Inside Raspberry Pi
An in-depth Digital Foundry interview on the remarkable capabilities of the upcoming $25 credit-card sized computer
The concept of a $25 computer may well border on the unbelievable, but within weeks the phenomenal Raspberry Pi initiative will finally bear fruit. After years in development, the first production run of 10,000 credit card-sized computers will be available for sale, the aim being to bring an affordable homebrew programming platform to the masses - a modern day successor to the home computers of the 80s, if you will.
In this interview, Raspberry Pi's executive director, Eben Upton, talks to Digital Foundry about the background to the project, the creation of the hardware, its wealth of potential across a number of fields - and what the games business can do to help it succeed.
And being Digital Foundry, we also talk about the device's surprisingly powerful technical specification. Can the $25 Raspberry Pi graphics core double the performance of the A5 chip found inside the £500 iPhone 4S on certain tasks? You read it here first...
Q: Raspberry Pi has caught the imagination of the tech, gaming and even mainstream press. Has the response to the concept surprised you?
Eben Upton: Absolutely. You have to remember that when we first came up with the idea, we were just looking at what we could do to improve admissions numbers for the University of Cambridge's Computer Science course. Even when we broadened the scope out to the wider UK education sector, we never anticipated this level of demand, or of press interest.
"The entire developed world seems to be experiencing trouble attracting sufficient numbers of young people into the hard sciences. Certainly educators in North America have been biting our hands off to get hold of Raspberry Pi devices."
Q: Are we right in saying that the charitable status of this project extends to the trustees donating their own time, effort and resources? How helpful has Broadcom been in this project?
Eben Upton: That's correct. Right now we have no full-time employees, though I spend about two days a week dealing with various operational issues as the foundation's executive director. Broadcom has been extremely helpful, both in providing our first round of prototype hardware (the alpha boards), and in letting us buy components at a reasonable price in relatively small batches. In addition, several Broadcom employees have spent a lot of their free time over the last six months getting a robust software stack together.
Q: You've identified lack of access to cheap computing as one of the reasons behind the lower quality of applicants to higher level computer sciences courses. However, surely PC ownership is now higher than it is has ever been? In what ways can Raspberry Pi provide a better or more accessible homebrew programming platform?
Eben Upton: Although PC ownership is at an all-time high, at least in middle-class households, we're still a long way from the situation in the 1980s where programmable home computers were almost ubiquitous among young people; a lot of this market has been eaten by games consoles, which provide little or no facility for programming or other creative activity (Little Big Planet being an honourable exception). Even in families with PCs, anecdotally we know that children aren't always allowed free reign to mess about with the machine; breaking a PC can be an expensive experience. We see the Raspberry Pi as offering a platform with pre-installed programming capabilities, which is cheap enough, and hard enough to break, that kids can be cut loose to experiment.
Q: Raspberry Pi runs Linux - can you give us an idea of the kind of programming environments you'd expect to see aspiring coders running on it?
Eben Upton: A range of environments, from Scratch, Logo and Kid's Ruby for younger children and beginners, via tools like YoYo Games' GameMaker, to fully-fledged professional tools like Python and C.
Q: With Windows 8 heading to ARM platforms, could Raspberry Pi run it?
Eben Upton: We've talked to Microsoft a little about this. It appears that Windows 8 will require an ARMv7 (Cortex) processor, so our little ARM11 isn't going to cut it. Perhaps a future version might go there; we certainly get a lot of people asking if they can run Windows applications on the device.
Q: As well as a suitable hardware platform, we're surely looking at a fundamental change in the way computer sciences are taught in schools. Do you see any progress being made there? What kind of discussions have you had in this regard?
Eben Upton: I think the developments over the last couple of weeks, including the Secretary of State's speech on IT education policy, suggest that there's been a sea change in Government opinion. We've had discussions with civil servants, but most of the credit here needs to go to people like Ian Livingstone, who have been pushing at this particular issue for a long time.
Q: Although a British enterprise, you're shipping Raspberry Pi worldwide. Are there similar problems with the quality of computer sciences students elsewhere, or do the needs for the device vary?
Eben Upton: Again anecdotally, the entire developed world seems to be experiencing trouble attracting sufficient numbers of young people into the hard sciences. Certainly educators in North America have been biting our hands off to get hold of Raspberry Pi devices. What has surprised us is the demand for units in the developing world, both for education and for use as a general-purpose productivity machine.
Q: There's something of a perception that Linux is something of a "nerds only" OS with little appeal to the mainstream. Raspberry Pi's reach with the mainstream is potentially phenomenal - can you see a situation where it champions Linux as a Windows alternative in a way we've not seen before?
Eben Upton: The key here is cost. Linux has always struggled to achieve a meaningful foothold on the desktop outside the geek community because the Windows tax is simply not that significant a component of the cost of even a $300 desktop machine. Driving down the cost of hardware as we've done helps make the "free as in beer" appeal of Linux much more obvious; it's hard to imagine someone being prepared to pay a $25 OS licensing fee for a $25 PC.
Q: Does Raspberry Pi's cheap price-point merely cover costs or does it give you a "fighting fund" for future products?
Eben Upton: Both devices are solidly profitable at the target price points. This is a real surprise to us; we'd expected to have razor-thin margins and to subsist for our fixed costs on selling higher-margin peripherals. As it turns out, I think we have better gross margins than Dell.
Q: The $25/$35 price-point was established very early on. As development and pre-production has progressed, what challenges have you had to overcome in coming in on budget?
Eben Upton: Mostly it's been about doing the boring stuff - good supplier negotiation, and picking out components which are a) available, b) not about to go end-of-life, c) reasonably priced and d) reliable. Securing manufacturing capacity at the right price has been a challenge; this is another place that Broadcom has helped, by giving us access to its Taiwanese sales and marketing team to negotiate off-shore manufacturing locally.
Q: Can you give us your thinking on the division between Model A and Model B - isn't the difference in RAM likely to cause some issues in the long run? Over time and with economies of scale, could Model A be phased out in favour of a $25 Model B?
Eben Upton: I think a $25 model B is probably a little out of reach, though as you say there are economy of scale advantages to having a single RAM SKU. Perhaps a RAM upgrade for the model A might make sense in the medium term.
Q: Storage comes via an attached SD card reader - does it support the faster speeds of SDHC cards? Assuming the OS is loaded from SD, would faster cards offer better overall performance, or is there a throughput limit on the bus?
"The games industry has an enormous part to play in solving this problem. We'd like to see companies chip in with tutorials, free asset packs, internships and coding competitions with decent-sized prizes."
Eben Upton: Faster cards do offer better performance. We currently have boot issues with the very fastest cards, which we hope to resolve soon, but you definitely see a difference between a good Kingston card and a no-name one.
Q: We see a lot of Android cellphones being overclocked quite comfortably - can this be done with the Raspberry Pi?
Eben Upton: The ARM is already fairly close to the edge at 700MHz; without overvolting the chip (which decreases lifetime), there's not much more than 100MHz of overclocking headroom on typical silicon. We do offer a clock speed tweak option in the boot configuration parameters, so if you get lucky and get a fast part you can exploit it.
Q: We know that you have a 700MHz ARM11 core in the Raspberry Pi SoC. However, you've described the Broadcom chip as a GPU with ARM elements grafted on. Can you give some idea of the rendering power of the graphics core? Is it true that it comprehensively outperforms NVIDIA's Tegra 2? If so, how?
Eben Upton: I was on the team that designed the graphics core, so I'm a little biased here, but I genuinely believe we have the best mobile GPU team in the world at Broadcom in Cambridge. What's really striking is how badly Tegra 2 performs relative even to simple APs using licensed Imagination Technologies (TI and Apple) or ARM Mali (Samsung) graphics. To summarise, BCM2835 has a tile mode architecture - so it kills immediate-mode devices like Tegra on fill-rate - and we've chosen to configure it with a very large amount of shader performance, so it does very well on compute-intensive benchmarks, and should double iPhone 4S performance across a range of content.
Q: The media decode engine is robust enough to handle 1080p30 Blu-ray h.264 files so Raspberry Pi would make for quite a superb little media player. Do you envisage media playback as part of the proposition or is it more of a "bonus" addition from the Broadcom SoC?
Eben Upton: All the media features are to some extent a bonus, but they've been a part of our thinking ever since I joined Broadcom five years ago (having spent a year trying to build a $25 PC out of openly-available parts like Atmel microcontrollers). I think there's a lot to be said for a device which is useful for something other than programming. The media features provide a "hook" to draw people to the platform; once we have them hooked, we can trick them into becoming programmers!
Q: From your Twitter feed we've seen that Raspberry Pi has caught the imagination of inventors, programmers and other innovators. What are some of the most surprising and exciting potential applications you've seen?
Eben Upton: I think the various balloon and satellite projects have taken us by surprise, but the most unusual one, which turned up very shortly after our first announcement, came from some guys who want to boot into a Sinclair QL emulator and put it inside a QL case. Most of the project suggestions we've seen have fallen into one of a small number of camps (media centre, car automation, home automation).
Q: We've already seen an open source Linux console released in the form of Open Pandora - is this close enough in terms of its technology to see its many apps and emulators ported across to Raspberry Pi?
Eben Upton: I don't know an enormous amount about Open Pandora, but it looks fairly similar in terms of feature set. I'm sure we'll see migration of applications in both directions between the platforms.
Q: There's a GPIO board in development for Raspberry Pi - can you talk us through the add-on and what applications you foresee it being used for?
Eben Upton: The Gertboard, as it's called, has been put together by one of the Broadcom engineers I mentioned earlier. It provides buffered digital I/O, brushed DC motor drive, and an Atmel AVR chip like you find in an Arduino to do analog and low latency I/O. We think it's got potential in a lot of small to medium scale embedded control markets, including home and factory automation, and obviously "physical computing" is a hot topic in IT education right now (witness the success of PICAXE and the like).
Q: You've discussed releasing a case for the Raspberry Pi. Can you envisage a third party economy growing up around the device? For example, could Logitech design an integrated mouse/USB hub casing? Would you have any objections to this?
Eben Upton: We'd have no objections at all, and in fact are working with a number of third-party case manufacturers, including a number who primarily build custom gaming-PC cases. Everything we do is aimed at fostering this sort of community activity, whether it's around software, accessories, or (in due course) licensed and open-source clone manufacture. We're a small organisation which does one thing well; we certainly don't want to stand in the way of third parties who want to add value and make money around the platform.
Q: Part of the reason that homebrew coding took off so dramatically in the 80s was because of the incredible range of games released at the time, and that anybody could join in and code. What part do you think today's games software industry can play in helping the Raspberry Pi initiative?
Eben Upton: The games industry has an enormous part to play in solving this problem. We'd like to see companies chip in with tutorials, free asset packs, internships, coding competitions with decent-sized prizes. One of the challenges facing kids today is that AAA content is so far beyond what they can reasonably hope to achieve, so casual games provide a nice target; this is why we're concentrating on working with companies like YoYo Games to give kids the tools they need to write the next Angry Birds rather than the next Modern Warfare.
Q: Let's look at the upcoming launch - are you still targeting 10,000 units for the first run? Bearing in mind the success you had with the £1 stickers (over 6,000 sold within a day in testing the site's ordering facilities), it's likely you'll sell through within hours. Where do you go after that?
Eben Upton: Another 10,000 units, and another! We're able to scale to the 100,000 unit per annum point with our existing supply of working capital, so we're hopeful in due course that we can catch up with demand. In the medium term, we want to enable other companies to manufacture these devices, which adds a whole new tier of scalability.