00:00SPEAKER- Hey everybody, Will here.
00:01SPEAKER- Hi, I'm Brad.
00:03SPEAKER- We are doing something a little bit different this week.
00:04SPEAKERI'm on vacation.
00:06SPEAKERYou're also on vacation?
00:08SPEAKER- No.
00:09SPEAKER- Well, you're about to be on vacation.
00:10SPEAKER- Or actually, yeah, the following week, yes.
00:11SPEAKERThere's like a day of overlap there, sure.
00:14SPEAKER- Yeah, you're off now.
00:15SPEAKERCongratulations. - Yeah, that's right.
00:15SPEAKERNo, thank you.
00:17SPEAKER- We are, instead of running a new tech pod this week,
00:19SPEAKERwe have a FOSPOD episode that we...
00:22SPEAKERLook, if you asked me to pick my favorite FOSPOD episode,
00:25SPEAKERit would be impossible for me to do so
00:27SPEAKER'cause they're all like children
00:28SPEAKERand I love them all equally,
00:30SPEAKERbut we have a fantastic episode of the FOSPOD this week.
00:35SPEAKERWe have Eben Upton from the Raspberry Pi Foundation
00:38SPEAKERand Raspberry Pi on to talk about
00:41SPEAKERmaking these little computers
00:43SPEAKERthat have literally changed my life.
00:45SPEAKERLike I don't think we'd be doing this show
00:48SPEAKERwere it not, like the Raspberry Pi is one of the things
00:50SPEAKERthat opened the door to doing this show for me.
00:52SPEAKER- I think I said it somewhere in this episode,
00:55SPEAKERI think it was pretty much my gateway into Linux
00:57SPEAKERis the first time I ever used Linux
00:59SPEAKERwas the first time I got a Pi.
01:00SPEAKERAnyway, yeah, so this is excellent.
01:03SPEAKERIf I do say so myself, an excellent conversation.
01:07SPEAKERHe's super fascinating to talk to.
01:08SPEAKERAlso, this is a never before heard episode.
01:11SPEAKERThis is brand new in both the FOSPOD feed
01:13SPEAKERand also going on the tech pod feed here.
01:15SPEAKERSo even if you're an avid FOSPOD listener,
01:17SPEAKERyou have not heard this one yet.
01:19SPEAKER- Yeah, it's fresh. - That's right.
01:21SPEAKER- Enjoy and have a piece of Pi
01:23SPEAKERwith your Raspberry Pi episode.
01:24SPEAKER(upbeat music)
01:26SPEAKERBrad, how many computers did you have in your house in 2010?
01:29SPEAKER- Let's see, a gaming desktop and a laptop,
01:33SPEAKERand I think that's it.
01:34SPEAKERHow about you?
01:35SPEAKER- That was probably me too.
01:36SPEAKERI might've had a server.
01:37SPEAKERI think I probably had a server by then,
01:39SPEAKERbut it was like a really low powered,
01:41SPEAKEReven for the time Adam processor, home server type deal
01:46SPEAKERthat was basically just like an SMB share.
01:48SPEAKERI didn't have anything beyond that.
01:49SPEAKERI had a router, which I guess you could argue
01:51SPEAKERis kind of a computer, but not much else.
01:54SPEAKERIt was probably running Linux of some sort,
01:56SPEAKERalthough you may not have known that,
01:57SPEAKERbut yeah, like even two computers at that time
01:59SPEAKERfelt a little bit extravagant,
02:01SPEAKERbut boy, little did we know what was on the way.
02:04SPEAKER- Yeah, how many computers do you have running
02:05SPEAKERin your house right now in 2023, Brad?
02:06SPEAKER- Gosh, well, okay, I've still got a desktop.
02:09SPEAKERI've still got a laptop.
02:10SPEAKERNow I've got this big Linux box over here,
02:12SPEAKERbut also at least four Raspberry Pis around here,
02:16SPEAKERmaybe more, and if you want to get into,
02:19SPEAKERdo containers running on the Raspberry Pis
02:20SPEAKERcount as computers, 'cause then we're up to like two dozen.
02:24SPEAKER- Yeah, I'm in a similar, like I have the desktop,
02:27SPEAKERthe streaming machine, the NAS, my kid's computer,
02:29SPEAKERmy wife's computer, and then about three,
02:31SPEAKERtwo, three Raspberry Pis at any given time.
02:33SPEAKER- They just keep multiplying, right?
02:35SPEAKERThe Raspberry Pi has become so indispensable,
02:38SPEAKERI have thought about adopting a policy
02:40SPEAKERof just having an extra on hand at all times.
02:42SPEAKERI mean, not super feasible right now
02:44SPEAKER'cause they're hard to come by for supply chain reasons,
02:46SPEAKERwhich you will hear about shortly,
02:48SPEAKERbut sometimes a project urge just strikes
02:51SPEAKERand you need a Raspberry Pi.
02:53SPEAKERMight be nice to just be able to reach
02:54SPEAKERinto the drawer and grab one.
02:57SPEAKER- Well, I just set up a home assistant yellow,
02:59SPEAKERwhich is a Raspberry Pi compute module based project
03:03SPEAKERthat freed up one of my old Pi threes,
03:05SPEAKERand I was really excited 'cause like,
03:06SPEAKERI would have more Pis in the house right now,
03:08SPEAKERexcept for you have not been able to buy them
03:10SPEAKERfor the last two years reliably, so.
03:13SPEAKER- Think about the Pi though, it's always useful.
03:15SPEAKER- You can never have too much Pi, that's what I say.
03:17SPEAKER(upbeat music)
03:23SPEAKER(upbeat music)
03:35SPEAKER- Welcome to the FOSPOD.
03:36SPEAKERI'm Will.
03:37SPEAKER- I'm Brad.
03:38SPEAKER- As always, this week's episode of the FOSPOD
03:40SPEAKERis brought to you by Google Open Source.
03:42SPEAKERThey bring all the value of open source to Google
03:44SPEAKERand all the resources of Google to open source.
03:46SPEAKERYou can find out more at opensource.google.
03:50SPEAKERThis week on the FOSPOD, we're talking to Evan Upton,
03:52SPEAKERco-founder and CEO of Raspberry Pi Trading Company,
03:54SPEAKERand well, I mean, he's just a fascinating guy, it turns out.
03:59SPEAKER- And the Raspberry Pi Foundation.
04:00SPEAKERThat's something I learned in the course of this interview
04:04SPEAKERwas the distinction between those two,
04:05SPEAKERbut dude, Raspberry Pi, it's like,
04:08SPEAKERperhaps as big an institution in the maker world,
04:11SPEAKERthe open source world, the--
04:13SPEAKER- Manufacturing.
04:14SPEAKER- Yes, I mean, the range of contexts
04:17SPEAKERin which these adorable little fruit flavored
04:20SPEAKERsingle board computers are used is kind of endless
04:23SPEAKERthese days.
04:24SPEAKER- Yeah, it's funny 'cause it's,
04:25SPEAKERwe were talking about this before we started the show.
04:28SPEAKERI can't really think of a project that, you know,
04:31SPEAKERin the normy, non-techie world,
04:33SPEAKERpeople are pretty much completely unaware of,
04:36SPEAKERbut like in our circles, everybody knows about this
04:40SPEAKERand it became such a ubiquitous, important part
04:43SPEAKERof the computing ecosystem in like less than a decade really
04:46SPEAKER'cause like the first Pi boards launched in 2012
04:50SPEAKERand there honestly wasn't a whole lot you could do with them
04:52SPEAKERif you couldn't write code in those early days.
04:54SPEAKERAnd then projects started building,
04:57SPEAKERlike the fact that this single board computer existed
04:59SPEAKERthat was 20 or $30 meant that projects would start tailoring
05:04SPEAKERtheir kind of single service computer projects
05:07SPEAKERspecifically for them.
05:08SPEAKERWhether it was like an arcade machine emulator
05:10SPEAKERor an old console emulator or a router or a DNS server
05:15SPEAKERthat strips out all advertising from the internet.
05:17SPEAKER- Yes, yeah, print servers, 3D print servers or interfaces,
05:22SPEAKERhome assistant, you name it.
05:23SPEAKER- Yeah, network VNC machines.
05:27SPEAKER- Smart mirrors.
05:28SPEAKER- Wait, smart mirrors?
05:29SPEAKERI haven't seen that one.
05:30SPEAKER- Dude, if you never, there are a lot of smart mirror
05:32SPEAKERprojects involving Raspberry Pis out there.
05:34SPEAKERThey look so cool.
05:36SPEAKER- Oh man, I just put a new medicine cabinet in.
05:38SPEAKERI should do a smart mirror in there.
05:39SPEAKER- You should slap a Raspberry Pi on the guts
05:42SPEAKERof an old monitor on the back of the mirror.
05:44SPEAKER- Yeah, the Pi is such a formative kind of gateway
05:48SPEAKERand open source for me.
05:49SPEAKERI'm pretty sure the first model B I bought
05:51SPEAKERwas the first real experience I had with Linux, honestly.
05:55SPEAKER'Cause like you said, those early projects,
05:57SPEAKERI mean, I wasn't writing code,
05:58SPEAKERbut those early, early projects that were available
06:01SPEAKERfor the Pi, you kind of couldn't set up
06:03SPEAKERwithout at least touching a bash prompt, right?
06:04SPEAKERAnd I don't think I had done a lot of that
06:06SPEAKERprior to getting my first Pi.
06:08SPEAKERSo a big deal for, I think, a lot of people's first steps
06:12SPEAKERinto the world of Linux and open source.
06:15SPEAKER- Yeah, so for folks who don't know,
06:16SPEAKERthe Raspberry Pi is a single board computer.
06:18SPEAKERIt runs ARM processors.
06:20SPEAKERThere's four main models and a bunch of little sub models
06:23SPEAKERnow, maybe five main models.
06:24SPEAKERI think the zero probably counts
06:25SPEAKERas its own thing at this point.
06:26SPEAKER- Yeah, yeah.
06:27SPEAKER- And you can use them for all sorts of things.
06:29SPEAKERThere's two things that make them different.
06:30SPEAKERWe talk about a little bit in the episode,
06:32SPEAKERbut one is that they're, pandemic concerns aside,
06:36SPEAKERtypically always available.
06:37SPEAKERSo they're not batch products.
06:38SPEAKERThey're products that they just continuously manufacture.
06:40SPEAKERYou can go to your local electronics store
06:42SPEAKERand get one.
06:43SPEAKER- They still make the original.
06:45SPEAKER- Yeah. - It's been out
06:45SPEAKERfor over a decade.
06:46SPEAKERThey are committed to making that thing
06:47SPEAKERfor several more years.
06:49SPEAKER- The other thing is, what used to be called Raspberry,
06:52SPEAKERand it is now called Raspberry Pi OS,
06:54SPEAKERis a fork of Debian that's basically designed explicitly
06:58SPEAKERto run on these boards.
06:59SPEAKERAnd there's an installer that you can download for it
07:02SPEAKERthat runs on Windows and lets you put the right OS files
07:05SPEAKERon a flash drive or a USB thumb drive
07:07SPEAKERor whatever it is you happen to need
07:10SPEAKERthat lets you basically jump straight
07:13SPEAKERinto the appliance mode for the Linux setup
07:15SPEAKERfor these devices.
07:16SPEAKERYou don't necessarily have to start from,
07:18SPEAKERfirst you install Debian,
07:19SPEAKERand then you apt-get install your,
07:22SPEAKERwhatever project it is you want,
07:23SPEAKERand then you have to do all the configuration stuff.
07:24SPEAKERIt just works.
07:25SPEAKER- Yes.
07:27SPEAKERYeah, that's really what it is.
07:28SPEAKERI think this thing has reached a critical mass.
07:30SPEAKERLike it's picked up that level of community support
07:32SPEAKERthat you don't have to be a Linux gray beard.
07:34SPEAKERLike there's just endless tutorials and tools and guides
07:37SPEAKERand all kinds of stuff out there to ease you into this.
07:40SPEAKER- To the point that often the projects say,
07:42SPEAKER"Hey, the easiest way to set this up
07:43SPEAKERis to just buy a Raspberry Pi,
07:45SPEAKERinstall it on the Raspberry Pi.
07:46SPEAKERDon't try to run it in a container
07:48SPEAKERon a normal Linux installer or whatever."
07:50SPEAKER- Yes, unless you're a loon.
07:52SPEAKERYeah, it's often as simple as here's our image,
07:55SPEAKERhere's how to flash it to an SD card.
07:56SPEAKERJust put it in there, press the power button, you're good.
07:58SPEAKER- I think that's all you need to know about this interview.
08:00SPEAKERLet's just get into it with the first time I heard
08:03SPEAKERabout a Raspberry Pi in person.
08:06SPEAKER(soft music)
08:11SPEAKEREvan, I went to a lot of Maker Faires in the early 2010s
08:13SPEAKERand the best part of that was always getting
08:15SPEAKERto interview people who'd made a fascinating stuff.
08:17SPEAKERBut I usually close those interviews by asking people
08:20SPEAKERwhat they were excited about at the Maker Faire.
08:23SPEAKERAnd in 2012 or 2013, I remember I went to one Maker Faire
08:26SPEAKERand five or six people within the first half day said,
08:31SPEAKER"Oh, there's this little computer the size of an Arduino,
08:33SPEAKERyou gotta check it out, it's really cool."
08:35SPEAKERAnd I'm curious, like that was the moment I was like,
08:38SPEAKER"Oh, I should be aware of this thing,
08:39SPEAKERI gotta learn about this."
08:40SPEAKERI'm curious if there was a moment for you
08:43SPEAKERwhen you realized the Pi was something
08:44SPEAKERthat like really filled a need,
08:46SPEAKERlike this personal need that maybe first nerds
08:49SPEAKERand then just normal people it turns out had.
08:52SPEAKER- Yes, we had, I mean, obviously this isn't something
08:57SPEAKERthat we'd ever imagined was gonna sell a lot of units.
09:01SPEAKERIt's the thing we were building to teach,
09:04SPEAKERto encourage kids to come to Cambridge.
09:05SPEAKERAnd so our idea of success was, you know,
09:08SPEAKERour idea of success was a thousand units
09:10SPEAKERand our idea of wild success was 10,000 units.
09:13SPEAKERI think the point where I realized that we were in trouble
09:18SPEAKERwas in the autumn of 2011.
09:21SPEAKERSo we had, we'd been working on it since 2008
09:24SPEAKERand we'd sort of told people in May of 2011
09:28SPEAKERthat we were doing it and we'd done this
09:30SPEAKERby going to see a guy called Rory Kethlen-Jones,
09:33SPEAKERwho's a BBC technology correspondent.
09:35SPEAKERAnd he took a little video of one of us
09:39SPEAKERholding up a very early Raspberry Pi prototype
09:41SPEAKERand talking about what we were trying to accomplish.
09:44SPEAKERAnd he got this vast number of YouTube views for his video.
09:47SPEAKERHe got me 600,000 YouTube views in a couple of days,
09:50SPEAKERwhich was kind of cool, but it's very easy to,
09:52SPEAKERyou want to apply a very large discount factor
09:54SPEAKERto that number to measure interest.
09:56SPEAKERSo yeah, we thought we were fine.
09:58SPEAKERAnd then we spent a lot of time in 2011
09:59SPEAKERgetting from the sort of point.
10:01SPEAKERIt was the kick that we needed.
10:02SPEAKERIt was the kick in the pants
10:04SPEAKERthat we needed to actually go from
10:06SPEAKERnoodling away at this thing
10:07SPEAKERto actually trying to turn it into a real project.
10:11SPEAKERSo we were spending 2011 going through that process
10:13SPEAKERof productionizing Raspberry Pi.
10:15SPEAKERBut we didn't imagine that we had 600,000 people
10:18SPEAKERwho were interested enough to buy a Raspberry Pi.
10:20SPEAKERBut the point where we realized that we were in trouble
10:22SPEAKERand that we got to sell a lot of these
10:24SPEAKERand that we would have to scale the organization
10:25SPEAKERto support that was when we put an operating system image
10:29SPEAKERonline in about November of 2011.
10:32SPEAKERAnd it got something like, and this was a,
10:34SPEAKERI think that this was a Fedora-based operating system.
10:38SPEAKERAnd it had something like 50,000,
10:40SPEAKERand you could only run it in QEMU, right?
10:42SPEAKERThere was no hardware out there.
10:43SPEAKERSo you could run it in a modified QEMU.
10:46SPEAKERWe had 50,000 downloads
10:47SPEAKERof a very badly broken operating system
10:50SPEAKERfor a computer that didn't exist yet.
10:52SPEAKERAnd that was the bit where we realized that there was,
10:54SPEAKERthat somehow this had tapped into,
10:57SPEAKERit had found in that kind of big parameter space
11:00SPEAKERof the things people were interested in,
11:01SPEAKERwe'd accidentally dropped a product
11:04SPEAKERinto a completely empty space
11:06SPEAKERwhere a lot of people were standing.
11:08SPEAKERIt was empty of products, but very full of people.
11:10SPEAKER- Yeah, it's probably that.
11:11SPEAKER- The idea that you could get a whole,
11:14SPEAKERI mean, so in those early days,
11:16SPEAKERit was novel to have a Linux ARM machine, right?
11:19SPEAKERLike finding compiled binaries and stuff
11:21SPEAKERfor Linux software, it wasn't,
11:24SPEAKERdifficult's not right, but it was unusual,
11:25SPEAKERI think, still then.
11:26SPEAKERAnd there was an inflection point
11:28SPEAKERaround the time the second Raspberry Pi came out
11:31SPEAKERthat that seemed to stop,
11:32SPEAKERthat all of a sudden everybody had Pi-compatible binaries
11:35SPEAKERfor everything and all their repos.
11:37SPEAKER- Yeah, I think it was suddenly when we started,
11:39SPEAKERI guess probably 2009, 2010 when we were first working
11:43SPEAKERon what would become the Raspberry Pi platform,
11:46SPEAKERARM Linux itself, the kernel was an unusual thing, right?
11:50SPEAKERI can't remember, 'cause it existed for a long time
11:53SPEAKERoutside the tree.
11:55SPEAKERI guess it had maybe been merged by then,
11:57SPEAKERbut ARM Linux was outside the tree for a long time.
12:00SPEAKERIt had probably been merged, but probably fairly recently.
12:03SPEAKERThere were various distros available,
12:07SPEAKERbut it was, yeah, it was all a bit,
12:08SPEAKEReverything was a bit special on how you built things.
12:12SPEAKERI remember, you know, trying to find a compiler,
12:16SPEAKERtrying to find a pre-built GCC to bootstrap yourself was,
12:20SPEAKERand to build your kernel, wasn't trivial.
12:22SPEAKERAnd yeah, there's just kind of a moment, you know,
12:24SPEAKERif you draw that line between, I guess,
12:26SPEAKERthat point, the kind of 2010, and probably, you could say
12:31SPEAKERmaybe the launch of the M1, max, was probably the point
12:34SPEAKERwhere ARM as a client architecture kind of emerges
12:38SPEAKERcompletely into the mainstream.
12:39SPEAKERThere is that point somewhere in the middle there
12:41SPEAKERwhere most, where large numbers of applications
12:44SPEAKERwere starting to provide binaries,
12:45SPEAKERgenerally Raspberry Pi binaries, generally binaries
12:47SPEAKERfor this slightly unusual, the Raspbian world, right?
12:51SPEAKERSo that's ARM7HF, so hard float.
12:56SPEAKERARM Debian, but built to remove the V7 specific instruction,
13:02SPEAKERso it could run on the ARM1176
13:05SPEAKERthat we had in Raspberry Pi 1.
13:06SPEAKERThose sorts of things started to appear
13:08SPEAKERin kind of 2013, 2014, 2015.
13:10SPEAKER- I'm gonna ask you to suspend your humility
13:12SPEAKERfor just a second here.
13:13SPEAKERWould you say the Pi was the first kind of primary driver
13:16SPEAKERof ARM adoption in Linux,
13:17SPEAKERor were there other solutions out there concurrently?
13:20SPEAKER- I mean, of course, yeah, Beagle, I would say Beagle,
13:23SPEAKERthe Beagle board predates us probably the first
13:27SPEAKERbroadly available.
13:29SPEAKERI guess it didn't have the continuous,
13:30SPEAKERone of the interesting things about Raspberry Pi,
13:32SPEAKERand of course, we've been challenged with this
13:34SPEAKERover the last couple of years,
13:35SPEAKERbut for most of its life, one of the interesting things
13:37SPEAKERthat's characterized Raspberry Pi
13:38SPEAKERis it's kind of continuous available,
13:40SPEAKERcontinuous high volume availability.
13:42SPEAKERAnd that was probably a thing
13:43SPEAKERthat the Beagle board didn't have.
13:44SPEAKERIt was something which was sporadically available.
13:46SPEAKERThey would build some and then sell them out
13:48SPEAKERand then build some more.
13:49SPEAKERBut it did, and it was probably three or four times
13:52SPEAKERthe price, four or five times the price of Raspberry Pi,
13:54SPEAKERbut it did fit that same niche.
13:56SPEAKERBut yeah, certainly I think we popularized,
13:59SPEAKERwe popularized it as a platform.
14:00SPEAKERSort of going onto that point about the M1 machines,
14:03SPEAKERof course, we were super happy.
14:04SPEAKERWe have this object called Raspberry Pi 400, right,
14:07SPEAKERwhich is a Raspberry Pi 4 built into a compact keyboard.
14:09SPEAKERWe launched that about a week before the M1 Macs.
14:12SPEAKERNow I grew up with, not owning, but lusting after,
14:16SPEAKERAcorn Archimedes, the original ARM computer,
14:20SPEAKERthe original ARM desktop.
14:21SPEAKERAnd it was kind of fun that we actually
14:23SPEAKERwere first back in market by a week.
14:25SPEAKERWe were, after a 20-year period in the wilderness
14:28SPEAKERwhere there were no kind of consumer-oriented PCs,
14:32SPEAKERwe, by a week, were first back in market.
14:35SPEAKER- Very basic chicken and egg question here,
14:36SPEAKERbecause the Raspberry Pi's foundation,
14:38SPEAKERlike the original mission statement,
14:39SPEAKERis educational outreach, right?
14:40SPEAKERSo was it a situation of we need to do educational outreach,
14:44SPEAKERand then through brainstorming, you came to,
14:46SPEAKERhey, we can make a tiny, very affordable computer?
14:48SPEAKEROr was it the opposite of, hey, we can make this tiny
14:51SPEAKERcomputer because the parts and the software are now emerging
14:54SPEAKERand then what can we do with it, oh?
14:56SPEAKER- Yeah, it's interesting, right?
14:58SPEAKERAnd it's all about how you tell stories, right?
15:00SPEAKERWhich bit you want to emphasize.
15:02SPEAKERI think, for me, it was a coming together of those,
15:06SPEAKERit was a perfect coming together of those two strands.
15:09SPEAKERAnd there wasn't really a chicken and egg.
15:11SPEAKERThat I was building little computers,
15:13SPEAKERactually quite Arduino-like little computers.
15:15SPEAKERSo there's this thing that if you search
15:17SPEAKERfor Raspberry Pi 2006 edition,
15:19SPEAKERthere's this rather cobbled together looking piece
15:22SPEAKERof error board that has an Atmel,
15:24SPEAKERso the same chip architecture as the Arduino Uno.
15:28SPEAKERIt's an Atmel 644, it's got 512K of SRAM,
15:32SPEAKERand it will drive a standard definition video signal,
15:34SPEAKERand it kind of, it bashes,
15:36SPEAKERactually uses a microcontroller both to run your code,
15:39SPEAKERbut also during the display period
15:41SPEAKERto bash out video addresses.
15:43SPEAKERTo get this SRAM to drop bytes onto the bus
15:46SPEAKERin the order required to draw something on the screen.
15:50SPEAKERSo I was building those things,
15:52SPEAKERand I was also a director of studies in computer science
15:54SPEAKERat the University of Cambridge.
15:55SPEAKERAnd so I was also experiencing the collapse
15:59SPEAKERin the number of young people
16:00SPEAKERwho are interested in computer science.
16:01SPEAKERAnd we saw this terrible, terrible decline
16:04SPEAKERbetween the late 1990s and the mid part
16:06SPEAKERof the first decade of the century.
16:09SPEAKERAnd so I was kind of exposed to both things.
16:11SPEAKERSo it was a sort of a natural coming together.
16:12SPEAKERAnd really, remember, Raspberry Pi is testing the hypothesis
16:16SPEAKERthat we used to have loads of kids
16:18SPEAKERwho were interested in computers
16:19SPEAKERbecause there were loads of programmable computers
16:21SPEAKERin kids' lives.
16:22SPEAKERAnd that when you took those computers away,
16:24SPEAKERbecause you were largely replacing with games consoles
16:26SPEAKERin a lot of children's lives,
16:28SPEAKERthe ultimate piece of very powerful non-programmable hardware
16:31SPEAKERand non-end user programmable hardware.
16:34SPEAKERThe disappearance of general purpose computing
16:36SPEAKERmade the kids go away.
16:37SPEAKERAnd if you bring back general purpose computing,
16:39SPEAKERthe kids will come back.
16:40SPEAKERThat's the hypothesis we're testing.
16:42SPEAKERAnd it was a kind of hypothesis
16:43SPEAKERthat was very easy to stumble upon
16:44SPEAKERif you happen to be both building little computers
16:47SPEAKERand confronting the disappearance
16:50SPEAKERof computing among young people.
16:52SPEAKERSo there's not really, it's a kind of a perfect,
16:55SPEAKERa perfect coming together of these two strands
16:57SPEAKERof certainly of my life.
16:58SPEAKER- It's funny 'cause it's an idea
17:01SPEAKERthat has shown up multiple times.
17:03SPEAKERYou know, growing up in America,
17:05SPEAKERwe didn't have the rise of programmable computers as much.
17:09SPEAKERThey were much less available, I think, here
17:11SPEAKERthan they were in the UK specifically.
17:13SPEAKERAnd it's a thing that comes up a lot
17:16SPEAKERwhen I talk to programmers of a certain age, I think,
17:18SPEAKERfrom the UK specifically.
17:19SPEAKER- The UK is weird and it certainly is a place.
17:22SPEAKERYeah, if you look at the number of 8-bit computers,
17:26SPEAKERI mean, of course, you know, the United States,
17:28SPEAKERyou had your Commodore 64s and you had your trash 80s
17:30SPEAKERand you had a bunch of other architectures
17:33SPEAKERand your Apple IIs.
17:34SPEAKERBut if you look at the number of distinct
17:38SPEAKER8-bit microcomputer architectures
17:40SPEAKERthat came out of a country that has a fifth
17:43SPEAKERof the population of the United States,
17:45SPEAKERit's just incredible, right?
17:46SPEAKERYou know, look, you look at the number
17:47SPEAKERof architectures that came out of Cambridge,
17:49SPEAKERthat came out, developed within five miles
17:52SPEAKERof where I'm sitting, you know, it's just,
17:53SPEAKERthere was something about the place
17:55SPEAKERthat made it a very fertile ground.
17:56SPEAKERAnd of course, many of those companies sold 10,000 units.
18:00SPEAKERThe two big ones, of course, were Sinclair,
18:02SPEAKERwho sold about 5 million of their Spectrum product,
18:05SPEAKERmostly in the UK, but did have some export success.
18:07SPEAKERAnd then, of course, Acorn, who sold about
18:09SPEAKERone and a half million BBC microcomputers,
18:12SPEAKERfailed as a computer company,
18:14SPEAKERbut then succeeded in terms of spinning off
18:16SPEAKERthis little chip design organization that became Arm.
18:19SPEAKER- Talking about the decline in kind of interest
18:22SPEAKERin computing among young people that you mentioned,
18:23SPEAKERI mean, there's this emerging theme in the tech press,
18:26SPEAKEReven these days, that basically says like,
18:28SPEAKERoh, you know, kids don't know what a file system is anymore.
18:30SPEAKERLike, they only know how to type on touch screens now.
18:32SPEAKERLike, do you still think that's true?
18:33SPEAKEROr, I mean, has your foundation made enough inroads?
18:35SPEAKERAre you seeing, are you seeing those tides turn
18:38SPEAKERthrough your efforts or otherwise?
18:40SPEAKER- I think if you imagine there's this scalar,
18:43SPEAKERthis scalar value that we care about,
18:45SPEAKERwhich is number of applicants to the University of Cambridge,
18:47SPEAKERwhere the deadline is the 15th of October,
18:50SPEAKERsomething like that.
18:51SPEAKERSo on the 16th of October every year,
18:52SPEAKERyou're presented with a scalar number.
18:54SPEAKERThat value went from about 600 in 1999
18:57SPEAKERto about 200 to 250 in 2008.
19:00SPEAKERAnd last year, that scalar was about 1,450.
19:03SPEAKERSo we have a little over twice as many applicants
19:06SPEAKERto computer science at the University of Cambridge
19:07SPEAKERas we had at the height of the dot-com boom
19:09SPEAKERwhen everyone thought that computer science
19:11SPEAKERwas a meal ticket, right?
19:13SPEAKERSo if you just look at that number,
19:15SPEAKERyes, there has been an absolute sea change
19:18SPEAKERover the last 10 or 12 years.
19:20SPEAKERAnd I think Raspberry Pi has had a part in that.
19:23SPEAKERI think that the kids who rock up at least
19:25SPEAKERhave the good sense to pretend that Raspberry Pi
19:28SPEAKERwas the reason for this.
19:29SPEAKERFascinating thing for me as somebody
19:31SPEAKERwho grew up writing computer games on my BBC Micro
19:34SPEAKERand then on my Commodore Amiga,
19:36SPEAKERand had imagined that the primary use case
19:39SPEAKERfor Raspberry Pi in education would be software,
19:41SPEAKERthe vast majority of these people
19:42SPEAKERsay Raspberry Pi in robotics.
19:44SPEAKERSo the vast majority of these people
19:45SPEAKERare not having a purely software-mediated interaction
19:49SPEAKERwith Raspberry Pi.
19:50SPEAKERIt's about physical computing,
19:51SPEAKERwhich means it's very, very lucky.
19:53SPEAKERI mean, that's obviously somewhere
19:53SPEAKERwhere Raspberry Pi kind of stands out.
19:55SPEAKERIt's not something that I was super excited.
19:57SPEAKERThey're those general purpose IO pins on a Raspberry Pi,
20:00SPEAKERwhich allow you to do all those kind of cool
20:02SPEAKERgeneral purpose computer,
20:02SPEAKERthose cool physical computing things.
20:05SPEAKERThat was Pete Lomas,
20:06SPEAKERwho designed the first-generation Raspberry Pi hardware,
20:08SPEAKERwas very keen to expose some of the interface
20:11SPEAKERand capabilities of the chip we were using.
20:13SPEAKERAnd I didn't care enough to argue with him.
20:16SPEAKERAnd of course, how lucky we were both,
20:19SPEAKERof course, in the education space,
20:20SPEAKERbut also that was the capability
20:22SPEAKERthat unlocked the industrial use of Raspberry Pi,
20:24SPEAKERwhich really now accounts for the majority of our sales.
20:27SPEAKER- I was gonna say the politics of ports
20:30SPEAKERon these kinds of little dev boards is really interesting,
20:34SPEAKERbecause you have the entire spectrum now, right?
20:37SPEAKERYou have a development board
20:39SPEAKERthat has a bunch of USB and HDMI and audio outs
20:41SPEAKERand all the things you need to make it
20:43SPEAKERbasically a little tiny, either single serving server
20:46SPEAKERor single serving desktop or laptop or whatever you want.
20:50SPEAKERBut then you also have the compute modules,
20:51SPEAKERwhich let you just drop an SOC on a board
20:54SPEAKERthat somebody else designs
20:55SPEAKERand plug into whatever ins and outs
20:57SPEAKERthat they're looking for.
20:59SPEAKERAnd I'm curious how you think about that going now
21:02SPEAKERand how it's changed and what you're thinking
21:04SPEAKERabout going forward.
21:06SPEAKER- The nice thing with the core platform,
21:07SPEAKERso the SBC, the single board computer platform,
21:10SPEAKERthat's kind of classic Raspberry Pi,
21:11SPEAKERit's still the dominant volume runner for us out of seven.
21:15SPEAKERObviously last year was a rather volume constrained year,
21:17SPEAKERbut if you look back to 2021,
21:19SPEAKERwhen we sold 7 million Raspberry Pi computers of all sorts,
21:23SPEAKERthe SBC product, the single board computer product
21:25SPEAKERwith its four USB ports and ethernet connector,
21:29SPEAKERthat was between five and a half and 6 million units.
21:31SPEAKERAnd so compute module was on the order of a million,
21:34SPEAKERon the order of a million units.
21:36SPEAKERSo the SBC is the volume runner.
21:39SPEAKERWe're lucky that we've never had to really make
21:42SPEAKERany decisions to prioritize educational users
21:46SPEAKERof that product or industrial use of that product.
21:48SPEAKERThe decisions you make to make it a great educational product
21:51SPEAKERor also make it a great industrial product,
21:53SPEAKER'cause particularly the kind of robustness,
21:55SPEAKERthe reliability and robustness, they're very important.
21:57SPEAKERWe always say, what's the worst environment?
22:00SPEAKERWhat's the more challenging environment?
22:01SPEAKERAn oil rig or a kid's bedroom, they're both pretty,
22:04SPEAKERI mean, I've got two kids and I've never been on an oil rig,
22:07SPEAKERbut I've been to some power stations.
22:09SPEAKERAnd a kid's bedroom is a pretty tough environment.
22:12SPEAKERSo we kind of, there we've kind of always been able
22:14SPEAKERto kind of like have a happy medium.
22:17SPEAKERYou never have to choose one consumer over the other.
22:20SPEAKERWe have now compute module, which is purely targeted
22:22SPEAKERto the embedded and industrial space.
22:25SPEAKERAnd of course we have Pi 400,
22:26SPEAKERwhich is purely targeted at the consumer space.
22:28SPEAKERSo we have put our toes in the water in both directions.
22:31SPEAKERThe nice thing is actually both of those are kind of,
22:34SPEAKERmost of the investments we make are at the platform level.
22:37SPEAKERSo deriving those specific platforms from the core platform,
22:41SPEAKERfrom the SBC platform is actually quite cheap
22:43SPEAKERfrom a cost and an engineering perspective.
22:46SPEAKERSo we have tried to kind of,
22:47SPEAKERwe put this product in the market,
22:48SPEAKERin the middle of the market,
22:49SPEAKERand then we just kind of try to feel out
22:51SPEAKERwhat people don't like.
22:52SPEAKERAnd the main thing that people don't like
22:53SPEAKERas they go to scale with the SBC
22:55SPEAKERis they don't like being enslaved
22:57SPEAKERby our form factor choices, right?
23:00SPEAKERAnd you do see people who, you do see industrial,
23:03SPEAKERthat I have a keen eye for the appearance of that boat tail,
23:10SPEAKERthe two pairs of USB connectors and the ethernet jack.
23:14SPEAKERAnd if you look carefully at lots of industrial products,
23:16SPEAKERyou see that boat tail a lot, right?
23:18SPEAKERIt's the sign that someone's built a big object,
23:20SPEAKERan industrial object,
23:22SPEAKERand there's been enough space inside it.
23:23SPEAKERAnd so their solution for how do they embed a Raspberry Pi
23:25SPEAKERis they put it inside, they have a wiring loom or a hat
23:28SPEAKERthat sits on top of it and mounts their circuitry.
23:31SPEAKERThe challenge that people have is when they try to address
23:34SPEAKERmore consumer-oriented use cases,
23:37SPEAKERthey don't want to have to follow our choices
23:40SPEAKERabout onboard peripherals,
23:41SPEAKERthey don't have to follow our choices about connector layout
23:44SPEAKERand that's really where the Compute Module products,
23:46SPEAKERthey've been kind of a slow burn.
23:47SPEAKERWe've actually been doing Compute Modules since 2014.
23:50SPEAKERAnd it's really only the last couple of years,
23:52SPEAKERand I think with the introduction of Compute Module 4,
23:54SPEAKERthat that started to take off like a rocket.
23:56SPEAKERSo it's still a small part of our volume,
23:58SPEAKERbut by far the largest growing segment of our business.
24:02SPEAKER- I mean, given the lead time on new board designs
24:05SPEAKERand hardware 4, especially the,
24:07SPEAKERlike I think about Home Assistant
24:09SPEAKERand folks like that in this space
24:11SPEAKERand like the time it took them to design the board
24:13SPEAKERand all that, if, as that Compute Module factor,
24:16SPEAKERform factor becomes a little more static or stable,
24:19SPEAKERI think it'll help those projects ramp
24:22SPEAKERand lean into that kind of stuff faster and better maybe.
24:25SPEAKER- Yeah, and we've had two distinct form factors
24:27SPEAKERfor the Compute Module.
24:28SPEAKERSo until up to Compute Module 3, Compute Module 3 Plus,
24:31SPEAKERwe have an SODIMM form factor,
24:33SPEAKERwhere you put a JEDEC SODIMM socket down on your board
24:35SPEAKERand then kind of slide it in
24:37SPEAKERand then clip it in like a memory module.
24:39SPEAKERAnd then Compute Module 4, we threw that.
24:41SPEAKERAnd that is that we have continuity of,
24:43SPEAKERwe have both backwards and forwards compatibility
24:45SPEAKERin that space.
24:46SPEAKERAnd then Compute Module 4, we replace that with this,
24:48SPEAKERa pair of high density managed impedance connectors
24:51SPEAKERthat you clip down onto the board.
24:53SPEAKERAnd that's definitely been a good choice.
24:56SPEAKERIt's a more space efficient choice.
24:58SPEAKERIt's a more cost effective choice.
25:00SPEAKERI guess the other decisions I think we made
25:01SPEAKERwith Compute Module 4, which were informed by experience
25:04SPEAKERwith earlier modules and have really helped us.
25:07SPEAKERWe put Wi-Fi on the board as an option.
25:09SPEAKERI don't really, there's a platform,
25:12SPEAKERI'm not sure if it's defunct now called Android Things,
25:15SPEAKERwhich is effectively IoT platform built around
25:17SPEAKERthe Android kernel and bits of the Android user land.
25:20SPEAKERAnd I had a very revealing conversation
25:23SPEAKERwith a member of the Google Android Things team in Tokyo,
25:28SPEAKERI think back in about 2018.
25:29SPEAKERAnd he said, well, you really need to put Wi-Fi
25:33SPEAKERon your module because what it means is
25:34SPEAKERas soon as you want to go to scale,
25:36SPEAKERyou have to really abandon.
25:37SPEAKERYou either have to do a bunch of Wi-Fi engineering yourself
25:39SPEAKERor you kind of have to abandon the Raspberry Pi platform
25:42SPEAKERand the Compute Module platform for another platform.
25:46SPEAKERSo we put Wi-Fi on as an option.
25:48SPEAKERThat obviously eases a lot of people's integration.
25:50SPEAKERIt eases a lot of people's migration from the SPC platform,
25:53SPEAKERwhich has integrated Wi-Fi into the modular space.
25:56SPEAKERSo that was one thing.
25:56SPEAKERThe other thing we did was we massively simplified
25:58SPEAKERthe power chain.
26:00SPEAKERSo earlier Compute Modules,
26:01SPEAKERyou have to give them a bunch of rails.
26:03SPEAKERYou have to give them a 3V3 rail
26:05SPEAKERand at least a 3V3 and a 1V8 rail.
26:07SPEAKERAnd I think another rail as well.
26:09SPEAKERYou have to put circuitry on your baseboard
26:11SPEAKERthat delivers those rails and brings those rails up
26:13SPEAKERand takes them down in order, in the proper order
26:15SPEAKERso that nothing latches up.
26:16SPEAKERCompute Module 4, you give it five volts
26:18SPEAKERand it makes its own rails.
26:19SPEAKERIt has a power management IC on there.
26:20SPEAKERIt makes its own rails, in fact, it gives you the rails back.
26:22SPEAKERSo it'll give you, it'll give you a 3V3 rail
26:24SPEAKERfor you to use, as long as you don't draw
26:26SPEAKERtoo much power from it.
26:27SPEAKERThat's been a huge, I struggle, I think, sometimes
26:30SPEAKERwhen we're, when we're speccing and designing these
26:32SPEAKERto understand what the pain,
26:33SPEAKERwhere the pain points are going to be.
26:35SPEAKERAnd so we put these, these products in the market
26:37SPEAKERand sometimes we find that we've,
26:39SPEAKERwe've done something that people find painful
26:40SPEAKERand it limits the success of a product.
26:43SPEAKERThat was definitely a big,
26:45SPEAKERa big factor in the success of CM4.
26:47SPEAKER- So, you know, we're, we're largely
26:48SPEAKERan open source podcast here.
26:49SPEAKERSo I'm curious about open source practices,
26:51SPEAKERlarge and small.
26:52SPEAKERI'm curious, is there anywhere in the hardware
26:54SPEAKERand software stack that you use open source practices?
26:57SPEAKERFor example, can anybody contribute
26:58SPEAKERto core parts of Raspberry Pi OS?
27:01SPEAKEROr is it still like a relatively kind of closed
27:03SPEAKERor professional effort?
27:06SPEAKER- So if you think about the evolution
27:09SPEAKERof the Raspberry Pi software environment from 2011, 2012,
27:14SPEAKERthrough to the present day,
27:16SPEAKERI think what you're seeing is we have this thing
27:18SPEAKERthat's occasionally at the cycle stop on ELF
27:20SPEAKERand that is sometimes disparagingly called the blob,
27:23SPEAKERwhich is a large closed source software binary
27:27SPEAKERthat runs on the closed CPUs in the platform.
27:31SPEAKERSo it runs on the VPU, effectively a DSP.
27:33SPEAKERIt's actually a fairly powerful DSP,
27:35SPEAKERbut really these days we only use the scaler core.
27:37SPEAKERThere is a kind of system management processor
27:40SPEAKERand early days, everything ran on there,
27:42SPEAKEReverything multimedia related ran on there.
27:44SPEAKERSo camera processing, video decode, video encode,
27:47SPEAKERdisplay scan out, controlling the HDMI subsystem.
27:51SPEAKERDid I mention 3D graphics? 3D graphics.
27:53SPEAKERSo all of that stuff ran inside the blob.
27:55SPEAKERAnd then what you've seen over the years
27:57SPEAKERis a kind of carving away of,
27:59SPEAKERso where you have an open,
28:00SPEAKERwhere you say we have an open fairly famously,
28:02SPEAKERwhere you had an OpenGL API on the ARM side,
28:06SPEAKERwhat that was was effectively a serialization shim
28:09SPEAKERthat would take the core, serialize them,
28:10SPEAKERpass them over a FIFO and down into this blob,
28:12SPEAKERwhich would then do what it was told.
28:13SPEAKERActually, that's kind of great for OpenGL
28:15SPEAKER'cause OpenGL is designed for that exact,
28:17SPEAKERfor that, that's a very natural implementation style
28:21SPEAKERfor OpenGL, but it wasn't very pleasing to people
28:24SPEAKERbecause people couldn't see what was going on inside.
28:26SPEAKERAnd we were very lucky that I think in about 2014,
28:32SPEAKERwe got permission to release the architecture documentation
28:36SPEAKERfor the 3D graphics core,
28:38SPEAKERthe video core for graphics core in the chip.
28:40SPEAKERAnd we released this and this is pretty unusual actually
28:42SPEAKERto have good public document.
28:44SPEAKERI was involved in designing the core
28:45SPEAKERand it's pretty unusual to have good public documentation.
28:47SPEAKERThat still isn't really the case certainly,
28:49SPEAKERsay for the Mali cores that you see in a lot of other SOCs.
28:52SPEAKERAnd Emma Anholt joined my team at Broadcom
28:56SPEAKERbasically to build a Mesa driver stack
28:59SPEAKERaround this, using this documentation.
29:02SPEAKERAnd so from probably 2014, 2015 onwards,
29:05SPEAKERwe did have the option to turn off
29:07SPEAKERthe 3D graphics element of the blob and use Mesa instead.
29:11SPEAKERAnd then Emma went on to develop a driver for video core six,
29:14SPEAKERwhich is the core that we use in the modern Raspberry Pis
29:17SPEAKERand that's then being further improved by Galia.
29:19SPEAKERSo on the 3D graphics side,
29:20SPEAKERyou saw us turning off a bit of the blob.
29:23SPEAKERAnd then in the camera processing side,
29:26SPEAKERyou saw us turn off the camera processing
29:27SPEAKERand move across to lib camera.
29:29SPEAKERIt displays scan out, you saw us turning off that element
29:32SPEAKERand moving across to KMS.
29:33SPEAKERAnd so really all that's left actually is video encode.
29:36SPEAKERH.264 video encode still runs inside the blob.
29:39SPEAKERAnd that's really, I mean, it's great
29:40SPEAKERbecause obviously we can leverage the efforts
29:43SPEAKERof these enormous teams that have put together
29:44SPEAKERthese very, very powerful middleware systems.
29:47SPEAKERWe can leverage some of those efforts.
29:49SPEAKERWe can contribute back to those efforts.
29:51SPEAKERIf you look at our education mission,
29:52SPEAKERit means that there's no point,
29:55SPEAKERit's really undesirable for there to be a point
29:57SPEAKERwhere you have to say, here be dragons, don't look inside.
29:59SPEAKERDon't look inside this black box that does OpenGL.
30:01SPEAKERYou really want someone who's sufficiently inquisitive
30:03SPEAKERto be able to go all the way down
30:04SPEAKERand see when I asked, please draw this triangle,
30:06SPEAKERwhat actually happened deep down at the bottom of the stack
30:09SPEAKERand what do we think is happening inside the hardware?
30:11SPEAKERAnd I suspect that video encode is still there in the blob.
30:16SPEAKERI suspect probably the next or next but one generation
30:19SPEAKERof Raspberry Pi will get down to a point
30:21SPEAKERwhere there's effectively only clocks, power, and reset.
30:23SPEAKERAnd that's fairly traditional.
30:24SPEAKERI mean, Intel chips have the Intel management engine
30:26SPEAKERin there, you often have some little one,
30:28SPEAKERmost systems have some little microcontroller,
30:31SPEAKERwhich is handling, bringing the big cores out of reset,
30:34SPEAKEReffectively, and we'll probably get to that point.
30:36SPEAKERThat'll be really desirable.
30:37SPEAKERAnd hopefully we may even get, when we get to that point,
30:39SPEAKEReven though there will still be a binary
30:41SPEAKERand probably a signed binary down there
30:44SPEAKERthat does those things, you know,
30:45SPEAKERhopefully we'll even be able to publish the source
30:47SPEAKERfor that binary so people can even understand
30:49SPEAKERwhat that's doing.
30:50SPEAKER- Speaking of new functionality
30:52SPEAKERand adding new functionality to the core
30:54SPEAKERand to the CPUs and to the Pis downstream,
30:57SPEAKERhow does stuff like machine learning acceleration
30:59SPEAKERand the kind of the stuff that happens in tensor cores
31:01SPEAKERand stuff like that, how does that apply to Pi going forward?
31:03SPEAKER- So we've got an interesting philosophy
31:05SPEAKERabout how we decide what to put into a Raspberry Pi
31:07SPEAKERhardware platform.
31:08SPEAKERAnd it's broadly, you can summarize it as we divide,
31:11SPEAKERwe take the cost of the feature and we divide it through
31:15SPEAKERby the fraction of people who will find it useful.
31:19SPEAKERSo a feature, a $1 feature that everyone uses
31:22SPEAKERor a 10 cent feature that's used by a 10th of people
31:27SPEAKERget the same priority in our decision-making.
31:31SPEAKERThis is actually a really, really hard barrier
31:33SPEAKERfor new features to overcome, right?
31:35SPEAKERThe only really, if you look at a modern Raspberry Pi,
31:37SPEAKERthe only qualitatively new feature
31:39SPEAKERthat's in a Raspberry Pi today
31:40SPEAKERthat wasn't in the 2012 product is wireless,
31:43SPEAKERis Wi-Fi and Bluetooth.
31:44SPEAKERAnd that's the only thing that gives a $1 to $2 feature
31:48SPEAKERand it's useful to almost everybody,
31:50SPEAKERpretty much everything else.
31:51SPEAKERThere's no ADC, there's no analog to digital converter
31:53SPEAKERon the Raspberry Pi because it'll cost 10 to 20 cents
31:56SPEAKERand we don't think that actually the vast majority
31:58SPEAKERof our users need one.
31:59SPEAKERAnd so what that philosophy tends to do
32:02SPEAKERis it tends to push lesser used features
32:04SPEAKERoff onto external hardware.
32:07SPEAKERAnd that's always been good.
32:08SPEAKERThat's always been good to us as a decision-making heuristic.
32:12SPEAKERAnd so, yeah, if you wanna do some machine learning
32:14SPEAKERon a Raspberry Pi, actually for 1.8 gigahertz,
32:18SPEAKERA72s is quite a lot of machine learning throughput.
32:20SPEAKERSo a lot of people's machine learning requirements
32:23SPEAKERwill actually fit onto the CPUs.
32:25SPEAKERIf you wanna do more machine learning than that,
32:26SPEAKERyou can go and buy a coral,
32:28SPEAKERsay a Google coral accelerator, they're about 60 bucks.
32:30SPEAKERThey've got about two Tops of inference on there.
32:34SPEAKERAnd that applies to everything from analog
32:37SPEAKERto digital converters all the way up
32:38SPEAKERto machine learning accelerators.
32:40SPEAKERThat's been good to us because it's actually created space
32:42SPEAKERfor other people in our ecosystem to innovate.
32:45SPEAKERYou know, the fact that we don't try and slurp
32:47SPEAKERall of the functionality into the platform
32:49SPEAKERmeans that there is a business.
32:50SPEAKERYou know, you could start a business selling really nice
32:53SPEAKERdigital to analog converters that you connect
32:55SPEAKERto a Raspberry Pi and we won't come and drive over you.
32:57SPEAKERYou know, we won't come and add those features
32:59SPEAKERinto our platform.
33:01SPEAKER- It's interesting, sometimes what's not on there
33:03SPEAKERleads to as much kind of learning opportunity
33:05SPEAKERas what it is, right?
33:06SPEAKERLike the original model B that I bought,
33:07SPEAKERas you say, it doesn't have a DAC on it.
33:09SPEAKERAnd I would, first of all, I was like,
33:10SPEAKERwhy does the music coming out of this thing sound like this?
33:13SPEAKERBut second of all, it was like, oh,
33:14SPEAKERso that's what PWM audio is, right?
33:16SPEAKERIt's like, I had no idea that that existed as a concept
33:19SPEAKERuntil I had to work around it
33:20SPEAKERbecause of what was not on that board
33:22SPEAKERand had to go buy a little, you know,
33:23SPEAKER$5 USB DAC or whatever.
33:25SPEAKER- And of course, that was a wonderful learning experience
33:28SPEAKERfor us because we always kind of knew
33:29SPEAKERthat you can do better than PWM audio.
33:32SPEAKER- Well, even if the hardware you have
33:34SPEAKERis effectively a filtered pin,
33:36SPEAKERyou can do a lot better than just piping PCM audio
33:40SPEAKERinto the duty cycle register for a PWM block.
33:44SPEAKERYou can do all sorts of clever Sigma Delta
33:47SPEAKERnoise shaping stuff, you know, one bit DAC effectively.
33:50SPEAKERWhat if it was a HiFi would be market,
33:52SPEAKERif it was a HiFi in my childhood at least
33:53SPEAKERwould have been marketed as a, you know,
33:54SPEAKERPanasonic would market these things as one bit DACs, right?
33:57SPEAKERAnd what we've ended up with, you know,
33:58SPEAKERif you go back and you run a modern Raspberry Pi
34:01SPEAKERoperating system on the Raspberry Pi one,
34:02SPEAKERand we're quite proud of the year,
34:03SPEAKERthat's kind of like a core bit of our ethos
34:06SPEAKERthat you can run today's operating system
34:08SPEAKERon the 2012 hardware and it performs pretty well.
34:11SPEAKERActually you find the audio quality is pretty good
34:12SPEAKERand that's because it's picked up a lot of noise shaping,
34:15SPEAKERa lot of fairly alarming noise shaping code over the years.
34:19SPEAKER- So since you mentioned old Pis,
34:20SPEAKERI feel like one of the ways you can tell you're a nonprofit
34:23SPEAKERand not a corporation is that you're still supporting
34:25SPEAKERdecade old products.
34:26SPEAKERSo I'm curious, I mean, there's a bunch of questions there.
34:28SPEAKERI mean, the big one for me is, you know,
34:30SPEAKERuntil pretty recently,
34:31SPEAKERyou were offering one 32 bit operating system
34:33SPEAKERthat ran across the entire product line.
34:35SPEAKERAnd I'm sure like philosophically that was, you know,
34:37SPEAKERit's easier to support one platform, et cetera, et cetera.
34:40SPEAKERAnd then, you know,
34:40SPEAKERyou finally started offering a 64 bit option
34:43SPEAKERfor newer models.
34:44SPEAKERI mean, does that increase the support burden
34:46SPEAKERkind of exponentially or is it something you've taken
34:49SPEAKERin stride to now be having kind of different
34:52SPEAKERdifferent Raspberry Pi OSs?
34:53SPEAKER- It was a real wrench for us having two
34:56SPEAKERand we limited ourselves to two, sir.
34:57SPEAKERI mean, there's another obvious choice which would be three.
35:00SPEAKERSo you could have an ARMv6HF raspbian,
35:04SPEAKEReffectively image that ran on Raspberry Pi one
35:06SPEAKERand Raspberry Pi zero.
35:07SPEAKERYou could have a ARMv7HF, ARMhf in Debian speak,
35:15SPEAKERthat runs on Raspberry Pi two.
35:17SPEAKERAnd then you could have the ARM64 version
35:21SPEAKERthat runs on Raspberry Pi three and later.
35:23SPEAKERSo we decided not to do that.
35:25SPEAKERSo there's nothing, there's no special Raspberry Pi two
35:27SPEAKERCortex-A7 build.
35:29SPEAKER- Yeah, it was a bit of a wrench.
35:30SPEAKERI think because of the way that the Debian world works,
35:33SPEAKERactually not too much of a wrench.
35:35SPEAKERYou know, an enormous amount of effort goes into,
35:38SPEAKERyou know, if you sort of pick
35:40SPEAKERwhat couldn't we have done Raspberry Pi
35:43SPEAKERwithout obviously Debian
35:45SPEAKERand the way that Debian thinks about the world.
35:47SPEAKERAnd so the fact that a lot of effort goes into making
35:50SPEAKERthe system fairly architecture neutral
35:53SPEAKERallows us to support two fairly radically
35:57SPEAKERdifferent architectures.
35:58SPEAKEROf course, ARM32 and ARM64
36:00SPEAKERwith relatively little engineering effort.
36:02SPEAKER- Have you had any coordination with any other
36:04SPEAKERkind of operating system groups that are non Linux?
36:06SPEAKERI mean, like for example, like FreeBSD has been trying
36:08SPEAKERto bring ARM up to tier one support recently.
36:10SPEAKERLike I tried to install FreeBSD on a Pi two
36:12SPEAKERlike three years ago, and it was a rough experience.
36:15SPEAKERLike, do you have any contact with other groups
36:17SPEAKERthat are making efforts like that?
36:18SPEAKER- We have sporadic, we have sporadic contact.
36:20SPEAKERSo I think you can sort of divide operating system support
36:23SPEAKERfor Raspberry Pi into let's say four buckets.
36:29SPEAKERSo you've got Linux, various Linux distros.
36:33SPEAKERYou have Windows, and of course there are sporadic efforts
36:37SPEAKERfrom Microsoft and from the community
36:39SPEAKERto make various incarnations of Windows work on Raspberry Pi
36:42SPEAKERsome pretty successful, some fairly ropy.
36:45SPEAKERYou have, so that was two, you have the BSDs
36:49SPEAKERand we do have some light touch interaction
36:52SPEAKERwith the various BSD ecosystems.
36:54SPEAKERGenerally trying to answer, trying to answer questions
36:58SPEAKERand also trying to provide where we can software
37:01SPEAKERthat interfaces with the various subsystems on the device
37:05SPEAKERwhich isn't GPL licensed because of course
37:08SPEAKERthey have a challenge and there are various subsystems
37:11SPEAKERwhere they can't really take bits of Linux kernel
37:13SPEAKERwhere kind of you're close to Linux,
37:16SPEAKERthe Linux kernel drivers being the reference,
37:19SPEAKERthe software reference architecture,
37:21SPEAKERbut of course they're inappropriately licensed
37:22SPEAKERand so we occasionally do work
37:24SPEAKERto try to help people with that problem.
37:26SPEAKERAnd then you have other operating systems
37:28SPEAKERand that's everything from, you know,
37:29SPEAKERvarious research operating systems to a RISC OS.
37:33SPEAKEROf course, you know, we've always liked the fact
37:34SPEAKERthat you could run the original arm operating system
37:37SPEAKERon Raspberry Pi and we do have a very friendly,
37:40SPEAKERa very friendly relationship with RISCOS Open
37:41SPEAKERwho are just up the road from us.
37:43SPEAKER(upbeat music)
37:47SPEAKER- So to change the topic a little bit,
37:49SPEAKERlast couple of years have been challenging
37:51SPEAKERfor a lot of folks with pandemic supply stuff.
37:54SPEAKERAnd I'm curious, I mean, I think we all follow
37:57SPEAKERRaspberry Pi locator here.
37:58SPEAKERSo it seems like things are getting a little bit better
38:01SPEAKERbut I'm curious how you're feeling about where you're at
38:04SPEAKERand what the future holds
38:06SPEAKERand if there's stuff you would do differently
38:07SPEAKERor if this was just an unavoidable side effect
38:10SPEAKERof, you know, a worldwide pandemic.
38:13SPEAKER- I think if I had my time again,
38:16SPEAKERyeah, I would, you know,
38:17SPEAKERwhat wouldn't I give today for a million, you know,
38:21SPEAKERfor an extra million?
38:22SPEAKERWell, of any of the Raspberry Pi chips, actually,
38:24SPEAKERI'm assuming 2711, which is what we used to build Pi 4s.
38:27SPEAKERBut what would I give for another million
38:28SPEAKERof any of those chips?
38:30SPEAKERI, and of course it should have been obvious, right?
38:37SPEAKERYou know, what was the core?
38:39SPEAKERThe core dynamic was a pandemic happened.
38:43SPEAKEREveryone assumed there would be a massive recession.
38:47SPEAKERSo stopped ordering businesses,
38:48SPEAKERassumed there'd be a massive recession.
38:49SPEAKERSo they went into de-stocking mode.
38:51SPEAKERAt the same time, everyone was at home
38:53SPEAKERneeding electronic equipment
38:55SPEAKERto continue to participate in society.
38:57SPEAKERAnd also in most, certainly European North America,
39:02SPEAKERhaving their income protected by various government schemes
39:06SPEAKERand also unable to spend money on services.
39:08SPEAKERSo you have the same income,
39:10SPEAKERbut you can no longer go on holiday or go to a restaurant.
39:12SPEAKERSo what are you gonna do?
39:13SPEAKERYou're gonna buy consumer durables.
39:14SPEAKERAnd so it was that pair of transients,
39:16SPEAKERthe negative going supply transient
39:18SPEAKERand the positive going demand transient were,
39:20SPEAKERyou know, if you sat someone down in a business school class
39:23SPEAKERand you described the situation to them in those terms,
39:26SPEAKERof course they would tell you what was gonna happen.
39:28SPEAKERBut at the time when we were in the thick of it,
39:29SPEAKERand of course we were,
39:30SPEAKERa lot of us were focusing on not getting COVID,
39:32SPEAKERnobody called this one right.
39:35SPEAKERYou know, the big car companies didn't call this one right.
39:38SPEAKERSo yeah, if I had my time again,
39:39SPEAKERobviously I would, what would I like?
39:43SPEAKERI would maybe have liked somebody to talk me through it
39:46SPEAKERin business school terms,
39:46SPEAKER'cause then I could have spotted what I needed to do.
39:49SPEAKERAnd so towards the end of 2020,
39:51SPEAKERyou see things start to come apart.
39:53SPEAKERSo you see first blowouts in component lead times.
39:57SPEAKERSo components that you would expect to be able to get on,
39:59SPEAKERon 20 week lead times, going out to 52 week lead times.
40:02SPEAKERWe were pretty quick in getting orders on
40:06SPEAKERin that environment.
40:07SPEAKERBut obviously if lead times instantaneously blow out by,
40:11SPEAKERsay three months in one day,
40:14SPEAKERit's inevitably going to create a three month void
40:17SPEAKERin your supply chain.
40:18SPEAKERThere's no way you can get orders on fast enough
40:20SPEAKERto fill that three months.
40:21SPEAKERIt's just an instantaneous, instantaneous hole.
40:24SPEAKER- And it's not like there's parts,
40:26SPEAKERit's not like you have warehouses full of parts in China
40:28SPEAKERsitting, waiting to go.
40:30SPEAKERStuff comes in and gets stuck on boards
40:32SPEAKERand the boards go out, right?
40:33SPEAKER- Yeah, and you have some inventory.
40:34SPEAKERI mean, so you have some component inventory
40:36SPEAKERand at any given day you have some component inventory
40:38SPEAKERand you have some finished goods inventory
40:39SPEAKERand your downstream channel has some inventory
40:42SPEAKERand industrial customers of that downstream channel
40:44SPEAKERhave some inbound inventory,
40:47SPEAKERwhich is your finished product, right?
40:49SPEAKERSo there's lots of inventory sloshing around in the system,
40:52SPEAKERbut no, it's certainly not the case
40:54SPEAKERthat you have millions of units of every component
40:57SPEAKERthat you need to build a Raspberry Pi.
41:01SPEAKERSo you have this blow out.
41:03SPEAKERThen once you get back into a situation
41:05SPEAKERwhere your orders that you've placed
41:06SPEAKERthat were now 52 week lead time arrive,
41:08SPEAKERyou end up in an allocation situation.
41:10SPEAKERSo you end up in a situation
41:11SPEAKERwhere even though you have orders on,
41:13SPEAKERyou don't get supplied at one-to-one ratio
41:17SPEAKERbecause there is more the total aggregate demand
41:19SPEAKERexceeds the total aggregate production capacity.
41:21SPEAKERAnd so the only thing you can really do
41:22SPEAKERis to give everyone a haircut.
41:24SPEAKERSo that was the kind of inbound situation for us,
41:27SPEAKERwhich has persisted really up to the present moment.
41:31SPEAKERIt's a situation which is now improving quite rapidly.
41:34SPEAKERAnd I think what we said when we gave an update in December
41:37SPEAKERwas broadly the first calendar quarter of this year
41:40SPEAKERwas pretty tough.
41:42SPEAKERYou know, it has some bright spots
41:43SPEAKERand I think if you go up by locator,
41:44SPEAKERyou'll see what some of those bright spots are right now.
41:46SPEAKERIt's things like Raspberry Pi 3A Plus.
41:48SPEAKERLater in this quarter, we'll see Raspberry Pi zero.
41:51SPEAKERWe'll see a significant improvement
41:53SPEAKERin Raspberry Pi zero availability,
41:54SPEAKERwhich is of course kind of cool
41:55SPEAKER'cause it's still my favorite Raspberry Pi product
41:57SPEAKERat the zero and zero wireless.
41:59SPEAKERSo we'll see some improvement
42:00SPEAKERover the course of this quarter,
42:01SPEAKERbut it's still a constraint quarter.
42:03SPEAKERNext quarter is broadly speaking a pre-pandemic quarter.
42:07SPEAKERSo you can see next quarter as being
42:09SPEAKERone and a half to two million units of inbound inventory
42:12SPEAKERacross the various components that we use.
42:14SPEAKERBut because you have backlogs,
42:17SPEAKERyou need to do better than that.
42:18SPEAKERSo that's enough to keep up,
42:19SPEAKERbut it's not enough to start to get back.
42:22SPEAKERSo I think what you'll see in the second quarter,
42:24SPEAKERmuch better availability,
42:25SPEAKERbut a continuation of things like single unit limits
42:28SPEAKERfor consumers and what we call active management
42:31SPEAKERfor industrial customers,
42:32SPEAKERwhere a customer orders something
42:36SPEAKERand rather than just taking their order,
42:38SPEAKERwe go back to them and we actively say to them,
42:40SPEAKERis that, do you really need that many units?
42:42SPEAKERIt's got anti-sales.
42:43SPEAKERWe have sales people here doing anti-sales.
42:46SPEAKERSo can I have 10,000 Raspberry Pis?
42:47SPEAKERWell, would 1,000 Raspberry Pis
42:49SPEAKEReach month for 10 months be enough, you know?
42:51SPEAKER'Cause you're trying to detect inventory building.
42:52SPEAKERWell, we call it kind of toilet roll hoarding
42:54SPEAKERin the context of the pandemic
42:57SPEAKERbecause that will destroy,
42:58SPEAKERthat even if you have a sufficient supply,
43:02SPEAKERthat kind of begging your neighbor behavior,
43:03SPEAKERwhich is totally understandable and totally defensible.
43:05SPEAKERAnd I would absolutely be doing it
43:06SPEAKERif I was buying these units.
43:10SPEAKERYou have to find ways to help people.
43:14SPEAKERIt's not about stopping people doing it.
43:16SPEAKERIt's about making your industrial customers comfortable
43:19SPEAKERthat if they don't do it, they'll be fine,
43:23SPEAKERthat we'll look after them,
43:24SPEAKERthat either they can disclose their true demand to us
43:27SPEAKERand that we will use that in a good way
43:29SPEAKERto help make sure that they don't go lines down
43:31SPEAKERrather than in a bad way to kind of give them a haircut.
43:34SPEAKERSo you'll see that though,
43:35SPEAKERthose behaviors will persist through Q2,
43:36SPEAKERbut it will be a good environment,
43:38SPEAKERbut it will be a heavily managed environment.
43:40SPEAKERAnd then from Q3 onwards this year,
43:41SPEAKERwe effectively have an unlimited supply
43:43SPEAKERof pretty much everything we need to make Raspberry Pis.
43:46SPEAKERThat will allow us to catch up on our backlogs
43:48SPEAKERand then get back into the situation we wanna be in,
43:50SPEAKERwhich is where we don't actually need
43:52SPEAKERto touch our customers so much.
43:54SPEAKERWe can just put units in channel.
43:55SPEAKERThe dream with Raspberry Pi,
43:56SPEAKERwhere we were for out of 11 years,
44:00SPEAKERfor nine out of 11 years in the market,
44:02SPEAKERwe've been in a situation where there are
44:04SPEAKERhundreds of thousands to millions of units
44:06SPEAKERof Raspberry Pi in channel.
44:07SPEAKERAnd if you wake up one morning as an industrial,
44:10SPEAKERas a hobbyist and you want 10, you can have 'em.
44:12SPEAKERIf you wake up as an industrial customer
44:13SPEAKERand you want 100,000, you can have 'em.
44:16SPEAKERAnd that kind of law of large numbers averaging effect
44:20SPEAKERmeans that actually not many people wake up
44:23SPEAKERon a given day and want 100,000 units.
44:25SPEAKERAnd therefore you can pull from the channel
44:27SPEAKERand it's as if there's an infinite supply of Raspberry Pis.
44:29SPEAKER- Yeah.
44:30SPEAKERYou just said that the Pi Zero is your favorite.
44:32SPEAKERI'm curious why.
44:33SPEAKER- I just, it's the, you can think of Raspberry Pi
44:37SPEAKERas having explored another use of Moore's law.
44:40SPEAKERSo the sort of classic PC industry use of Moore's law
44:43SPEAKERis to pick a price point.
44:44SPEAKERAnd in every, let's say 18 months, two years,
44:46SPEAKERfill that price point up with twice as much compute.
44:49SPEAKERAnd so you get an exponentially increasing amount
44:51SPEAKERof compute at a given price.
44:53SPEAKERThe other thing you could do with Moore's law
44:54SPEAKERis you can pick an amount of compute
44:56SPEAKERand you can ride an exponentially decreasing price curve.
44:59SPEAKERAnd you can sort of see Raspberry Pi
45:01SPEAKERas a piece of hardware having said in 2010, 2011, 2012,
45:06SPEAKERhey, what would a 2000 era PC cost now actually
45:10SPEAKERbecause of Moore's law?
45:11SPEAKEROh, it's $35.
45:12SPEAKERAnd that was kind of great.
45:13SPEAKERAnd then we launched Raspberry Pi 2 in 2015.
45:15SPEAKERAnd what did we do?
45:15SPEAKERWe picked $35 and we filled it
45:17SPEAKERwith an exponentially increasing amount of compute.
45:19SPEAKERWe'd actually just gone back to doing what,
45:21SPEAKERat a different price point,
45:22SPEAKERwe've gone back to doing what everybody else does
45:24SPEAKERwith Moore's law.
45:24SPEAKERAnd really Raspberry Pi Zero, I love it
45:26SPEAKERbecause it's an example of a few months later
45:30SPEAKERrealizing that we'd conformed
45:34SPEAKERand then asking ourselves,
45:35SPEAKERwell, what if we didn't conform?
45:36SPEAKERWhat if we said, let's take the 2000?
45:38SPEAKERWhat now what does a 2000 era PC cost?
45:41SPEAKEROh, it costs five bucks or 10 bucks.
45:43SPEAKERAnd so I like it for that.
45:45SPEAKERAnd the projects you get within a very,
45:47SPEAKERthey're very kind of embedded.
45:50SPEAKERIt shows up in some of my favorite hobbyist projects
45:52SPEAKERlike the high altitude ballooning because it's very light.
45:55SPEAKERIt shows up in some of those projects.
45:57SPEAKERSo I do love it.
45:58SPEAKERAnd we've sold a lot of them.
45:59SPEAKERI mean, it's been, I don't know what the total numbers are
46:03SPEAKERbut it's probably three or 4 million.
46:04SPEAKEROut of the 50 million,
46:05SPEAKERroughly 50 million Raspberry Pis we've sold,
46:07SPEAKERprobably three or 4 million of them
46:09SPEAKERhave been these tiny little computers.
46:11SPEAKER- It lives in kind of a fun spot
46:13SPEAKERbetween like a microcontroller and a full blown PC, right?
46:17SPEAKER- Yeah, and of course we've explored with Pico
46:19SPEAKERwhich came out a couple of years ago,
46:21SPEAKERwhich has our own Silicon on it.
46:23SPEAKERIt's the first product
46:24SPEAKERthat has Raspberry Pi designed Silicon on it.
46:26SPEAKERBut we kind of then have explored what does the world below
46:29SPEAKERzero look like?
46:30SPEAKERAnd we have this $4 Pico product and $6 Pico W product.
46:34SPEAKERJust all trying to understand, how low can we take?
46:38SPEAKERAnd then the sock, which you can also buy the microcontroller
46:41SPEAKERfrom us and this is a 50 cent microcontroller.
46:43SPEAKERSo, how low can we take that general purpose
46:46SPEAKERcomputing mission?
46:47SPEAKER- One of the many things that's always fascinated me
46:49SPEAKERabout your efforts is how long old models
46:51SPEAKERremain in production.
46:52SPEAKERI mean, the Pi one and two you've committed through 2026
46:56SPEAKERto keep making them.
46:56SPEAKERI'm curious, when the supply constraints really kicked in,
46:59SPEAKERI mean, everybody wants to buy the latest and greatest
47:01SPEAKERof course, but like there's plenty of hobbyist projects
47:03SPEAKERthat run just fine on a three or a two,
47:06SPEAKERor my old model B is still sitting here
47:07SPEAKERdoing something useful.
47:09SPEAKERDid those old models being still in the channel,
47:11SPEAKERdid that take on kind of a renewed importance
47:12SPEAKERwhen nobody could get fours and CM fours anymore?
47:16SPEAKERLike did those have an uptick?
47:18SPEAKER- It certainly helped that we had access
47:20SPEAKERto two process notes.
47:21SPEAKERSo we had core Silicon across two process notes,
47:24SPEAKER40 nanometers for Pi three and earlier and 28 nanometers.
47:28SPEAKERSo you kind of, yeah,
47:29SPEAKERbeing exposed to more process notes is good.
47:31SPEAKERIf I had my way,
47:32SPEAKERI would have been exposed to three or four process notes,
47:33SPEAKERbut yeah, two was certainly better than one.
47:35SPEAKERYeah, we saw people migrating in both directions.
47:38SPEAKERSo we saw people driven by month to month
47:40SPEAKERavailability behavior.
47:42SPEAKERSo we saw people who were using Pi four era hardware
47:46SPEAKERlooking to see whether applications could be optimized
47:49SPEAKERto run on early hardware.
47:50SPEAKERWe saw people who were using compute module three migrate.
47:53SPEAKERWe have a, we have a sort of a,
47:55SPEAKERwhat's currently a pseudo product called the CM four S
47:58SPEAKERcompute module four S,
47:59SPEAKERwhich is the brains of compute module four
48:02SPEAKERin the compute module three form factor.
48:04SPEAKERAnd that was developed before the pandemic
48:06SPEAKERas a way to give people a migration path
48:08SPEAKERup to the modern Silicon platform
48:10SPEAKERwithout having to retool their baseboards.
48:12SPEAKERBut it actually ended up doing duty as a way for,
48:16SPEAKERin the periods when we had better 28 nanometer availability
48:19SPEAKERand the 40 nanometer availability,
48:21SPEAKERit did duty as a way for people to get continuity of supply.
48:25SPEAKERThere's a product that it is on our,
48:27SPEAKERit is a real and official Raspberry Pi product,
48:30SPEAKERbut it has never really been launched.
48:32SPEAKERIt's on our website, but it's never really been launched.
48:35SPEAKERIt's only been used as a sort of a tactical aid
48:38SPEAKERto some of our industrial customers.
48:41SPEAKERYeah, it's been, it's interesting.
48:42SPEAKERAnd even today, as I say, even today,
48:44SPEAKERyou're gonna see us come back into availability
48:47SPEAKERand kind of broadly the order in which our products were launched.
48:52SPEAKERSo you're gonna see the older, the 3A+ come in,
48:54SPEAKERyou see 0W come in, and then probably a Pi 4
48:58SPEAKERwill be towards the end of the consumer,
49:01SPEAKERunrestricted volume consumer availability.
49:04SPEAKERRaspberry Pi 4 will probably be towards the end
49:05SPEAKERof the transition.
49:07SPEAKER- Are you all in on ARM for the foreseeable future?
49:09SPEAKERDo you have any thoughts on RISC-V, for example,
49:12SPEAKERwhich is making a lot of headlines as an open source
49:15SPEAKERinstruction set in the kind of the same embedded space?
49:17SPEAKER- It's very hard for me not to be all in on ARM, right?
49:19SPEAKERAs somebody who spent his childhood drooling
49:22SPEAKERover economic communities and as a former colleague
49:25SPEAKERof Sophie Wilson, of course, who was responsible
49:27SPEAKERfor the design of that instruction set architecture.
49:30SPEAKERIt's very hard for me not to be all in on ARM.
49:32SPEAKERI think RISC-V, RISC-V is an interesting development.
49:35SPEAKERPeople, it's important not to understate the challenges
49:39SPEAKERthat a new instruction set architecture face
49:42SPEAKERin terms of adoption.
49:43SPEAKERThere's an enormous amount of accumulated
49:47SPEAKERinvested engineering effort in the ARM architecture
49:50SPEAKERand the Linux kernel and the various user lands
49:55SPEAKERthat run on top of it.
49:56SPEAKEROptimization, so you go look at the FFmpeg source code,
49:59SPEAKERa huge number of neon optimizations for 32 and 64-bit ARM.
50:04SPEAKERActually, that accumulated base of engineering investment
50:07SPEAKERpales in comparison to the even more enormous level
50:10SPEAKERinvestment in the Intel architecture.
50:12SPEAKERIf you go to some of these places,
50:14SPEAKERyou find fast paths for everything from NNX to SSE,
50:18SPEAKERto SSE2, onwards, right?
50:20SPEAKERA whole museum of Intel multimedia instructions.
50:24SPEAKERSo, and that stuff that needs to be replicated
50:27SPEAKERin the RISC-V space.
50:28SPEAKERIt's not to say it won't be replicated there,
50:29SPEAKERbut it does need to be replicated
50:31SPEAKERand that would be a necessary precursor
50:34SPEAKERto us being able to do any sort
50:35SPEAKERof RISC-V based Raspberry Pi product.
50:37SPEAKERSo I'm very satisfied with the ARM architecture
50:41SPEAKERis the ARM ecosystem is quite a good place to hang out.
50:46SPEAKERAnd it's notable that other,
50:48SPEAKERquite a lot of other computer companies
50:50SPEAKERhave chosen to go hang out there recently.
50:51SPEAKERAnd just as the Linux kernel provides a venue
50:56SPEAKERfor cooperative collaborative innovation.
50:58SPEAKERSo, the ARM instruction set and the various cores
51:02SPEAKERthat you can license that execute that instruction set,
51:05SPEAKERthey provide the house a similar sort of,
51:08SPEAKERit's a similar environment.
51:09SPEAKERAnd so there's a similar source of scale advantage.
51:12SPEAKERSo it's gonna be interesting to see what happens
51:13SPEAKERwith RISC-V.
51:14SPEAKERI think in the microcontroller space,
51:16SPEAKERthose advantages are much less pronounced.
51:19SPEAKERAnd I think where you do see penetration
51:21SPEAKERand I think you'll continue to see penetration
51:23SPEAKERof the RISC-V instruction set architecture
51:25SPEAKERis in microcontrollers and deep embedded.
51:27SPEAKER- Okay, I was gonna say,
51:28SPEAKERone of the things we talk about on the FOSS pod a lot
51:30SPEAKERwith smaller projects is funding
51:33SPEAKERand the organization of the, of the organization.
51:36SPEAKERHow the business of making these large open source projects
51:39SPEAKERworks and I'm, I mean, I think the funding
51:42SPEAKERis probably pretty obvious,
51:43SPEAKERbut I bet there's probably stuff
51:44SPEAKERthat people don't understand.
51:45SPEAKERAnd I'm curious what the organizational structure is like
51:47SPEAKERand what your kind of day-to-day involvement
51:49SPEAKERand what your job started as and how it, what it is now.
51:54SPEAKER- The, so the funding, the funding is interesting.
51:57SPEAKERThe foundations are Topco's a charity.
51:59SPEAKERThe thing I run is Raspberry Pi Limited.
52:02SPEAKERSo I run the engineering subsidiary of the foundation.
52:05SPEAKERAnd so we do the engineer, if you buy Raspberry Pi computer,
52:07SPEAKERit was designed by us and the software was written,
52:10SPEAKERthe Raspberry Pi specific bits of software written by us
52:12SPEAKERand the brand is maintained by us.
52:14SPEAKERSo we do that bit and then we make money
52:17SPEAKERand then we give that money.
52:18SPEAKERSo we reinvest some of the money
52:19SPEAKERin making the Raspberry Pi stuff
52:21SPEAKERand then we give the surplus money
52:22SPEAKERto the foundation that does educational things with it.
52:25SPEAKERAnd we've returned on the order of 35 million pounds,
52:28SPEAKERI think over the years to the foundation
52:30SPEAKERto fund its educational, its educational work.
52:32SPEAKERSo that's the kind of structure.
52:34SPEAKERWhere did the money come from in the first place?
52:35SPEAKERA handful of us put money in,
52:37SPEAKERI think we put in about a hundred thousand.
52:40SPEAKERThe founding trustees of the foundation
52:43SPEAKERlent the foundation about a hundred thousand pounds
52:45SPEAKERbetween us, actually for a period of very few months.
52:48SPEAKERAnd then those loans were then repaid.
52:50SPEAKERSo that's for the initial capital.
52:51SPEAKERAnd actually until 2000 until 2021,
52:56SPEAKERthat was until the autumn of 2021,
52:57SPEAKERthat was the only money that had gone into Raspberry Pi.
53:00SPEAKERSo the whole of Raspberry Pi was bootstrapped off that money
53:04SPEAKERand probably about 30 or 40,000 pounds of donations
53:07SPEAKERfrom high net worth individuals in Cambridge
53:09SPEAKERwho thought it was a good idea.
53:11SPEAKERSo that's kind of fun.
53:13SPEAKERAnd then we did a $45 million funding round
53:15SPEAKERin the autumn of 2021.
53:18SPEAKERA couple of absolutely fantastic supporters of ours
53:21SPEAKERput in $45 million into the trading business,
53:24SPEAKEReffectively bought new shares in the trading business
53:26SPEAKERand the trading business then ceases to be a hostel
53:28SPEAKERvery much a majority owned subsidiary of the foundation,
53:31SPEAKERbut no longer a wholly owned subsidiary.
53:33SPEAKERBut that's the only external money
53:35SPEAKERwe've actually brought in.
53:36SPEAKERIt turns out how my job has evolved,
53:38SPEAKERI guess it's evolved from kind of overseeing everything,
53:42SPEAKERevery aspect of Raspberry Pi early on
53:44SPEAKERthrough to very much that focus on the commercial
53:46SPEAKERand engineering side of it, which is natural for me
53:49SPEAKER'cause I'm a software engineer originally,
53:51SPEAKERI'm a chip engineer subsequently.
53:54SPEAKERAnd so kind of having the opportunity really to focus in
53:58SPEAKERon that aspect of things has been great.
54:01SPEAKERWe found ourselves a fantastic chief executive
54:03SPEAKERfor the foundation, Philip Colligan,
54:04SPEAKERwho poached from Nesta, the National Endowment
54:06SPEAKERfor Science, Technology and the Arts in the UK.
54:09SPEAKERHe was deputy chief executive at Nesta
54:10SPEAKERand he joined us in 2015.
54:13SPEAKERI think he was about for the kind of the early years
54:15SPEAKERof Raspberry Pi actually, the engineering function was big
54:18SPEAKERand the charitable function was small.
54:20SPEAKERAnd so Philip was probably only at that point
54:22SPEAKERabout the fourth employee here of the foundation
54:26SPEAKERand then grew that very, very quickly
54:28SPEAKERthrough both organic growth and acquisition
54:30SPEAKERto about 150 or organic growth and merger
54:33SPEAKERto about 150 people.
54:34SPEAKERSo that's the foundation to us.
54:36SPEAKERThe headcount's now rather bigger than us.
54:38SPEAKERAnd he stayed with us, I mean, he's been with us,
54:40SPEAKERhe will have been with us eight years this year
54:42SPEAKERand he's transformed it into,
54:45SPEAKERI think he's transformed the foundation into something
54:47SPEAKERthat we didn't envisage when we started.
54:52SPEAKERThe existence of the Raspberry Pi computer
54:53SPEAKERwas supposed to be the contribution, right?
54:55SPEAKERWhen now we have a foundation, which obviously
54:58SPEAKERwe continue to ensure the Raspberry Pi computers exist,
55:01SPEAKERbut the foundation does much more.
55:03SPEAKERConventional is kind of conventional is not the way.
55:05SPEAKERConventional sounds disparaging,
55:06SPEAKERit's not really the right word,
55:08SPEAKERbut they do much more regular charitable,
55:11SPEAKERthe sorts of things you would imagine
55:12SPEAKERthat educational charity would do.
55:14SPEAKERThey train teachers and they create,
55:15SPEAKERthey train teachers and they run clubs
55:17SPEAKERand they create courseware and things like that.
55:20SPEAKERAnd so it's kind of wonderful to have
55:21SPEAKERboth that original strand of the kind of anarchic strand
55:24SPEAKERof just making sure general purpose computing is everywhere
55:26SPEAKERand then also to take some of the money you've made from that
55:28SPEAKERand then go and do more traditional interventions
55:31SPEAKERin the space.
55:32SPEAKER- But for the most part, the day-to-day
55:33SPEAKERof both the charity and the charitable foundation
55:36SPEAKERand the trading company are funded
55:38SPEAKERby obviously sales of hardware, right?
55:40SPEAKER- Yes, that's right, yes, yes, which is cool, right?
55:44SPEAKERWe made a lot, we sold 50 million computers
55:46SPEAKERand we don't make very much money when we sell out,
55:48SPEAKERwe don't make very much profit when we sell a Raspberry Pi,
55:50SPEAKERbut if you sell 50 million of them,
55:52SPEAKERthen you do make a bit of money.
55:54SPEAKER- I was gonna say like the idea,
55:55SPEAKERlike it sounds like a lot of money
55:56SPEAKERwhen we talk about it in normal human terms,
55:58SPEAKERbut the idea that you bootstrapped a company
56:00SPEAKERthat sold 50 million computers on $100,000
56:03SPEAKERis kind of shocking, right?
56:05SPEAKERLike it's an incredible thing.
56:07SPEAKER- I mean, to some extent it's a testament
56:09SPEAKERto the business model.
56:10SPEAKERFor the longest time we were a licensing company,
56:12SPEAKERso we're taking a leaf actually out of arms book.
56:15SPEAKERWe designed the Raspberry Pi
56:18SPEAKERand we licensed that design
56:20SPEAKERand the use of the Raspberry Pi brand
56:22SPEAKERand obviously the community behind it
56:25SPEAKERto two partners who then made Raspberry Pis
56:27SPEAKERand paid us a royalty.
56:30SPEAKERAnd the nice thing about that business
56:31SPEAKERis as you grow a traditional manufacturing business,
56:33SPEAKERyou kind of see it sucks capital in
56:35SPEAKERbecause you need to fund inventory.
56:37SPEAKERYour, as we call the working capital loop expands,
56:42SPEAKERyou need to fund all of that inventory,
56:45SPEAKERcomponent inventory, finished goods inventory,
56:47SPEAKERreceivables waiting for your customers to pay.
56:49SPEAKERYou need money as your business gets bigger,
56:51SPEAKERso you suck capital in.
56:53SPEAKERBecause we were a licensing company,
56:55SPEAKERthe unit economics were a little bit less good
56:56SPEAKER'cause you're only getting a royalty
56:57SPEAKERrather than getting the full margin on the product.
57:01SPEAKERBut it does mean you don't need to suck money in
57:02SPEAKERand that's why the foundation was able
57:04SPEAKERto retain 100% ownership of Raspberry Pi limited
57:07SPEAKERwhile it grew and thus we're able
57:10SPEAKERto devote all of the entirety of the spare money,
57:13SPEAKERthe entirety of the surplus profit.
57:15SPEAKERWe were able to devote the entirety of that
57:17SPEAKERto funding the foundation rather than having to share it
57:20SPEAKERbetween the foundation and people who had come in over time
57:23SPEAKERto contribute capital to the growth of the organization.
57:25SPEAKER- So then that raise in 2021 was to transition
57:29SPEAKERoff of the licensing?
57:30SPEAKER- It was certainly to transition to more of a blended model.
57:33SPEAKERWe are a much more blended organization now.
57:35SPEAKERSo certainly last year, really a substantial majority
57:39SPEAKERof Raspberry Pi computers were actually made by us
57:41SPEAKERor we commissioned the manufacturing.
57:43SPEAKERNobody makes anything right.
57:44SPEAKERSo we paid some people to make it for us.
57:46SPEAKERBut yeah, it was a traditional company
57:48SPEAKERwith us inside the working capital.
57:50SPEAKERThat's obviously capital intensive.
57:51SPEAKERAlso a certain amount of that money has obviously gone
57:53SPEAKERto pay for sort of advanced R&D stuff
57:56SPEAKERbecause of course we continue to try
57:57SPEAKERto make cool new Raspberry Pi things and that's not cheap.
58:00SPEAKER- It's kind of remarkable to me to just sit here
58:02SPEAKERlistening to you talk about this
58:03SPEAKERbecause 50 million units in a decade is a number
58:06SPEAKERthat any kind of publicly traded corporation
58:08SPEAKERwould salivate over, right?
58:09SPEAKERBut yeah, you're working with pretty thin profit margins.
58:12SPEAKERThe money mostly is going back into like fairly noble
58:15SPEAKERkind of educational outreach and so forth.
58:17SPEAKERI mean, I don't really have a question here.
58:19SPEAKERIt's just kind of, this is almost like a unicorn
58:21SPEAKERof an operation to see something at this scale
58:24SPEAKERthat is not trying to maximize shareholder value constantly.
58:27SPEAKER- It's great, right?
58:28SPEAKERIt's a wonderful, and most of us who are involved
58:30SPEAKERin Raspberry Pi have had some sort of business experience
58:32SPEAKERbefore, a more traditional business experience.
58:34SPEAKERSo I think for all of us, it's kind of been a wonderful toy,
58:39SPEAKERI guess, just this experience that no one's built
58:43SPEAKERan organization like this before.
58:45SPEAKERAnd it's interesting to see how far we can take it.
58:47SPEAKERI mean, we've sold 50 million computers.
58:48SPEAKERWe fab our own chips.
58:50SPEAKERIn fact, again, nobody fabs their own chips.
58:51SPEAKERWe pay people to fab our own chips.
58:52SPEAKERWe design our own chips.
58:53SPEAKER- You design your own chips.
58:54SPEAKER- And then TSMC fab them for us.
58:57SPEAKER- Just like Apple.
58:58SPEAKER- Yeah, just like everybody else.
59:00SPEAKERAnd it's kind of interesting to see the breadth
59:05SPEAKERof what you can do within this interestingly constrained
59:10SPEAKERenvironment.
59:11SPEAKER- Yeah, I think you're probably uniquely positioned
59:13SPEAKERto know more about what people are using Pi's for
59:16SPEAKERthan pretty much most anybody we would talk to.
59:19SPEAKERAnd I'm curious if there are any things
59:21SPEAKERthat you were really excited about.
59:22SPEAKERI think you mentioned high altitude ballooning a minute ago.
59:24SPEAKER- The hab stuff's great.
59:25SPEAKERI love the space stuff in general.
59:27SPEAKERSo I love our own Astro Pi program
59:29SPEAKERwhere we've put Pi's on the station
59:31SPEAKERfor kids to run the code on.
59:33SPEAKERA lot of European children in Europe have done that.
59:36SPEAKERIt's a European space agency collaboration.
59:39SPEAKERSo that's a fun one.
59:40SPEAKERI love that actually,
59:42SPEAKERalthough there are two Astro Pi's on the space station,
59:46SPEAKERthere's a bunch more, there are on it at any given day,
59:48SPEAKERa huge number more Raspberry Pi's on the ISS
59:50SPEAKERbecause they get used as payload controllers
59:52SPEAKERbecause they're very robust.
59:54SPEAKERAnd so often actually we have no idea
59:56SPEAKERhow many Raspberry Pi's there are on the ISS.
59:57SPEAKERWe only tend to find out about these uses
59:59SPEAKERwhen they get down mast, so that's kind of fun.
1:00:04SPEAKERWhat else?
1:00:05SPEAKERI just like all the people who've built businesses around it.
1:00:08SPEAKERWhether they're resellers
1:00:09SPEAKERor people who make DAC accessories
1:00:13SPEAKERor people who've built it into some product
1:00:17SPEAKERthat they've dreamt up.
1:00:19SPEAKERJust this idea of enabling entrepreneurship,
1:00:22SPEAKERenabling distributed, I love this.
1:00:25SPEAKERI think entrepreneurship, capitalism, business,
1:00:27SPEAKERthey're wonderfully powerful tools
1:00:29SPEAKERfor good change in the world.
1:00:31SPEAKERAnd it's wonderful to have had the opportunity
1:00:33SPEAKERto kind of unlock that kind of innovation,
1:00:35SPEAKERwhich it was getting hard to do, right?
1:00:37SPEAKERYou know, was able to walk down to an electronics store
1:00:41SPEAKERand get 6502 out of a bin for about the same price
1:00:46SPEAKERthat the biggest company in the world
1:00:47SPEAKERwould have paid for a 6502, right?
1:00:49SPEAKERIt was a very level playing field in the 1970s and 1980s.
1:00:52SPEAKERAnd the playing field is not level at all anymore
1:00:54SPEAKERin terms of access to advanced technology.
1:00:56SPEAKERIt's very hard, it's not that advanced technology is expensive
1:00:59SPEAKERfor some of these companies.
1:01:00SPEAKERIt's just unavailable.
1:01:01SPEAKERAnd so taking really advanced technology
1:01:03SPEAKERand making it available at a very flat price.
1:01:06SPEAKERYou know, one thing that characterizes Raspberry Pi
1:01:08SPEAKERis not just the thin margins,
1:01:09SPEAKERbut the very flat pricing structure.
1:01:11SPEAKERSo people come to us and say, "Oh, so Raspberry Pi is $35.
1:01:14SPEAKERHow much will I have to pay for 1,000 Raspberry Pis?"
1:01:18SPEAKERAnd the answer is $35,000.
1:01:21SPEAKERWe know you'll eke out 1,000, it might be $34,000,
1:01:25SPEAKERbut it's not $25,000 or $15,000.
1:01:28SPEAKERAnd that's something that's a commitment
1:01:31SPEAKERwe've stayed very true to all the way through.
1:01:33SPEAKERAnd I think it's obviously important
1:01:35SPEAKERfrom the educational side,
1:01:37SPEAKER'cause it means that the people
1:01:39SPEAKERwho are at the bottom end of the volume curve are kids.
1:01:42SPEAKERSo you don't want to be rinsing the kids.
1:01:44SPEAKERBut it's also the thing that's powered
1:01:46SPEAKERthis kind of community,
1:01:47SPEAKERthis ecosystem of innovation around the platform.
1:01:50SPEAKER- Well, and just so people understand,
1:01:52SPEAKERI mean, when you look at what Intel or Apple
1:01:55SPEAKERor Nvidia spend developing a new chip,
1:01:58SPEAKERthey're looking at billions of dollars, in some case,
1:02:00SPEAKERover a really long period of time.
1:02:02SPEAKERAnd I'm curious what your,
1:02:04SPEAKERlike when you're developing a new Pi model,
1:02:05SPEAKERpart of it is that you license arms,
1:02:07SPEAKERso you get a lot of stuff with that licensing cost.
1:02:09SPEAKERBut I'm curious what you spend developing a new Pi model
1:02:12SPEAKERand what that-
1:02:13SPEAKER- Well, remember most big Raspberry Pi products
1:02:16SPEAKERuse merchant silicon,
1:02:17SPEAKERwell, use effectively merchant silicon.
1:02:19SPEAKERSo we don't develop the chips
1:02:20SPEAKERthat are in the big Raspberry Pis, right?
1:02:22SPEAKERThat's protocol.
1:02:23SPEAKERSo we're not involved in that.
1:02:25SPEAKERWhen we do the,
1:02:26SPEAKERbut just doing the, our bit of that work is,
1:02:30SPEAKERthere's many millions of dollars now.
1:02:31SPEAKERSo Raspberry Pi 4 was,
1:02:34SPEAKERI mean, I could go and look it up,
1:02:36SPEAKERbut Raspberry Pi 4 was probably
1:02:37SPEAKERa two or $3 million program.
1:02:38SPEAKERRaspberry Pi Pico was probably
1:02:41SPEAKERa seven or $8 million program
1:02:42SPEAKERbecause that did have a chip development in it, right?
1:02:45SPEAKERSo that has the RP 2040 microcontroller wrapped up in it.
1:02:48SPEAKERSo it's probably RP 2040 is about a $5 million program
1:02:50SPEAKERjust on its own.
1:02:51SPEAKERAnd then you have the board development
1:02:53SPEAKERand the software development on top of that.
1:02:55SPEAKERSo it's only getting more expensive.
1:02:56SPEAKERI would imagine, you know, we are,
1:02:58SPEAKERwe've kind of taken a pause on new big Raspberry Pis
1:03:00SPEAKERbecause of the, well,
1:03:02SPEAKERwe focus on just making the existing Raspberry Pis available,
1:03:04SPEAKERbut I would imagine that any subsequent Raspberry Pi
1:03:06SPEAKERwill be a, would be a more than $10 million effort,
1:03:10SPEAKEReven without us, you know, even, you know,
1:03:13SPEAKERwithout a chip development in.
1:03:15SPEAKER- I guess what I'm getting at though,
1:03:16SPEAKERis that the benefit of having the Raspberry Pi
1:03:18SPEAKERas this, as this unit of compute and SOC
1:03:21SPEAKERthat you can jam into your projects,
1:03:22SPEAKERis that if you want to build a project
1:03:25SPEAKERthat needs a general purpose computer,
1:03:27SPEAKERyou don't need to do that work and effort.
1:03:29SPEAKERYou can just say, okay,
1:03:30SPEAKERwe have this thing that has these ins and outs
1:03:32SPEAKERand jam it into the thing that you want to build.
1:03:34SPEAKERAnd it's a lot more accessible to the smaller market.
1:03:37SPEAKER- Yeah, we have this,
1:03:38SPEAKERwe have this idea of reluctant hardware companies
1:03:42SPEAKERthat actually the world was full of reluctant hardware
1:03:44SPEAKERcompanies when we started as a digital signage companies,
1:03:47SPEAKERthin client companies,
1:03:49SPEAKERpeople building widgets to go into factories to log data.
1:03:52SPEAKERAnd all these people who were really software companies,
1:03:55SPEAKERthey all wanted to be software companies.
1:03:56SPEAKERWhat they're differentiating value
1:03:58SPEAKERwas in whatever software stack it was that they provided.
1:04:01SPEAKERBut in order to get that software stack,
1:04:02SPEAKERyou can't sell people software.
1:04:04SPEAKERThese are something to run on.
1:04:05SPEAKERAnd the PC was too expensive
1:04:07SPEAKERand there weren't readily available,
1:04:09SPEAKERembedded PCs at a respectable price point.
1:04:12SPEAKERAnd therefore people were building a little thin client,
1:04:14SPEAKERgoing to do all the work of going out and sourcing,
1:04:17SPEAKERcausing a look at a memory and bringing up a board,
1:04:19SPEAKERwriting a board support and getting an OS port
1:04:22SPEAKERjust so they get to the point
1:04:23SPEAKERwhere they could then run their application,
1:04:24SPEAKERwhich was there.
1:04:25SPEAKERAnd that's just so much wasted effort.
1:04:27SPEAKERAnd that's really the, yeah.
1:04:29SPEAKERWhat I'm proudest of,
1:04:30SPEAKERI think with Raspberry Pi's contribution there
1:04:33SPEAKERis that we freed a whole heap of companies.
1:04:35SPEAKERAnd so I talked about the boat tail.
1:04:37SPEAKERNow, actually, you'll see a thin client on a desktop
1:04:40SPEAKERand you look at it and it's got a boat tail on the end
1:04:42SPEAKER'cause it's actually a Raspberry Pi in a fancy box
1:04:44SPEAKERrunning somebody's differentiating software application.
1:04:46SPEAKERAnd that suit the kind of contribution to the public good,
1:04:51SPEAKERnot having thousands of talented engineers wasted
1:04:55SPEAKERdoing basically undifferentiated grunt hardware design
1:04:59SPEAKERis really, it's a good contribution.
1:05:01SPEAKERAnd I think ranks actually probably alongside
1:05:04SPEAKERin terms of what those engineers can do
1:05:06SPEAKERnow they're freed up,
1:05:08SPEAKERprobably ranks alongside the educational impact
1:05:12SPEAKERas in terms of good done by Raspberry Pi.
1:05:15SPEAKER- So I had a big last grand question
1:05:17SPEAKERand you've kind of actually alluded to an answer already,
1:05:19SPEAKERbut I wanted to ask what emotion do you feel
1:05:21SPEAKERwhen you hear the words Raspberry Pi 5?
1:05:25SPEAKER- Terror.
1:05:27SPEAKERSadness, sadness in that the sort of progress
1:05:32SPEAKERtowards Raspberry Pi 5 has been derailed
1:05:34SPEAKERby all of the challenges,
1:05:37SPEAKERpandemic and post-pandemic that have confronted us.
1:05:41SPEAKERI mean, obviously one day there will be a Raspberry Pi 5.
1:05:43SPEAKERWe've told people it's not this year and it isn't this year.
1:05:46SPEAKERI think once we get back into our traditional
1:05:51SPEAKERsuper robust stock position for Raspberry Pi 4,
1:05:53SPEAKERwe'll be in a position to think about it,
1:05:54SPEAKERbut it is some distance away.
1:05:56SPEAKERThese, as I say, they're not cheap programs
1:05:58SPEAKERand the fact they're not cheap
1:06:00SPEAKERimplies that they're not quick
1:06:01SPEAKER'cause it takes time to spend that much money,
1:06:03SPEAKERbut we'll get there.
1:06:04SPEAKERAnd we're lucky that this happened to us
1:06:06SPEAKERwhen Raspberry Pi 4 was the flagship platform
1:06:10SPEAKERnot Raspberry Pi 3, 'cause of course Raspberry Pi 4
1:06:11SPEAKERis enormously more performant
1:06:13SPEAKERand has a lot more headroom in terms of how much
1:06:17SPEAKERmore performance we can eke out of it
1:06:19SPEAKERby clever software optimization.
1:06:21SPEAKER- That was kind of my follow-up question.
1:06:22SPEAKERThe 4 to me is kind of the first,
1:06:25SPEAKERI shouldn't say real Pi, but like real computer.
1:06:27SPEAKERYou know what I mean?
1:06:28SPEAKERIt's a quad core, like good IPC,
1:06:29SPEAKERlike it finally has full speed ethernet and all that stuff.
1:06:31SPEAKERI mean, in a world with no pandemic,
1:06:33SPEAKERdo you think the Pi 4 would have lasted longer
1:06:35SPEAKERthan its predecessors even without all the external factors?
1:06:39SPEAKER- Yes, yes, absolutely.
1:06:40SPEAKERIt's the first one that delivered
1:06:41SPEAKERon our dream really of a no compromises.
1:06:45SPEAKERYou can sit down and we had people, lots of kids did.
1:06:47SPEAKERI mean, we had a lot of charitable work
1:06:48SPEAKERgetting these out to people in the pandemic.
1:06:51SPEAKERIt's the first one that you can sit down in front of
1:06:54SPEAKERand use as a no compromises entry-level PC.
1:06:59SPEAKERYou're not gonna run a big CAD package on it,
1:07:01SPEAKERbut you can run a web browser on it.
1:07:04SPEAKERAnd so it was good that that came along.
1:07:07SPEAKERIt's good that was available.
1:07:08SPEAKERAnd obviously Pi 4 and then Pi 400 during the pandemic.
1:07:11SPEAKERAnd it was one of the things that sort of sustained me
1:07:14SPEAKERwhen I was locked in my house
1:07:15SPEAKERwas all the pictures of kids getting their Pi 4
1:07:18SPEAKERand then Pi 400 kits.
1:07:21SPEAKERThese kids who, you know, we have so many,
1:07:22SPEAKERit was such a crime.
1:07:28SPEAKERSending all the kids in the country home to work from home
1:07:32SPEAKERand forgetting that huge numbers.
1:07:36SPEAKERI think it was 700,000 households in the UK
1:07:38SPEAKERhad kids, but no general purpose computer was just...
1:07:42SPEAKERI mean, it didn't create the inequality.
1:07:45SPEAKERIt made us all, it reminded us all the inequality was there.
1:07:48SPEAKERAnd so to the extent that we were able to get units out
1:07:50SPEAKERto a few tens of thousands,
1:07:52SPEAKERwe have some great philanthropic support.
1:07:54SPEAKERWe're able to get a few tens of thousands of units out
1:07:56SPEAKERto kids mostly in the UK.
1:07:58SPEAKERJust the looks on their faces when they opened the box
1:08:00SPEAKERand it was a computer inside was amazing.
1:08:02SPEAKER- I think there's no better place
1:08:03SPEAKERto wrap it up than that, Eben.
1:08:06SPEAKERIf people want to find out more,
1:08:07SPEAKERwhat are the best places to find you?
1:08:10SPEAKER- So raspberrypi.com for Raspberry Pi limited stuff.
1:08:15SPEAKERSo engineering and products, raspberrypi.org
1:08:20SPEAKERfor the Raspberry Pi Foundation
1:08:21SPEAKERand all that good educational work.
1:08:27SPEAKER- That'll do it for us this week.
1:08:28SPEAKERThanks everybody for listening.
1:08:30SPEAKERThanks to Eben for coming by and chatting with us
1:08:32SPEAKERand being extremely generous with his time.
1:08:34SPEAKERI know he's always really busy.
1:08:36SPEAKER- Yes, it was a joy, honestly.
1:08:37SPEAKERI mean, I thoroughly enjoyed every interview
1:08:39SPEAKERwe've done for the show,
1:08:40SPEAKERbut the Raspberry Pi phenomenon has gotten so big
1:08:44SPEAKERthat it was a real honor to get to talk to him
1:08:46SPEAKERand see that he's still an engineer at heart.
1:08:48SPEAKERI mean, you never take that out of,
1:08:50SPEAKERno matter how busy he is with other business
1:08:53SPEAKERand foundation stuff,
1:08:54SPEAKERbeing able to get into the weeds on a technical level
1:08:56SPEAKERwith him about the stuff that we talked about
1:08:58SPEAKERwas just super gratifying.
1:09:00SPEAKER- Well, it's interesting to see
1:09:01SPEAKERthat he still seemed a little surprised
1:09:04SPEAKERthat it's become this kind of juggernaut, right?
1:09:07SPEAKERWhich is fun.
1:09:09SPEAKERI often tell people that my favorite part about this show
1:09:11SPEAKERis talking to people in different stages
1:09:13SPEAKERof the open source software lifecycle.
1:09:15SPEAKERAnd in this case, this is a hardware project
1:09:17SPEAKERthat is barely 10 years old at this point,
1:09:20SPEAKERa little bit more than 10 years old,
1:09:21SPEAKERand is contributing greatly
1:09:23SPEAKERto a massive number of sectors around the world,
1:09:28SPEAKERfar beyond what I think they ever anticipated.
1:09:30SPEAKERSo you can find out more about Raspberry Pi
1:09:32SPEAKERat raspberrypi.org for the foundation
1:09:34SPEAKERor raspberrypi.com for the board.
1:09:38SPEAKERAnd that's Pi, like the number 3.14,
1:09:41SPEAKERnot the thing that you put in your mouth.
1:09:44SPEAKER- It's Pi, not P-I-E.
1:09:45SPEAKER- The other thing I will say is the thing we talked about
1:09:46SPEAKERwhere Raspberry Pi 3s are now becoming more available,
1:09:51SPEAKERwe are starting to see that happen in the real world again.
1:09:55SPEAKERSo that's, if you want to get a Pi,
1:09:57SPEAKERyou don't necessarily, F4s are great.
1:10:00SPEAKERIf you can't get a 4 at the real price,
1:10:02SPEAKERa 3 is fine for most of the projects
1:10:04SPEAKERthat we've talked about on this show
1:10:05SPEAKERand on the tech pod in the past.
1:10:07SPEAKER- A 3 is no slouch, you know,
1:10:08SPEAKEReven if you've got an old 1 or a 2 lying around,
1:10:10SPEAKERthose have their uses as well.
1:10:13SPEAKER- Yeah, so this week's episode, as always,
1:10:15SPEAKERis brought to you by Google Open Source.
1:10:16SPEAKERThey bring all the power of open source to Google
1:10:18SPEAKERand all the power of Google to open source.
1:10:20SPEAKERYou can find out more about them
1:10:22SPEAKERat opensource.google, thanks Google.
1:10:24SPEAKERAs always, this week's episode was produced by Matt Purdy
1:10:27SPEAKERand edited by Sabrina Hill, thanks Matt and Sabrina.
1:10:29SPEAKERAnd we will be back in a couple of weeks
1:10:31SPEAKERwith another episode.
1:10:32SPEAKERI think we're gonna talk about KiCad on the next one.
1:10:34SPEAKER- Mm-hmm, we are.
1:10:36SPEAKER- See you all then.