The Raspberry Pi Guy interviews Eben
A few weeks ago Matt Timmons-Brown, The Raspberry Pi Guy, visited Pi Towers to interview Raspberry Pi Trading CEO Eben Upton. Over the course of an hour and a half, Matt drew Eben’s detailed answers on some of the topics that hundreds of you put forward in response to a call for questions.
When might we see an updated Model A and Compute Module? What’s next for the Raspberry Pi camera module and Pi NoIR? How will a Raspberry Pi look, ten years from now? Are there any plans to update Minecraft: Pi Edition for the Pi 2’s hardware?
Watch Matt’s interview for the answers to these and many more. You’ll certainly learn things you won’t have seen elsewhere.
“This is not a product announcement! This is speculation about industry trends! Make sure that that bit goes in the video.”
The Raspberry Pi Guy
Thanks for having me! I hope that I gave Eben the grilling the community so desperately craved for! Subscribe to see more interviews with key Raspberry Pi employees in the future.
For more information on the display you can watch the release here: https://www.youtube.com/watch?v=k7byf2h_bT8
Thanks Matt & Eben, you’ve given the answers to all the questions I asked.
Calling it now – New Raspberry Pi Model C with:
– 8 USB Type-C ports (Including those for display and charging)
– Quad Core 64-Bit ARMv8+
– VideoCore 5 with 128 QPUs
– Next generation video codec from the Alliance for Open Media capable of 4K 60Hz video playback.
– Gigabit Ethernet
– Ditching of the extra USB Hub for reduced cost
– Punch card reader and tape drive for backwards compatibility with existing BBC Micros!
On a more serious note, I hope the whole Y2038 thing does not become a serious issue with all the embedded Pis still running a 32 bit OS.
As always, thanks to all those involved in the interview!
Well, I think one day Gigabit ethernet and 4 usb c ports may well happen. Better USB speeds would help with so many projects and SD cards are getting faster and faster. Who knows where we will be in 5/10 years?
Punched cards?! YaaaaY! Been using the US’ NSA secret files from the Kennedy years as beermats for years…
S’ppose my COBOL learning course no longer useless!
Mechanical teletype (80-volt, natch) is my next interface project….oh, and a COBOL interpreter, goes without saying….
W. H. Heydt
COBOL *interpreter*? Surely you jest. Not only should COBOL be compiled, but according to apt-cache search, there is a COBOL compiler for the Pi.
As for the TTY…20ma current loop interface, right?
What about a single SATA port ? Or two? more CSI lanes for higher-res camera? Before the Pi has more than 4GB ram, we don’t need ARMv8 64bit CPU’s (64bit v8 instruction set is very complicated and I’d therefore say power hungry by itself). Maybe some 32bit big-LITTLE core would help more – to allow Pi to be used both for high-performance desktop and for embedded apps. There was A17 core from ARM if I can remember correctly.
In the interim before overcomplicated USB-C comes, displayport would be great instead of HDMI.
Any chance of a transcript? The “CC” option on youtube is hit-or-miss, mostly miss. I’m interested in the info, but I can read much faster than it takes to say all this.
Unless you’re a really proficient touch typist, transcribing things takes a fair while, and the youthful Matt is suppose to be lending his attention to the college course he started less than 48 hours ago.
(As it happens, I wondered how long it would take me to transcribe this when I watched it; I was a moderately proficient touch typist about a decade and a half ago. It didn’t take long for me to conclude that, yes, it’d be very interesting for many to have a transcript; and that it’d definitely depend on someone else having the time!)
I spent the time to transcribe and post a summary of the first 15 minutes, it appeared here briefly but is now gone. :( Are over long comments auto-deleted?
Hopefully the work on Epiphany will include being able to change the homepage without going into the back-end, better bookmarking, and fixing of the HTML5 video issues.
Please forget about weston/wayland for now. It is too far down the line to be useful for Raspberry Pi, at least for the current generation of SoC. There is loads of work that needs done and it would be a big diversion from the foundation’s core aims if it decided to do that work. In my view it would be much better to wait and see where weston/wayland ends up a few years down the line. Hopefully by that time we will see either KDE or Gnome ported across, and maybe one of the lightweight desktops as well.
1,000 for that Astro Pi case? Wow. Depending on its complexity I could probably make that in pewter to retail for under £80 … anyone know how I can get the design files?
W. H. Heydt
Those cases are machined in a special grade of Aluminum with great attention to surface finish (i.e. no burrs or rough spots anywhere). The ones that are to be flown apparently cost $15K *each*, so $1K would be the cheap version.
Hey Eben, regarding Africa / Developing Countries: You should get in touch with the guys from ProjectCC (the project leader is a friend of mine). They are a small but very dedicated charity from Milwaukee, which emerged out of a student project.
They are setting up computer labs in developing communities for schools and universities in Africa, Jamaica, India, but also for non-profits in the neighborhood in Milwaukee to give people access to Computers.
They are not *from* Africa, but built some good contacts there through their work and know locals on a personal, friendship basis.
http://www.projectcc.org (the website is a bit outdated but they are still active)
W. H. Heydt
Please, please, please… When the replacement camera modules are designed, please build in a way to use C-mount lenses with them. (I’d love to use my F2.7 152mm and my F2.0 63mm macrofocusing lenses in conjunction with a Pi…not to mention the availability of 25mm F0.8 lenses)
+1 to that. In fact any kind of removable lens option would make it seriously wonderful.
I do a lot of photography and thermal imaging for fun and chose the (sadly Android-only) Therm-App thermal camera simply because it’s relatively inexpensive for good thermal (about $1000) but more importantly uses interchangeable screw-in germanium lenses. I’ve used it with everything from a 8″ telescope downwards. I’m not suggesting a Pi Ultra-Noir (though it’d be VERY nice!) but just pointing out that interchangeable lenses are a huge benefit.
W. H. Heydt
Yes… What would really be fun would be an adapter that would let me use my OM mount 600mm F8 catadioptric lens…
The problem is that C mount lenses are designed to be used with 1/2″ or 1″ vidicons, with the modern solid state sensor being only a fraction of the size what used to be a wide angle lens will actually be a telephoto.
True… but the net effect is an even greater zoom. So a cheap ‘normal’ [angle of view] lens intended for a TV camera will act as a telephoto on the smaller sensor.
(Granted, there may also be some issues with stray light bouncing around and causing reduced contrast, but that’s generally easy enough to deal with at a design level by simply making everything other than the sensor black.)
I think the main thing is, if you’ve got access to the naked sensor via some kind or recognised mount, it then becomes easy to hack all sorts of other lenses onto the camera and learn whilst you’re doing it.
W. H. Heydt
Well…*yours* may be designed for vidicons. Mine go with my Kodak Cine Special 1 (using an adapter), and it’s the mount that was used by a lot of 16mm cine cameras.
If you want a smaller format, there is D-mount that was used with the old 8mm (not “Super 8”) cine film.
I was glad to here that MCPI edition is not dead! It would be awesome to incorporate more blocks and especially if there were mobs that could be controlled from the python or sonic pi API. Imagine a great big crowd of minecraft people following you! Awesome!
Ok, my notes on the first half. Anyone want to do the 2nd half?
Eben Upton Interview
unofficial loose paraphrase of first 15 minutes by J.Beale
We cancelled the R-Pi display. No, just kidding.
Here it is, 7-inch 10-point multitouch display, it was a real effort to get it through compliance and into manufacture. 20k RPi displays in existence at the moment (18 Aug.)
We could have done this a lot faster if it had been more expensive.
This is generally the story with Raspberry Pi, you’re trading speed for cost.
We spent a lot of time making this cheap, US$60 plus tax.
Hopefully the first of several display products, we’d like to do a 10-inch high resolution one, there are some challenges there; those panels tend to be LVDS and this is a DPI panel so there’ll be a respin of this board to do that, the good news is that I think LVDS will be easier to get through compliance.
With Model B 2 success, updated Model A anytime soon?
We’re kind of in the same position this year as we were in 2012, RPi’s are in short supply, not desperately short 2012 supply but there’s massive demand for the Model 2B.
We’re constrained by the supply of the Broadcom devices, every BCM device we have in the pipeline is earmarked for the 2B. I don’t think you’ll see a Model 2A from us this year, but it’s an obvious product. If we don’t do one next year I’ll be very surprised.
Same answer for updated compute module. Compute Module 1 has been a pretty successful little product, CM2 is unlikely in Q4 this year.
What is next for the RPi camera module?
The OV5647 got EOL’d about a year ago, we did a lifetime buy this year and our stockpile will take us well into next year at the earliest. There will be a follow-on product when the stockpile’s burned down, but no word on spec yet, the goal is to provide something at least as good, at least as good a price, the $25 price has been pretty popular with people.
Has there been any more development of the replacement Maynard/Wayland desktop?
No. Simple answer. We paused work on this, 1) we want to spend the money on the web browser. This is done by a collaborative consulting company in Cambridge. We spend the money on Epiphany instead. 2) we got Eric Anhold working on the accellerated X graphics drivers. There is at least an even chance that the future is accellerated X, rather than Wayland. Once Epiphany is a little further along, we’re doing a bunch of Webkit 2/Epiphany. In the meantime Simon’s been doing all this work on the X desktop. The question of what accelleration framework we do is a race between Wayland getting mature, and Eric’s stuff.
Update on Astro-pi project?
Astro-pi / Sense Hat should have gone on sale on the 24th at the price of $30. Competition has been very successful around Tim Peake’s flight, very good entries, lot of engineering work going on to get that ready for flight, both HW and SW particuarly the IMU software stack. Final day to deliver the HW, I think it’s in early October, Tim will fly in December. We’re flying 2 units up, We’d like to find ways to fly more units. The units will stay up when Tim comes back down, potentially available to run this in other ESA countries. It’s early days but we’d love to do something with NASA around this. We’ve had a really high level of engagement from both classes of the kids. I think Jonathan said we’re one of the fastest projects to ever go from concept to flight hardware, which is fun. We’d love to find a way to sell the cases, but they are incredibly expensive to make, I think a thousand pound piece of kit, not much market for them. I’m a big space geek. None of us are going to Kazakstan to watch the launch, annoyingly.
What is your reaction to the success of the R-Pi, and what will it be like in 10 years time?
Wow. It’s crazy, isn’t it. The run rates have become completely insane, between 200K and 300K units a month. 1/3 of RPis are now RPi2s. It’s great. From an engineering pt of view we’re doing cool stuff, Educational, we’ve hired Phillip Colligan as Foundation Chief Executive about a month ago, we’re kicking up a notch the big push into education mission. The volume of sales enables us to support that mission. It’s wierd. I was talking to one of the distributors yesterday about their next chip order. Their next 1.2 million unit chip order, that’s 80% of the lifetime volume of the BBC Micro. In one chip order.
It’s wierd, but its good. We’re hungry for a few things, to make it work in some areas where it’s not currently successful. It would be lovely to get a bit more cost down, to bring the new quad-core device to some lower price points. There’s a lot to do. Last year we bulked up the software team with people we hired from Broadcom. This year we’ve grown the hardware engineering team, Mike Stimson(?) joined us at the start of the year. Roger Fontain joined us (both ex-broadcom), we now have quite a decent in-house engineering team.
In 10 year’s time, obviously it will look like this (holds current R-Pi board). So what can you fit into that box in 10 year’s time? Obviously more processing power, not vast amounts more. I don’t think we’ll have 100x more, not desktop form factor, I think we’ll have more than we have. I hope we’ll address a little bit more memory, more hope for more memory than significantly more processing power. I think high speed interfaces, pretty obvious that USB-C is going to be a big thing, I expect in 10 years time that end of the board will look a little different. I think we’ll still have ethernet, we like ethernet, might be a bit faster ethernet. Conceivably in a maximally USB-C world we might not even have that, you can move a lot of video over USB-C cables. This is not a product announcement, this is a speculation about industry trends.
W. H. Heydt
Eben may be a bit conservative on the “not 100 times faster” for 10 years out. Extrapolating from two data points (Yes, I know), we have the Pi2 6 times faster than the Pi1 in 3 years.
If the generation time stays at 3 years, and the capability increase by a factor of 6 with each generation, 3 generations from now would be 216 times the processing power. (If only two generations–figuring the originally anticipated 5 years from Pi1 to Pi2–then it would “only” 36 times in 10 years, which is still pretty impressive.)
That’s a good point, but I think Eben was suggesting that we can’t make that assumption. That is, the Pi’s compute power will not continue to grow at the same rate for 10 years. Not sure if that’s part of larger industry trends or the Pi’s own roadmap, but I understood the Pi1 -> Pi2 step size is thought to be a one-time deal.
Yes, it’s sad to see how the Pi and plenty of other stuff so much more expansive here in Brazil. My Pi 2 was pretty expansive. We also have the most expasive iPhone in the world!