Embarrassingly Parallel Computing - Steve Furber

Parker Dillmann
Welcome to circuit break from MacroFab. A weekly show about all things engineering, DIY projects, manufacturing, industry news and ARM Processors. We're your hosts, electrical engineers, Parker Dillmann and Stephen Kraig. This is episode 437.

Stephen Kraig
This week, our guest is Steve Ferber. Steve is professor emeritus of computer engineering at the University of Manchester in the UK. He is best known for his work on the BBC microcomputer and the first ARM processor at Acorn Computers in the 19 eighties, and more recently for the Spinnaker project that has delivered a machine with a 1000000 ARM Processors for brain modeling applications.

Parker Dillmann
So thank you so much, Steve, for joining us on the podcast this week.

Steve Furber
It's a pleasure.

Parker Dillmann
So before we, like, completely jump into it, Steve, what what like, how how did you get started, like, just I mean, there there's a lot of stuff there to unpack of of your history of starting with the BBC microcomputer and arm and beginning with arm processors. Because like nowadays, there's an ARM Processor in practically almost every single device nowadays. Let's let's wind the clock all the way back and talk about, like, your early career and how did you like get started with like the BBC micro, computer? Oh, what actually actually, I bet you most of our listeners don't even know what the BBC microcomputer is actually as well.

Steve Furber
So my story really starts at Cambridge University in the UK. I went to Cambridge, to study mathematics as an undergraduate and in in in was there for 4 years, studying mathematics. And then because I had a long interest in flying, I got the opportunity to take a PhD in aerodynamics in the engineering department. So I became a very sort of theoretical engineer. It was still mainly math, so a bit of experimentation.

Steve Furber
And my interest in flying sort of manifest in terms of I did some model flying in my earlier years, and I joined the university glider club in Southern Airfield every Wednesday afternoon for a year, amassing a total of about 54 flying minutes over that year. And through all of that, III decided that that actually might be simpler to get involved in in in, flight simulation. That might be a more productive way of of doing this flying thing. They're actually trying to get near real airplanes, which seem to occupy a lot of time and or money. And around the time I was thinking about this, I I got to hear about a bunch of students who were starting a new student society called the Cambridge University Processor Group.

Steve Furber
And this was a student who built computers for fun. And we're we're talking here in late seventies. And I thought, if I'm interested in building some kind of flight simulator, then knowing a bit about computers is an obvious place to start. So I got involved in the process to group. I started building computers, which in those days involved very scary things like ordering microprocessors from shops in California using credit cards, which for imprecunious students were very frightening things to do.

Steve Furber
And I I bought, a 2650. Anybody remember the Signetix 2650? That was an 8 bit microprocessor. Ordered some memory chips, and started putting these things together to form a computer. Now I such should say that in in the Cambridge University Processor Group, the real men built their computers out of TTL, and it's only the wimps like me who who went to this newfangled microprocessor stuff.

Parker Dillmann
So so the p the wimps were using integrated circuits. I guess, more integrated versus TTL. But, no, I I've never heard of the Syngetix 2650 I'm looking at right now.

Steve Furber
It didn't it didn't stay around long. Anyway, so I so I got involved in the student society and and and built machines that started working and, got on with my engineering, PhD and and then I had a research fellowship sponsored by Rolls Royce after that. And while I was doing that, I was approached by Herman Houser, who told me he was thinking of starting a consultancy in the microprocessor business and was looking for a few willing people to do odd jobs for him. And that's how I got drawn into what later became Acorn Computers. When I got involved, it was very embryonic, and it was called Cambridge Processor Unit Limited or CPU Limited.

Steve Furber
And I remember the first job we did was developing, a microprocessor based controller for a fruit machine, which I think in the US, you call a 1 armed bandit. Yeah? Is that the right terminology? Up until the late seventies, these were largely electromechanical devices, and they were just beginning to think about switching them to microprocessor control. And and and we developed this control system.

Steve Furber
Then 1 of the other guys that was involved in that started thinking about designing a hobbyist microprocessor kit. And in the UK, Clive Sinclair and and and Chris Curry, who was involved with Herman, used this National Semiconductor thing called the MK 14, which they sold as a hobby's kit. And and Sophie Wilson looked at this and said, I can do better than that. And and and produce this single board. You know, this the kind of computers that were around then, which were 8 bit micros with hexadecimal displays and hexadecimal keypads.

Steve Furber
And, usually, you bought these as a kit, and you'd assembled them yourselves and put them together. And and and that thing that Sophie designed was the Acorns became the Acorn System 1. So Acorn was originally a trading name, not rather than the company name. Only later did the company rename itself to follow that trading name. So that's how I got drawn in through through this fruit machine controller development and and and building stuff for fun.

Steve Furber
I was still primarily working at the university in my research fellowship. But I'd I'd I'd started building computers firstly as a hobby, and then I was using these computers in my aerodynamics research to to monitor these these fluid flow systems experiments that I was building. And it was a kind of deal with Acorn that I designed stuff for them. They give me bits to play with. I'd use those bits to design more stuff and and and to give the designs to them.

Steve Furber
It was a very informal relationship between us. Then Acorn moved on from the system 1, and it built these rack based systems that were really fairly professional microprocessor controllers, with floppy disk drives and things, and they they were very expensive. And Chris Curry decided they wanted something cheaper to get it back into the hobbyist market, and they developed the Acorn Atom. And there are lots of stories I could tell you about the Atom. I did not have a hand in developing the Atom.

Steve Furber
But, it was sold as a kit, and we had all sorts of rejects sent back by people who built these kits. 1 said, you know, I understand that solder is hot and heat is bad for integrated circuits. So I very clear if he glued all these chips into circuit board, and it still didn't work. That was the the last product Acorn sold as a kid. The the Atom, became fairly well known in the UK.

Steve Furber
And, in fact, it was responsible for, some leading technology. 1 thing that came out of the Cambridge Computer Lab was a local area network technology that was built into the atom called Ethernet, so you could network classrooms of of atoms together. But then we heard about the BBC's interest in in developing, in having a computer around which they could base a series of TV programs. And they opened up the bidding for this. And and Acorn put in a proposal, which was basically based on the machine I'd built the year before at home for monitoring my fluid dynamics experiments.

Steve Furber
And we turned this into a circuit and and and into what also it be became the BBC Micro because Acorn got the contract. And there was a famous week when the beginning of the week, we had nothing. And at the end of the week, the BBC were coming going to come and see what we had, and we put the prototype together. And and and by the Friday, when the BBC arrived, indeed, the hardware was just working, and the software was running on it. And, they were suitably impressed by Acorn's ability to move things forward very quickly that Acorn got the contract.

Stephen Kraig
Wait. Wait. You're saying you developed a computer within a week, or I should say built a computer within a week?

Steve Furber
I as I say, I we have the concept at the beginning of the week. I built this machine at home the previous year, which embodied quite a lot of the principles in this concept. But we haven't started building the actual machine, which was, if you like, based on the concept, but running twice as fast. And and so we did we did the prototype development from scratch, you know, wire wrapping the circuit board. The first the first stage was actually drawing the circuit diagram on a piece of paper, which I did.

Steve Furber
And then wire wrapping the prototype, and then debugging it, and then getting the software up and running. Yeah. That took a week.

Parker Dillmann
So this is oh, no. Let's back up a little bit. So a fruit machine, by the way, is is we would call that a slot machine in the United States so it's a it's a gambling machine you said 1 armed bandit and I'm like I think that's a slot machine

Stephen Kraig
yeah okay

Parker Dillmann
and I want to talk about so you built a computer at home this is the year prior to run your fluid dynamics experiments. You gotta talk about that.

Steve Furber
Yeah. To say I built it to run those experiments will probably be slightly misleading, and I built it for fun, because I liked building digital electronic circuits. And having built it, I then used it, to basically do data logging and and simple things, linked to my fluid dynamics experiments. And in the engineering department, we had we had computers. I think I think the small 1 we had was a computer automation LSI 4.

Steve Furber
If that means anything to anybody, I'm not sure these days. But it was a fairly big and heavy and immobile thing. So having, you know, home built rack of microprocessor based boards that that I've developed was just a more convenient way of interfacing into these sensors in the experiments. These were doing things like measuring pressure at different points in in an axial flow compressor stage.

Stephen Kraig
You know, actually so I've I've I've 1 question that's gonna rewind the clock even a little bit more. You had mentioned your your schooling. You you originally went for to school for maths and then got your PhD in AeroAeronautical, was it? Yeah. Aeronautical.

Stephen Kraig
What what I'm curious about is had was was electronics involved at all in that? Or did you get all of your electronics training from the processor group at Cambridge?

Steve Furber
I I pretty much got all the training from the processor group. There was there was a little bit of computing in the, in the maths degree that I took. I think that was that was using another computer nobody's ever heard of called a modular 1 computer. I think running a language called Focal. Does that sound right?

Steve Furber
I don't know. I do remember that it it used Hewlett Packard storage tubes. Right? So the computer were right on the display. And then when you wanted, you had to raise the whole display and start again.

Steve Furber
So they were basically a a storage oscilloscope technology. But yeah. So so all all my digital electronics, I learned through building stuff in the processor group. And, you know, friends in the processor group who helped debug debug things. There were people there who knew a lot more than I did about putting chips together to build systems.

Stephen Kraig
Oh, that's fascinating.

Steve Furber
The the technology we used then was was was a wiring pen. So you basically got these preformed circuit boards, and you put chip sockets into the circuit boards and, you know, soldered them at the corners to hold them in place. But then you made the connections with a wiring pen and wrapping wires around combs and then around pins. And when you soldered them, the insulation on the wire melted and you had a connection. Right.

Steve Furber
Right.

Stephen Kraig
Which which makes for a for a good a good circuit connection, but it must take forever to build a a sizable computer.

Steve Furber
Yeah. I mean, you're you're you're making connections 1 at a time. I mean, it's it's like wire wrapping.

Parker Dillmann
Right.

Steve Furber
But quite a lot cheaper to implement, and and you don't end up with a board that's got, you know, 2 inch long pins sticking out the back. But it's a it's a more compact format. But it's debuggable and repairable, and you can take wires off and Right.

Stephen Kraig
It's it's rapidly prototypable.

Steve Furber
Yeah. Yeah. It's a rapid prototyping system, basically.

Parker Dillmann
So what were what were the compared to, like, nowadays, what were the challenges back then when this kind of technology was just starting to kick off? What was different now then than now?

Steve Furber
I think 1 1 major difference was that the the speeds that were being used in those boards were in the megahertz region. And and the there's some manual wiring and running lots of wires through the same combs, so very close to each other. That all worked at the megahertz range, and it clearly would not work with the kind of gigahertz frequencies that we see today. Today's technology is much more demanding in terms of signal integrity running around the system. In those days, you you really could I mean, you could wire chips together in patch boards and expect them to work at the speeds that were being used then.

Steve Furber
You wouldn't stand a chance these days of of doing that kind of thing. What was challenging then was was coping with some of the higher frequency stuff. The clock I used in my system, I think, was 16 megahertz. And getting that crystal oscillator to work reliably required a fair amount of help from my friends, because we didn't have the tools. We, you know, we didn't have oscilloscopes and things that you need to to really get crystal oscillators working properly.

Parker Dillmann
Yeah. You just kinda had to bank on people who have done it before and had that experience, because you couldn't see what you were doing really.

Steve Furber
That's right. Yeah. Yeah. I mean, the may the major debugging tool in those days was was a logic probe. I don't know if you've come across those, but a logic probe, you basically have a couple of clips that you clip across the 5 volt rail, because everything was 5 volts.

Steve Furber
And then the thing had a pin at the front and a green LED and a red LED, and the LED would light up to tell you if the signal was high or low. And and, you know, so you tended to build circuits that you could run and then stop. When you stop them, you could then go measure all the signal levels and see if they were right.

Parker Dillmann
Yeah. Imagine trying to see a LED blink at 16 megahertz. Probably wasn't working.

Steve Furber
Yeah. As I say, you you needed to get the thing into a static condition in order to sensibly use the logic probe.

Parker Dillmann
0III can't imagine trying to because, like, nowadays, you use with, like, a digital logic analyzer, which but can record all that over time, and then you can go back and look at the data. Because you're you're talking about, basically, that, but you'd have to run per clock cycle, and then read all your data bus, and write that all down and then, you know, iterate the next cycle and then measure all that again.

Steve Furber
Yeah. I I remember 1 occasion where III coded a sort of simple screen operating system and and put it in a apron, And and there was a bug, and it was running. You entered the first command, press return, and then something went wrong. Debugging that was exciting because after you'd entered the command and push return, what your brain system did was it scrolled the screen. Okay.

Steve Furber
And scrolling the screen is basically copying every byte representing a pixel on the screen up 1 location. And then you had to sit there with a single step button, pushing this single step button. I think it took me about 45 minutes just pushing this button as fast as I could to get through the screen scroll before it got to the point that I was interested in where the bug came up.

Stephen Kraig
That is fun.

Parker Dillmann
So after the, you know, working on the BBC micro, you started working on Arm. So how did that transition happen?

Steve Furber
The the story there has has been told a number of times. I mean, the the BBC Micro was a huge success for Acorn. And in the early negotiations with the with the BBC predicted, you know, sales of about 12, 000 BBC Micros. And and that turned from 12, 000 into 1 and a half 1000000, in in 2 years. So everybody's estimate of of of the degree of interest in home computing was completely wrong.

Parker Dillmann
So, actually, so from 12, 000 so, like, BBC is, like, we're gonna build 12, 000 of these over 2 years to you said 1 point how many million?

Steve Furber
1, 100, 000.

Parker Dillmann
How was trying to get get those, like, I can't imagine, like, the logistical jump to make that work Cause you're you're planning 12, like, yo, you're gonna get 12, 000 microcontrollers. Oh, now we actually need 1 point, you know, 2, 000, 000.

Steve Furber
Yeah. It was it was an interesting challenge to Acorn's manufacturing people. I mean, all the manufacturers subcontracted. Acorn itself didn't didn't build things. So what it had to do was just bring in more subcontractors to to have more assembly lines producing the product.

Steve Furber
But the BBC Micro was a big success in in that sense. And that sort of launched Acorn into quite a prominent position, particularly in the UK and some other regions. So they're quite successful in in Australia, New Zealand, Netherlands, Germany. The attempt to sell the BBC Micro in in the US, was not a big success. In fact, it was rather an expensive failure.

Steve Furber
That's a different story. As far as the the development team was concerned, at that point, the world was moving from 8 bit microprocessors to 16 bit microprocessors. And and we at Acorn were looking at all the 16 bit microprocessors built in mechanism built in mechanism to support a second processor. So it's very easy to hang, any any microprocessor you like to build a little system with the microprocessor and some memory hanging into the 2nd processor port and and then see what it would do. And we did that with a lot of microprocessors, most of the ones that were around at the time.

Steve Furber
And we didn't like any of them. And there were 2 main reasons we didn't like them. 1 was we used extensive real time, performance in the BBC micro. 6502 has quite good real time response, And all of these 16 bit microprocessors had much worse real time response. I mean, the the figure I remember is the National Semiconductor 32016 had a memory memory divide instruction, which took 360 clock cycles to complete, and and the clock was 6 megahertz.

Steve Furber
So you're talking 60 microseconds. And during that time, it was not interruptible. So if you were trying to handle a double density floppy disk stream data stream, which generates a byte every 32 microseconds, you couldn't without paying for more hardware. The real time response was the first thing. And the second thing is that we determined by then that the main thing that that affected how fast a processor went was the the the processor's ability to use memory bandwidth.

Steve Furber
And the memory was the expensive component in any, small microprocessor system. If you paid for the memory, what you really wanted was a processor that would use all the memory bandwidth. And none of the processors that were available from the big semiconductor manufacturers at the time would do this. This was partly because, a lot of them had architectures inspired by the 19 seventies mini computers. So they had very complex instruction sets and they just couldn't keep up with the memory.

Steve Furber
So we were sort of frustrated by these 2 factors when our boss at ACORN, Hermann Hauser, started dropping papers on our desk about the, the risk work at Berkeley and Stanford, and how a postgraduate class had designed a microprocessor that was very competitive in in a year. And and that got us thinking about, you know, maybe if you could do it with a postgraduate class, we could we could design our own microprocessor. So that's where the Arm idea came from. We visited National Semiconductor's design facility in Haifa in Israel, where they were developing the first 2 0 16, and they were on revg orh. I can't remember which.

Steve Furber
And they still haven't got the bugs out. And, again, this was a result of the complexity of the instruction set. So this this new idea from Berkeley of of reduced instruction set computing, really simplifying down the instruction set to its basics, and using the silicon resource for stuff that gave you more performance than complex instructions, such as pipelining the processor. That kind of appealed to us. And, Sophie Wilson started playing with set architecture, and and and then I took that and turned that into a microarchitecture.

Steve Furber
We gave it to our very small VLSI design team to put together. And 18 months later, from from the start, much to our surprise, we had a processor that worked really well. None of us had any experience in processor design before that. We were just you know, we before the armed, we all thought processor design had a kind of mystique to it, and only these big semiconductor houses could do it. At the end of the armed development, we realized it's it's just another piece of logic like everything else.

Steve Furber
And and and if you go about it in a sensible way, you can do it with a with a very small resource. You could then. It it was a particular sort of peak of simplicity, I think. The ARM used fewer transistors than some of the 8 bit microprocessors around at the time, and we pushed very hard to keep it simple.

Stephen Kraig
Did I hear that right? You said the first processor that came back functioned and and worked well? Yeah. Wow. Were there any bugs at all?

Steve Furber
There was 1 very minor bug, in an obscure corner of the barrel shifter operation. So 1 of the barrel shifter functions didn't work as designed. A piece of metal there had ended up in the wrong place. But apart from that, the whole process had worked exactly as we had expected, in in including performance. You know?

Steve Furber
It it what I said earlier about memory bandwidth, that was the reason why Acorn went straight from 8 to 32 bits. Okay? Because the 32 bit bus gives you twice the bandwidth of the 16 bit bus if you can use it. So all was 32 bits. And but it was very much inspired by the risk work.

Steve Furber
You know, we didn't implement the big register windows function that on the Berkeley risk machines because we thought it was too expensive for our application domain. But we did take most of the other risk ideas, the load store architecture, you know, the reg the regular, largest register file, separation of of load store instructions from data processing instructions, all of that we, we we took from the the the risk philosophy. With a few optimizations because, you know, we kinda thought the risk examples were academic prototypes. We were a commercial company, so we wanted things a bit tighter, a bit less slack in the code density. And Sophie Wilson, who who did the ISA work, of course, had written all Acorns BBC BASIC interpreters, so she had a very good understanding of what was required to support a high level language.

Steve Furber
So the arm had some some features which which you might not think of as very risk, such as the load store multiple instructions, which would load or or store the any subset of the register set in the single instruction. And and and, you know, that was clearly useful for procedure entry and exit and stuff like that.

Parker Dillmann
So what was the first high level language that was was for the compiler for this microcontroller? What was it c, or is it was it before then?

Steve Furber
I think the first the very first high ish level language was BASIC.

Parker Dillmann
Okay. It was BASIC?

Steve Furber
I mean, the BASIC interpreter was kinda written before we got the silicon, because we did we did have emulators. So we built software emulators so we could develop code and debug it before we had silicon. I think the first compiled language probably was c. There were various people got very interested in importing their favorite language to the ARM when we we had the early prototype systems available.

Stephen Kraig
What was the first commercially available arm after you've done this prototype?

Steve Furber
It depends what you mean by commercially available. Right? We did sell the old development system based on that very first Arm silicon, but that was sold in, I guess, tens or or may maybe the order of 100 units. That was the second process. This is the BBC Micro.

Steve Furber
So you you got a box that you could plug into the BBC Micro, and that box had a long processor and and 1, or I think more typically 4 megabytes of memory in it. So the first 1 that was sold in in significant volume was was arm 2, which process shrunk. So arm 1 was 3 micron CMOS. Arm 2 was 2 micron CMOS. And arm 2 was the processor around which the first Acorn Archimedes products were developed.

Steve Furber
And the Archimedes sold in in not huge, but significant 50, 000 units a year or something like that. So so it's it's not huge numbers, but but that those were the first volume sales. And the arm 2 was also licensed to VLSI Technology, who who were the silicon manufacturer, and they made it available to to third parties. And, you know, Radius, I think, built a graphics accelerator for the Apple 2 that was based around the arm 2.

Parker Dillmann
Yeah. The other in my computer architecture class, we actually built a microcontroller that the peripherals for a microcontroller with an arm 2 core. Yeah. So we did all, like, the VLSI design around that arm 2 core. So that's kinda cool.

Steve Furber
Yeah. I mean, was the arm 2 core a hardcore? Or or or or or or was that also I think it was synthesized on to FPGA?

Parker Dillmann
It was synthesized.

Steve Furber
Sorry. It was?

Parker Dillmann
It it was synthesized.

Steve Furber
Synthesized. Right. Okay. Yeah.

Parker Dillmann
It was all on a on a Spartan FPGA, I think, if I recall. Okay. Yeah. So arm ARM stand used to stand for, Acorn risk machine. When when did they change it to, advanced risk machines?

Parker Dillmann
When was when did that happen?

Steve Furber
So that happened in 1990 when this this was actually must have been very shortly, a week or 2 after I left Acorn to take up my position at Manchester. Apple came knocking on Acorn's door saying that, they quite like to use the arm in the Newton products that they were developing, but they'd be more comfortable if it wasn't owned by a competitor. So how about setting up the Arm activity as a joint venture? And they were pushing on an open door because Acorn was quite keen to shed the cost of supporting ARM development. Its business wasn't growing fast enough to really fund that.

Steve Furber
And Advanced Risk Machines Limited was set up as a joint venture between Acorn, Apple, and VLSI Technology had, I think, had a minor role in that, but but but they were the 3rd founding company. And it was at that point when Arm was changed from Acorn Risk Machine to Advanced Risk Machine.

Parker Dillmann
How'd you feel when you heard that?

Steve Furber
III thought that was fine. I think I think they, know, they obviously couldn't keep the the Acorn name if they were gonna move us into a separate company.

Parker Dillmann
I guess it's it's different if it was called, like, the Steve risk computer or something like that.

Steve Furber
Do you

Parker Dillmann
think why do you think Arm is everywhere now nowadays?

Steve Furber
So People ask about the success of Arm. And, you know, as with all of these things, there's quite a lot of serendipity comes into the story. I think Arm was fortunate in that, it was established as a separate company at the beginning of the nineties, just when people were beginning to talk about systems on chip. So through the eighties, a microprocessor was a chip. Okay?

Steve Furber
And then in the nineties, the number of transistors that Moore's Law was delivering was beginning to get to the point where you could put a microprocessor and, quite a lot of the rest of the system on the same chip. And because ARM was small and simple, it occupied less of the real estate of an early SoC than did its competitors. So there's more room for the other stuff. And I think that made a difference. Also, Arm turned out remarkably low power.

Steve Furber
Again, that can be attributed to its simplicity. But if you're building an SoC, then you want a processor that doesn't burn too much of the power budget, because, again, you've got more power budget for everything else. And and and so I think the the the the those technical aspects were significant in the very early days, and that enabled ARM to get the Nokia business. Now Nokia in the nineties was the biggest mobile phone handset manufacturer in the world, and they were basically setting the pace for GSM, which is the early digital mobile phones. And the fact that Arm got into Nokia meant they got into nearly everything else as well, in the mobile phone area.

Steve Furber
And and that was huge. But also, a factor in Arm's success has been the business model that Robin Saxby developed when he came in as CEO of Arm, when it was set up as Advanced Risk Machines. And and and that is this IP licensing model. And and and that basically means that that that ARM is available to anybody, you know, at at at cost. But if if you base that on a royalty only basis, then royalties are are quite tough for cash flow, because the money comes little and late with royalties.

Steve Furber
So Saxe, we developed this model where you pay an upfront entry fee to get your license, and then you pay downstream royalties as well. And, of course, the upfront license fee is very good for cash flow. It's it's big, and it's early. And that kept on going and building without any external financial input, right through their early years. And so the so there's the technology in the at least in the first instance, there's the business model.

Steve Furber
And then, there's what Arm did, particularly in the nineties, by way of road mapping, which was not all about making bigger and faster processors. It was about understanding that when you put the processor into a system on chip, debugging becomes a whole new ballgame. Right? The model, through to the end of the eighties, was the in circuit emulator. You know, you pull your processor out of the circuit board, plug it in in circuit emulator, and then you can see what's happening and debug your software.

Steve Furber
Once the processor is embedded in an SoC, that model breaks, because you need to build a new in circuit emulator for every chip, and that's just not gonna work. So they they developed the on chip debug technology. So they gave you effectively ICE functionality around the processor on the chip. And they really focused on making the arm easy to use. We we had this sort of view that, you know, that clever designers like to build stuff to impress their clever designer friends.

Steve Furber
But if you want to address a big market, actually, you want to make stuff simple. And you have you have to worry about the large teams of designers out there who perhaps aren't quite as clever as your friends, and build something that that, you know, the engineer on the job can can you can work with. And and Arm, I think, did a very good job of focusing on that aspect of the business, of making the Arm easy to use in very complex SOCs.

Parker Dillmann
So looking back, did you ever or did you ever anticipate, or were you, like, blown away by how popular the BBC Micro and then what became Arm?

Steve Furber
We were certainly blown away by the pop popularity of the BBC Micro, you know, which exceeded all predictions, as did, of course, a lot of other computers. This isn't unique to the BBC Micro. In the UK, Sinclair was building the zx8081, and those markets exploded too. So it wasn't just the BBC Micro. Everybody was surprised by interest in home computing and the way people could find applications.

Steve Furber
You know? So BBC Micros were used in large numbers just as dumb terminals attached to mainframe computers, for instance, because they were cheaper than the standard commercial terminals. So there are all sorts of surprises for people. And certainly with Arm, I don't think anybody in the early days could have predicted the success today. I mean, the scale is just mind boggling.

Steve Furber
I think that the last number I heard is that over 250, 000, 000, 000 ARM powered chips are being shipped by ARM's very large partner network. So you're talking about 40 chips for every human on the planet is arm powered. Nobody anticipated that, but the growth is just has been sort of rapid and exponential since ARM limited was formed in 1990.

Parker Dillmann
So let's, talk about something more recent, the Spinnaker project, which we touched on a little bit earlier. So can you explain what the Spinnaker project is?

Steve Furber
Yeah. Spinnaker is a machine that was conceived slightly over 20 years ago as a massively parallel computer optimized for brain modeling applications. Brains are are based on networks of biological neurons, and the way these neurons are assembled in in very large numbers makes for an embarrassingly parallel model if you try and compute it. And so Spinnaker was was intended to exploit that embarrassingly massive parallelism by building an embarrassingly massively parallel computer. And we set ourselves the goal early on of putting a 1000000 ARM processors into this machine.

Steve Furber
It was clear even from the outset that with a 1000000 ARM processors, you don't get close to the scale of the human brain, but you are in the region of of mouse brain scale, potentially. It's it's based on, in many ways, a relatively conventional, highly parallel computer. What's unconventional is firstly that the processors are all low end energy efficient embedded processors, rather than high end math processors that you find in typical HPC applications. And the second is the way we implemented the connectivity. The brain is hugely connected.

Steve Furber
Each neuron connects to many thousands of others. And so to implement a model on a machine, you need to replicate that degree of connectivity. And that required that we moved away from the typical point to point communication you find in a in a normal, parallel computer towards a multicast communication. So when the message is sent by 1 processor, it can be delivered to 100 or 1000 of others, and and and and but effectively, the communication path is a tree across the machine. And if you like, the key innovation in Spinnaker is how we implement the routing for those trees, because they're they're completely flexible.

Steve Furber
So you can define any tree structure, and therefore, you can model any part of the brain. Different parts of the brain have quite different connectivity diagrams. So we we virtualize the connectivity through this innovation on

Parker Dillmann
So were these cores actual, like, silicon cores, or was this all synthesized on a lot of different FPGAs? Like, what's the actual architecture like?

Steve Furber
It it it it's based on an ASIC. So, yes, the the the the ARM cores are synthesized in the standard ARM synthesis root way using ARM's preferred tool flow. But then they're they're implemented with 18 cores on a chip, on spinnaker 1, with this interconnect technology joining them together on the chip and and and forming the links to other chips. So the the core of Spinnaker is is a centimeter square ASIC with 18 arm cores on each ASIC.

Parker Dillmann
Okay.

Stephen Kraig
How does the user define these network trees in, the project?

Steve Furber
So the the the goal with with Spinnaker is is to enable regular users to use the machine without understanding any of this interconnect technology. And and users can describe the networks they want to simulate in a language called Pine, which simply stands for Python Neural Network. And that language was developed by collaborators at CNRS in Paris. And and you can run a Pine model on your laptop using a simulation engine such as Brian or 1 of several others. When you've developed your Pine model, if you want it to run faster, you can then simply port that model to Spinnaker, And the software stack that we developed at Manchester will take that Pyne model, allocate, processor resources to modeling the various neuron populations in the model, and it will compute the routing tables to make the necessary connections.

Steve Furber
For a standard user who's mainly interested in, if you like, the neuroscience and the network level, They don't need to know anything about what's happening at the low level in Spinnaker. The enthusiastic user who wants to develop, for example, their own neural model or their own learning rule, both of which are implemented in code, has to dig a bit deeper. But again, they can do that within an environment where the software stack handles all the routing. So I think very few of our users actually get directly involved in understanding how the routing works, and in driving it themselves. That's all handled by the software stack, and it's basically it's a problem of mapping 1 graph, which describes the neural network you want to simulate onto another graph, which describes the machine.

Steve Furber
Okay. And and the goal in the communications architecture is to is to virtualize communication, so that that mapping can happen as efficiently as possible for as many different problems as possible.

Stephen Kraig
Right. So the user's abstracted from having to go that deep.

Steve Furber
Yes. As I say, I suspect very few Spinnaker users have done anything hands on with the routing.

Parker Dillmann
So given the current rise in, like, neural nets and AI, what's your thought on, like, the current state of that kind of technology?

Steve Furber
It's been a very interesting parallel development because we've been working on Spinnaker for over 20 years now. And over those same 20 years, neural networks as used in connectionist AI, have come from nowhere to being absolutely the dominant technology. And there are similarities and differences between the networks that we run on Spinnaker and the networks that are used in AI. People tend to refer to AI networks as as second generation, and Spinnaker spiking networks as 3rd generation. So there are similarities and and quite fundamental differences.

Steve Furber
I think what's interesting is that going forward, people are beginning to see conversions. And some of the problems currently facing AI, such as the completely unsustainable energy demands of of current leading edge large language models might be addressed by using more brain like approaches of the sort, supported by platforms such as Spinnaker. If you look at those big AI networks, they're dense. So the matrices that describe the connections are dense matrices, and that's why they run very nicely on GPU's and and similar. If you look at the brain, nothing in the brain is is dense.

Steve Furber
Okay? All the all the connectivity in the brain is sparse, and the activity is sparse. And if you could capture that sparseness in an effective large language model, then you might see orders of magnitude reduction in the energy requirements to run that model. And that would be very important. I mean, we it hasn't been compellingly demonstrated yet, but I think we're not far off.

Stephen Kraig
So, just out of curiosity, outside of your groundbreaking technology, I'm I'm curious what your hobbies and other interests are.

Parker Dillmann
Yeah. So so I was while I was going through and building up these notes, Steve, I saw a picture on your Wikipedia page of you playing a bass guitar. So earlier you were talking about flying. How did you go from fly you went from flying or thinking about flying? Did you act oh, you spent you said an hour in a glider?

Steve Furber
I I spent a cumulative 54 minutes flying time over 1 year of standing on an airfield every Wednesday afternoon at the Cambridge University Glider Club and ending up in the air very little, which I decided in my 3rd year when I had my final exams was probably not the most efficient use of my time. So that was my 1 year of of real flying as it were.

Parker Dillmann
That's where you got into processing. But how'd you go into the bass guitar?

Steve Furber
Well, I've I've been playing guitar since my teens. Okay? And and III played in a in the Christian music group at at at university. There's what in the sideline is where I met my wife, so it had lots of benefits. They they also used to say in this music group, it wasn't a great music group, but it was a great marriage bureau.

Steve Furber
So I and, you know, III don't consider myself as sort of, you know, the top tier musician, that I'm a sort of I can technically hack my way through things on both 6 string and bass guitar. Mainly, I've played bass for the last 30 years, because that the the the church music groups I've been in, there's been more need for a bass than a regular guitar. But just recently, I've been playing a bit more 6 string because, for various reasons, we haven't had a keyboard player. And just playing bass doesn't really work for leading singing. You you need a bit more middle than that.

Steve Furber
Yeah. So, you know, me middle range music has has has has been a long term interest of mine and or or activity. And and and and I enjoy that because it's a completely different mental process from from what what you regularly do. I mean, if you if you play music in in a in a band or church group, you've got to sort of listen to the others, and and and fix in, and adapt, and work out what to do when somebody gets something wrong. Right?

Steve Furber
Which happens quite a lot at the level I play at. Just just patch things over, and it's a it's a very different mental process. Otherwise, in terms of hobbies, yeah, I guess I'm I'm not really done that much on the flying front for a long time now, not even with the models. I like walking. I go out for a bit of a walk every day to try and keep my art working for a bit longer.

Steve Furber
You know? It's a nice fall of hazards when you get to my age.

Stephen Kraig
So so I I like it. It it goes math, aeronautical engineering, fluid dynamics, digital processor, bass guitar. I'm

Steve Furber
III like it. That's that's a good sequence. Yeah. I feel like I've I've made to my hobby, my my day jobs 2 or 3 times in my career. I've sort of shifted focus.

Steve Furber
And then, you know, I I have no regrets about choosing maths for my first degree because it's a very adaptable subject. Right? You can use it almost anywhere. By the way, of course, in in the UK, we do maths in the plural. I know in the US, you tend to just do 1 math.

Stephen Kraig
Right. Right. Yeah. So actually, the shoehorn that do 1 other question we had was, advice for aspiring engineers. You said, you had turned your hobby into your career multiple times.

Stephen Kraig
So do you have any advice for future, engineers and computer scientists?

Steve Furber
Well, I I think the main advice I give to young people who are at the point of sort of deciding which degree subject to do or what career to choose Unless you have a true vacation, so you know you want to be a doctor, in which case, you know, you better do a medical degree. Right? That's that's pretty easy. But if you're not sure what you want to do, and most young people aren't at that stage, then think about what keeps the most doors open. You know, so doing a sensible core STEM subject, such as maths or physics or computer science, you know, you can turn your hand to most things with that with that kind of background.

Steve Furber
So don't specialize too soon is is is is probably my advice to people at that early career stage. And and and then you should choose a degree which interests you. Right? Because degrees are are hard work, they're supposed to be. You have to you have to find the motivation to keep going some of the time, and and that's easier if you're interested in what you're doing.

Parker Dillmann
Let's say computer science didn't work out. What would have been the backup career for you?

Steve Furber
I I think I could have stuck with my fluid dynamics. I could, I could have got somewhere as a, as an academic researcher in the fluid dynamics space, I think. The decision point I made there was the end of my research fellowship, was do I continue with the fluid dynamics, or do I go join Acorn? Was basically, the decision there. And Acorn's situation, having secured the BBC Micro contract, was so interesting that it was a fairly easy decision to to to go to Acorn.

Steve Furber
But I think had that not been, available, I would probably have continued further with fluid dynamics.

Parker Dillmann
And if you had, if you could go back in time to that period and tell yourself a piece of advice, what would it be?

Steve Furber
I think I I don't know. I think I think, you know, just find stuff that interests you, and and and and try and, you know, try and earn your living from stuff you find really interesting. And then it, you know, it it it feels less like work. If it's stuff you'd like to do. Would you would you still do this if nobody paid you?

Steve Furber
Okay. You need somebody to pay you to pay the bills, but if you didn't need that, would you still do this? And and, you know, with with with what I'm doing in computing, that's still true. You know, I'm now retired. I don't need anybody to pay me to do anything, but I'm still doing stuff because it's interesting.

Parker Dillmann
III view it as, would you do your day job if it all it gave you was lunch?

Stephen Kraig
Yeah. I like that.

Steve Furber
That's about right. And and I mean, there probably aren't that many people who can answer yes to that. But I I think if you can answer yes to that, you're very fortunate. And I always have been able to answer yes to that.

Stephen Kraig
You certainly made the right decision back back then.

Steve Furber
You know, life is unpredictable. You can only optimize locally. Right? It's Do

Parker Dillmann
you have anything else, Steven?

Stephen Kraig
No. I think I think that's great. We we really appreciate you coming on and spending an hour with us, Steve.

Steve Furber
Yeah. It's been good fun.

Parker Dillmann
Steve, is there if anyone's got any questions for you, how can they reach out and talk to you?

Steve Furber
People seem to find me on LinkedIn with that too, a stroll.

Parker Dillmann
Put the link, to your LinkedIn in our show notes.

Steve Furber
Okay. Yeah. So people can find me on LinkedIn and contact me through that. I don't read the communication on LinkedIn every day, but I do tend to notice if somebody sent a message eventually.

Parker Dillmann
Well, thank you so much, Steve. Okay.

Stephen Kraig
Yeah. I'll do the outro. Okay. Thank you for listening to circuit break from Macrofab. We were your host, Steve and Craig

Parker Dillmann
and Parker Dillman. Take it easy. Later, everyone. Alright, Steve. I'm gonna just have a little outro here I I need to

Steve Furber
say. Sorry?

Parker Dillmann
I just have a little outro to say. Thank you Yes You Breaker for downloading our podcast. Tell your friends and coworkers about circuit break podcast from Macrofab. If you have a cool idea, project, or topic you want us to discuss, let Steven and I and the community of Breakers know. Our community where you can find personal projects, discussions about the podcast, and engineering topics and news is located atform.macfab.com.