driven,

thing – special purpose though! Could solve certain differential equations. And even

Tommy Flowers, and Colossus, we know, special purpose: could decrypt Tunny traffic –

Lorenz cipher as it later become known as. So, you’ve got everything

happening that if it’s general-purpose it’s not yet electronic; if it’s

special-purpose it is electronic but we want it to be general-purpose. So, at the

end of the war was the perfect time to get everybody together and say: “Look, now

that the war’s over we all want to find the way to do it correctly. To build

general-purpose, probably binary-based because they’re more reliable, all

electronic digital computers. How do we do it”? And who better to lead the charge and run a Summer School, and be associated

with it, than uncle Johnny of course! And the Moore School at Pennsylvania, to their

great credit, did this. They decided that the successor to ENIAC would be a thing

called EDVAC. They said, yeah, it’s going to take us three or four years to do

this but in the meantime here we are, 1946, why not all of you, all over the

world, who are interested in the quest to build general-purpose, all electronic,

digital computers, probably based on the binary system … we’ll hold a Summer School

in the Moore School of Engineering, 1946, welcome everybody. Did Turing go to it?

He was, I believe, at that time at the National Physical Laboratory in the UK

No he didn’t. Hated conferences did Turing. He wasn’t

a clubbable character. He couldn’t stand small-talk. Classic shy mathematician not

at all like von Neumann, right? So Turing wasn’t, if you

like, the UK representative there. And one wonders whether he would also been held

back by Bletchley Park and the Official Secrets Act because he’d only just left a

few years before. The representative from the UK was somebody who was Turing’s exact

contemporary. They had both done mathematics in the early 1930s, at

Cambridge. They had both got first-class degrees. Did they get on? Not very well!

But who’s this other person? His name is Maurice Wilkes – Maurice Vincent Wilkes – and

by the vagaries of job allocations around about World War II, he didn’t end

up at Bletchley Park, did Maurice, he ended up in radar. But he knew enough about

mathematics and electronics to be in a good position to do, or make, a von

Neumann machine in the UK in the period from about 1946 onwards. But we’d better

stop there because Maurice and his EDSAC is an extra story.

## 100 comments found