Eben Upton is a busy man. Formerly of IBM and Intel, he’s been at the helm of two major start-ups – Ideaworks3D and Podfun – and currently whiles away his time as a system-on-chip architect and associate technical director for semiconductor giant Broadcom.
We grabbed some time with Upton between meetings at Broadcom’s South California offices, where he was – as is typical – working on several tasks simultaneously. “I’m attempting to move from being 75 per cent in the UK and 25 per cent in the US to the other way around,” he explained after a brief hiatus in the interview as he organised his next meetings with Broadcom colleagues and filled out a dry cleaning form so he’d have clean clothes to wear, “to get a little bit more involved in the guts of things in this organisation.”
There’s another project that’s demanding increasing amounts of his attention, however, and it’s not to be found in the sunny climes of South California: Raspberry Pi.
A project born of the minds of Upton and David Braben, a man perhaps best known for his creation with Ian Bell of the classic BBC model B space adventure Elite, Raspberry Pi has a simple premise: “It’s a project to make a very cheap computer,” Upton explained. “A computer for $25 without networking – the Model A- or $35 with – the Model B.”
With no lesser intent than to boost computing education around the world and train a new generation of hackers – “we think that the lack of programmable hardware for children – the sort of hardware we used to have in the 1980s – is undermining the supply of eighteen year olds who know how to program, so that’s a problem for universities, and then it’s undermining the supply of 21 year olds who know how to program, and that’s causing problems for industry,” Upton explained – Raspberry Pi is a British initiative to produce low-cost yet powerful hardware as a purely charitable act.
What made you decide to found Raspberry Pi as a non-profit charity, rather than as a traditional company?
“I guess the biggest factor is that it’s very hard to deal with really huge companies like Broadcom and our RAM supplier when you’re as small as Raspberry Pi. Typically, if you’re a small organisation like us, you’d obtain components via a distributor because large companies don’t want the hassle of working with you. The problem with a distributor is a distributor comes with a margin, and on small quantities of stuff there are really substantial margins.
“We’re buying ten thousand RAM chips, but ten thousand is tiny to a big company. You can’t expect them to deal with you unless you have some pitch, and, of course, our pitch is that we’re a charity. Nobody’s trying to make a buck. If they think we’re trying to make a buck here, they’ll screw us. Instead, they’ll help us out by not overcharging us for the small quantities we’re buying, so that really helps.
“The other thing is that those of us involved in the charity, we’ve all got other things going on: some of us are academics, and there are people like me who’ve been academics and are now in industry, and none of us need another start-up in our lives. None of us need another money-making start-up, we’ve all done reasonably well for ourselves at all the other things we’ve been doing. So, there wasn’t ever really that incentive to try and turn it into a money-making venture.”
Can you see the Pi becoming a realistic alternative to the likes of the Qualcomm Dragonboard or the Texas Instruments Beagleboard for ARM developers?
“Yeah. It’s not entirely clear to me why the Beagleboard is so expensive. Somebody in that Beagleboard value chain has got to be making a pile of money – I mean, $175 for a Pandaboard or $100 for a Beagleboard? Somebody’s got to be amassing a pile of cash there, because that’s a $10 chip in that device. I don’t know why they’re so expensive. Raspberry Pi, in terms of multimedia, outperforms any other dev board in existence – which is nice.
“In terms of general purpose computing, it’s got this 700MHz ARM11, and our benchmark shows it’s about 20 per cent slower than a Beagleboard for general purpose computing. But, you know, it’s a quarter of the price – somewhere between a sixth and a quarter of the price – so yeah, I expect that our first customers are going to be Beagleboard-type customers.”
The first batch of finished boards was originally scheduled to go on sale towards the end of November: what caused the delay?
“Erm, I guess, attention to detail? We planned to launch in the fourth quarter. I think that for a little while that equated to November, for some reason – I think I said ‘oh, it’d be nice to get it out in November’ – and there had been a feeling that we’d somehow committed to November. So, yeah, we’re a little bit late, but it just took us longer than we’d expected to get a PCB design that we’re happy with.”
When the early beta boards were produced, you noticed a show-stopping problem with the design: what was the reaction like when the problem was first spotted?
“I think we had a very unpleasant twelve hours. Obviously, the problem then would be if it was something really irrecoverable then we would have to ditch the beta boards, and that would put another two months in the time line; as it was, it was a simple enough thing that we could go straight to production with a slightly modified version of the beta board schematic. There’s a limited amount that can go wrong on a board that size, and we were away.”
What other issues have you encountered trying to develop a single-board computer that is so compact, at the size of a credit card?
“PCB routing has been a real issue. It’s not just size, it’s also simplicity: the board is a very simple board, in terms of the number of layers and the kind of features that we use in the PCB design, because features cost money. So, it’s easy to make something which is big and simple, and it’s easier to make something which is small and complicated, but it’s hard to make something which is both small and simple. So, yeah, a lot of effort went in to that, and it is one of the reasons we took so long doing it.
“The alpha boards took about three days to lay out, but you can see they’re big and they’re very expensive. I mean, those are $150 boards – a lot of that is because we were making them in small volume – but they were big and they were eight layers, whereas the final board is small and six layers. That stuff translates directly into routing effort.”
As you ramp up production of the Raspberry Pi, will Broadcom be able to supply you with the system-on-chip components at the price level you need to produce such a low-cost finished product?
“There’s been some speculation that we’re getting a sweetheart deal from Broadcom – an unsustainable sweetheart deal. I can say – putting my Broadcom employee hat on briefly – if we sold all our chips to Raspberry Pi, that’d be just great. We haven’t given a price that means that Broadcom is losing money on this activity. It’s not losing money. The understanding that I have with Broadcom is that they are happy to continue providing chips for this project forever.”
The original beta boards you showed off were running Canonical’s Ubuntu Linux, but that’s no longer a supported option. What changed?
“We thought that we would try to use Ubuntu – Ubuntu’s a very professional-looking system. The problem with Ubuntu, it turned out, was that Ubuntu’s organisation – Canonical – had decided to drop support for ARM earlier than ARMv7. They’re only supporting Cortex-class ARM processors, and we don’t have a Cortex-class processor, we have an ARM11 [ARMv6].
“You know, with the majority of Linux distributions somebody got in touch with us from their organisation saying ‘how can help you get our distribution running on Raspberry Pi,’ which was fine. Ubuntu got in touch with us to say ‘how can we stop you saying that Ubuntu runs on Raspberry Pi?’ Which I thought was pretty brutal, actually. So, yeah: they don’t support our chip, they’re not interested in supporting our chip, they’ve been quite vocal about trying to stop us from saying Ubuntu, so we stopped saying Ubuntu. You could run an obsolete version of Ubuntu – Ubuntu 9.04, which I think is Jaunty – you can run Jaunty on the Pi, but you can’t run any more recent version that that.
“It’s unfortunate, but there is at least one distribution keeping its ARMv5/ARMv6 line alive specifically in order to help us, which isn’t a cost free exercise. We’ve been very fortunate that Red Hat has been extremely helpful to us, but with Canonical… Yeah. That’s the way it goes.”
Do you have any plans to introduce a Model C, with built-in Bluetooth, Wi-Fi or other additional features?
“I think we’re being very careful… I mean, obviously we’re very careful not to talk about future products. You know about this: you talk about your future product, people don’t buy your current product in order to wait for the future product, then you don’t have any interest in your current product so you can’t afford to make your future product and no-one’s happy. This has happened to companies before, so we don’t talk about future products.”
For more information on the Raspberry Pi project, visit www.raspberrypi.org.
SoC: Broadcom BCM2835 Multimedia Processor, comprised of:
CPU: Single-Core ARM1176JZ-F (ARMv6 ISA) at 700 MHz
GPU: Broadcom Dual-Core VideoCore IV Media Co-Processor
RAM: 256MB (Model A & B)
USB: 2x USB 2.0
Video: 1x HDMI, 1x RCA Analogue Video
Audio: 1x HDMI, 1x 3.5mm Analogue Jack
Storage: SD Card
Networking: None (Model A) or 10/100 Ethernet (Model B)
Additional Connectivity: GPIO, UART, I2C, SPI, CSI, DSI, JTAG
Actual Size: 85.6mm x 53.98mm
Related Raspberry Pi articles
- How to set up the Raspberry Pi
- Five alternatives to Raspberry Pi
- Raspberry Pi review – the price is right, but the software is not
- Raspberry Pi sells out within one hour