1 year ago

Eben Upton talks Raspberry Pi 3

We speak to Eben Upton to find out what makes the Raspberry Pi 3 tick....

It’s been four years since the launch of the original Raspberry Pi, and project co-founder Eben Upton has but one word to describe that period: “Packed!”he laughs from his home in a small Cambridgeshire village. “Just sort of non-stop.”

“There’s just this sense of not really having stopped,” he tells us on behalf of the growing Raspberry Pi team. “Of just having worked continuously.”

The full interveiew, including comments from James Adams, will be available in The MagPi 43, out Thursday.

The culmination of the team’s ambitions, the Raspberry Pi 3 excites Eben for one main reason: “Connectivity. The other stuff is just quantitative change, it’s just faster, more,” he explains. “The wireless and Bluetooth is the big step change on this device. It’s something people have asked us for for a long time. It’s been this missing element of the platform.

The radio chip that makes the Pi 3 wireless

The radio chip that makes the Pi 3 wireless

“It wasn’t on the original device because it was a million miles beyond us at the kind of scale we were at back then,” Eben admits, thinking back to the early days of the Foundation. “We sold 800,000 Raspberry Pis before we hired our first employee. There’s no way you could do this sort of stuff with no employees. It’s eaten a couple of man-years, probably, getting radio onto the device.”


Conformance testing, too, would have been difficult. “You can do all the testing for an unintentional radiator in a couple of days. Pete [Lomas] and I did the original one; Pete, James [Adams], Gordon[Hollingworth], and I did the subsequent products,” Eben recalls. “The radio stuff, you give it to a guy and he takes six weeks and just rams through all of these tests. The test campaign for Pi 3 has cost us, basically, £100,000,” some ten times that of its radiofree predecessors.

The radio chip isn’t the only new feature, of course. The new BCM2837 system-on-chip has been developed specifically for the project by Broadcom. “It’s kind of a mixture of being able to make a business case for it, and then all those people at Broadcom whobelieved in the mission and were prepared to do the work to get it over the line,” Eben explains of how the Foundation was able to convince the multinational chip giant to build custom chips for the project. “That’s people from fresh graduate engineers all the way up to people in the C-suite at the top. Tricky, though!”

USB and PXE network boot

Even with the chip designed and taped out in March of last year, the Foundation had some final input for Broadcom in order to add twonew features: direct USB massstorage and PXE network boot capabilities. “Gordon rewrote the boot ROM for the chip and then provided an updated boot ROM to Broadcom, saying ‘shove this in the chip, it’ll work’,” Eben laughs. “And it does!

The Raspberry Pi 3 in its full glory

The Raspberry Pi 3 in its full glory

“The other interesting thing about the chip is for all other ones the implementation work was done entirely in Cambridge; this one was a collaboration with Broadcom’s settop box engineering group in Aztec West [business park] in Bristol.”

Eben readily admits that not all the capabilities of the new parts are going to be used at launch, however. “Although it is a 64‑bit core, we’re using it as just a faster 32-bit core,” he reveals aboutthe Pi 3’s central processing unit. “I can imagine there’d be some real benefits [to 64-bit code]. The downside is that you do really create a separate world. To access that benefit, you’d have to have two operating systems. I’m hoping that someone will come and demonstrate to me that this is a good idea. But there are some really compelling advantages to still being basically ARMv6, and because it’s [Cortex-]A53 it’s a really good 32‑bit processor.”

Aware of the high demand that a Raspberry Pi launch typically encounters, talk turns to production. “We’ve already made 30,000 of it,” Eben discloses, still some weeks ahead of the release. “I’d like to do at least three million units, and that requires you to build, say, 60,000 a week. We’ll probably build around 100,000 a week in the UK, plus whatever Embest builds in China, for a month or two, then you back off to maybe 60,000 a week.”

Subscribe now and get a code to order Pi 3 first!

Subscribe now and get a code to order Pi 3 first!

Excited by all this and the possibilites of the Raspberry Pi 3? Want to get one right now? Subscribers to the print version of The MagPi can jump the queue at The Pi Hut to get priority handling on your orders. The good news is, if you’re already a print subscriber, you should be getting an email about it right now. The better news is, if you subscribe to The MagPi before Sunday 13 March 2016 you too can get a code to jump the queue. Subscribe today, jump the queue and never miss another issue.

Find out more about the Raspberry Pi 3 on the Raspberry Pi website, including other places you can try to buy it from.

  • You keep coming out with new boards, BUT, you don’t change to USB 3.0!
    What’s with that? I know you can replace the controller chip so USB 3.0
    will work!

  • Dean Woodyatt


    it’s not just the silicon on the SOC that needs changing for USB3….
    look at it as a law of diminishing returns when updating a product at a fixed price

    They made a good call with the wireless connectivity

  • OK, I’ll give you that. Maybe the built-in Wi-Fi will stay connected; because
    the external ones don’t! And I’ve tried 3 of them– same problem.
    They connect, and then they don’t respond for 30 seconds or so, then they
    connect again.

  • Kyle Douglas

    Disable power saving for that wifi chip mate, will keep it connected 100% of the time

  • If were talking about the Pi 3, I’ll have to look for that command.
    If were talking Pi 2, what is the command or files to modify?

  • cyclicredundancy

    “direct USB massstorage” – does that mean Pi3 will be able to boot directly from usb flash memory? 🙂

  • radoo

    Just bought a bunch of these and I was wondering what’s the status on PXE boot?
    It didn’t work out of the box and I am booting pc’s just fine on PXE on the same network cable even.

  • nigra truo

    Well, why 3.0? The PI can’t even use the full speed of USB 2, the CPU would be too slow for that kind of transfer rate. Also, I assume that just the licensing costs for 3.0 would be double what the PI costs. I hear people clamor and ask for SATA too, but that also does not make the least sense: the PI is just too slow to serve as a good NAS. Shoveling all that data around is hard work. Some people just want the PI to be what they want, not what it is really good at. These are the people that could not care less about GPIO and GPIO is much more integral for what the PI really is and what it does best. I often hear the Raspberry Pi Foundation say “this is not useful or required for an device for education” and they got a good point there.

  • nigra truo

    Huh? Very strange, I have tons of PIs and have used them for years, developed on them via wireless, never ever had that problem. You might want to check your network. Also make sure you get a edimax wifi dongle.

  • WildWalker

    I think the reason for Sata is connectivity not speed. At lot of people (me included) host websites from the Pi, and using SD cards is a bit of a worry. A SATA interface you could connect a small 2.5 hdd to, and boot from, would be very nice. Again it’s not about performance, just a bit of reliability.

    I too read (when the Pi came out) that it would at some point boot from USB, but alas this has not transpired.

    I agree that the GPIO functionality should not be compromised however, in fact I would love some analogue inputs for some of my projects (be nice to use joypad input for menus on LCDs rather than buttons for example)

    You said that “Some people just want the PI to be what they want” well that’s true for everyone, you included. It’s just that the Pi is much closer to some peoples ideal than others, no reason why the wishes of others can’t be considered though right?

  • nigra truo

    You know, analogue GPIO ports would be a huge boon, I have been amazed of how much stuff is analogue and my next project is to create a analogue digital converter for measuring stuff. This is a lot of work to do when you have to do a dedicated circuit or getting a hat for it, but very easy (and probably cheap) to build into the Pi. It would also directly serve the Pi’s core market: Education, you often need analogue inputs for that, and I feel that many projects probably switch to the Arduino, because it does have both analogue and digital inputs.
    I see the rationale behind SATA, never looked at it like this. And it is true that SDcards are extremely unreliable, you can actually just wait for them to break.
    Booting over USB? I’m not a big fan of that. I have done backups to a lot of USB attached HDs and the experiences were not really cool: USB tends to disconnect for no reason whatsoever once in a while and of course, no auto reconnect. SATA is much more robust for that. I guess USB was never meant for a permanent reliable connection.

    How about doing a site similar to Ubuntu Brainstorm? Where people could propose what they would like in the Pi and then others vote ideas up and down? So you could see what the most wanted ideas are. I don’t know if the Raspberry Pi Foundation would even heed something like this or care, but they might.
    Even though their focus is education, features that are wanted by the market means that these features are used much in the market and what else is education than preparation for the real life usage in the market?

  • nigra truo

    OR, did the “the powersource is not strong enough, my Pi crashed” in truth be “wifi dongle disconnected and won’t reconnect”?