WEBVTT

00:00:00.000 --> 00:00:01.360
<v Chris Gammell>This is The Amp Hour Podcast.

00:00:02.980 --> 00:00:05.780
<v Chris Gammell>Released October 22nd, 2023.

00:00:08.080 --> 00:00:09.540
<v Chris Gammell>Episode 648.

00:00:11.280 --> 00:00:14.120
<v Chris Gammell>The RP1 and beyond with the Raspberry Pi Harper team.

00:00:30.000 --> 00:00:31.720
<v Chris Gammell>Welcome to the Amp Hour.

00:00:31.940 --> 00:00:33.440
<v Chris Gammell>I'm Chris Gammell of Contextual Electronics.

00:00:33.980 --> 00:00:36.860
<v James Adams>Hi, I'm James Adams, CTO, Hardware at Raspberry Pi.

00:00:37.120 --> 00:00:41.800
<v Liam Fraser>Hi, I'm Liam Fraser, a mix of software and hardware engineering at Raspberry Pi.

00:00:42.340 --> 00:00:42.740
<v Liam Fraser>Awesome.

00:00:42.860 --> 00:00:45.460
<v Chris Gammell>Welcome back and congrats on the new silicon.

00:00:45.600 --> 00:00:46.580
<v Chris Gammell>That's what we're going to be talking about today.

00:00:46.580 --> 00:00:48.480
<v Chris Gammell>The RP1 silicon was just released.

00:00:49.080 --> 00:00:51.660
<v Chris Gammell>And we're going to do a little recap on last time you guys are here.

00:00:51.740 --> 00:00:53.580
<v Chris Gammell>You were here back in 2021.

00:00:54.220 --> 00:00:58.200
<v Chris Gammell>And we'd love to get an update and hear about all the new things that are happening.

00:00:58.800 --> 00:00:59.100
<v James Adams>That's right.

00:00:59.100 --> 00:01:00.300
<v James Adams>Well, thanks for having us back, Chris.

00:01:00.440 --> 00:01:01.220
<v James Adams>Glad to be here.

00:01:01.660 --> 00:01:05.800
<v James Adams>Yeah, let's have a wander through what Raspberry Pi has been doing, I guess, on chip and product

00:01:05.800 --> 00:01:06.940
<v James Adams>side, if you want to talk about that.

00:01:07.280 --> 00:01:08.520
<v James Adams>So where would you like to start?

00:01:09.040 --> 00:01:14.700
<v Chris Gammell>Well, I think one of the things that has come out as part of this is just that the RP2040

00:01:14.700 --> 00:01:16.320
<v Chris Gammell>wasn't the initial target, right?

00:01:16.340 --> 00:01:18.160
<v Chris Gammell>You guys were driving towards this recently.

00:01:18.960 --> 00:01:21.500
<v Chris Gammell>The RP1 is one for a reason, right?

00:01:21.520 --> 00:01:23.280
<v Chris Gammell>It was the first thing you guys started working on, right?

00:01:23.600 --> 00:01:23.880
<v James Adams>Yeah.

00:01:24.260 --> 00:01:27.640
<v James Adams>So I guess the naming sounds confusing.

00:01:27.640 --> 00:01:29.740
<v James Adams>But actually, yeah, RP1 is the first.

00:01:29.880 --> 00:01:33.260
<v James Adams>It's called RP1 because it's the first Raspberry Pi silicon that we started.

00:01:33.780 --> 00:01:36.120
<v James Adams>We've got RP2, which is RP2040.

00:01:36.340 --> 00:01:39.960
<v James Adams>And RP3, which is our chip that goes on the 02.

00:01:40.840 --> 00:01:45.120
<v James Adams>So yeah, that's just the order that we started the chip programs.

00:01:45.120 --> 00:01:49.100
<v James Adams>I guess the genesis of RP1 was back in 2015.

00:01:50.000 --> 00:01:57.680
<v James Adams>And we knew that we have this natural architecture of a main processor on the Raspberry Pi and then

00:01:57.680 --> 00:01:59.040
<v James Adams>some kind of IO controller.

00:01:59.520 --> 00:02:03.960
<v James Adams>So back in the early days, that's the LAN 9512 on the very early boards.

00:02:03.960 --> 00:02:12.540
<v James Adams>And then four, which is basically USB to USB, sorry, a USB 2 to a USB 2 hub and Ethernet, 10100 originally.

00:02:13.000 --> 00:02:17.140
<v James Adams>So we kind of wanted to improve on that really.

00:02:17.620 --> 00:02:20.980
<v James Adams>And most people from Raspberry Pi are from a chip design background.

00:02:20.980 --> 00:02:27.340
<v James Adams>And the thought was that we would like to improve on that offering.

00:02:27.440 --> 00:02:30.560
<v James Adams>So back when we started this, we had USB 2, right?

00:02:30.620 --> 00:02:31.540
<v James Adams>We had LAN 9512.

00:02:31.700 --> 00:02:31.760
<v James Adams>Yeah.

00:02:32.140 --> 00:02:32.300
<v James Adams>Yeah.

00:02:32.960 --> 00:02:38.360
<v James Adams>Well, we also, on the Brawcom chips, we had these MIPI channels spare.

00:02:38.360 --> 00:02:40.660
<v James Adams>So we had CSI and DSI.

00:02:40.820 --> 00:02:47.040
<v James Adams>So these could do multi-gigabits of bandwidth, primarily for camera and display.

00:02:47.200 --> 00:02:52.180
<v James Adams>But the initial thought was, well, could we repurpose these channels to actually transport

00:02:52.180 --> 00:02:55.800
<v James Adams>data and do an IO chip by plugging it to those?

00:02:56.420 --> 00:03:02.520
<v James Adams>So this is before we kind of came up with the revelation that, well, actually, we should

00:03:02.520 --> 00:03:04.320
<v James Adams>do a standards-based thing and adopt PTI Express.

00:03:04.320 --> 00:03:08.780
<v James Adams>The original idea was to do something funky with these spare interfaces.

00:03:09.400 --> 00:03:13.800
<v James Adams>And so we started the team in a kind of conservative way, as we do at Raspberry Pi Engineering.

00:03:13.980 --> 00:03:18.980
<v James Adams>So we had three people that we knew from our chip design days who had gone off to do a startup

00:03:18.980 --> 00:03:21.000
<v James Adams>and the startup was running out of money.

00:03:21.840 --> 00:03:27.620
<v James Adams>And so we employed them as contractors to kind of scope out how could we do this?

00:03:27.700 --> 00:03:29.200
<v James Adams>Can we start looking at costing it?

00:03:29.240 --> 00:03:30.620
<v James Adams>Can we start looking at where we get the IP?

00:03:30.620 --> 00:03:34.800
<v James Adams>And so really, we started a chip design team really absolutely from scratch.

00:03:34.880 --> 00:03:35.400
<v James Adams>So we had no one.

00:03:35.460 --> 00:03:36.920
<v James Adams>Then we had three people as contractors.

00:03:37.420 --> 00:03:41.980
<v James Adams>And gradually, as we built the specification, we started to find the IP.

00:03:42.200 --> 00:03:45.740
<v James Adams>We started to build the more detailed specification on how are we really going to do this?

00:03:45.800 --> 00:03:51.160
<v James Adams>We did some software work on, well, is it feasible to reuse these channels as data channels?

00:03:51.840 --> 00:03:52.780
<v James Adams>And yeah, it kind of grew from there.

00:03:52.800 --> 00:03:54.700
<v James Adams>And eventually, we were quite committed.

00:03:54.700 --> 00:04:01.400
<v James Adams>So we pulled those guys on as full-time employees and started employing more ASIC guys.

00:04:02.000 --> 00:04:04.180
<v James Adams>And so now we've got a reasonable-sized team.

00:04:04.260 --> 00:04:07.920
<v James Adams>I think it's still a very small team by ASIC design standards.

00:04:08.580 --> 00:04:08.660
<v Chris Gammell>Right.

00:04:09.060 --> 00:04:09.160
<v Chris Gammell>Yeah.

00:04:09.300 --> 00:04:11.060
<v Chris Gammell>Intel laughs at your...

00:04:11.060 --> 00:04:11.660
<v Chris Gammell>Absolutely.

00:04:11.900 --> 00:04:13.440
<v Chris Gammell>We have more interns than you guys have.

00:04:13.440 --> 00:04:14.620
<v Chris Gammell>Yeah, pretty much.

00:04:14.700 --> 00:04:14.920
<v Chris Gammell>Right.

00:04:14.920 --> 00:04:21.680
<v James Adams>But we do, you know, we employ people who are a lot of people who've done this for a long time, right?

00:04:21.680 --> 00:04:23.180
<v James Adams>So they really know what they're doing.

00:04:23.340 --> 00:04:23.400
<v James Adams>Yeah.

00:04:23.660 --> 00:04:28.700
<v James Adams>And we're lucky to, because we were in the space before, we knew the good people to go and pull on.

00:04:28.900 --> 00:04:32.100
<v James Adams>And it's taken a little bit of time, but we've managed to recruit a superstar team.

00:04:32.480 --> 00:04:34.540
<v James Adams>And, you know, it's paid off, right?

00:04:35.040 --> 00:04:38.100
<v James Adams>We do manage to do quite a lot with a small team.

00:04:38.100 --> 00:04:43.860
<v James Adams>And I guess it kind of speaks to why RP1 took so long, actually, is that it's a complicated chip.

00:04:44.080 --> 00:04:48.420
<v James Adams>We kind of went through this cycle of deciding on one sort of architecture.

00:04:48.780 --> 00:04:53.140
<v James Adams>And then we did actually have a test chip with these MIPI channels in, and it did actually work.

00:04:53.420 --> 00:04:54.660
<v James Adams>But it kind of missed the boat.

00:04:54.880 --> 00:04:55.420
<v James Adams>It was late.

00:04:56.160 --> 00:05:00.260
<v James Adams>And that's largely just the kind of, you know, the effort to build it.

00:05:00.440 --> 00:05:01.600
<v James Adams>It just took longer than we expected.

00:05:01.960 --> 00:05:04.460
<v James Adams>You know, you're trying to bring up a whole team from scratch.

00:05:04.460 --> 00:05:13.380
<v James Adams>And once we'd, I think we basically taped up, as we taped out that first test chip, we really, we'd already decided, okay, we're going to switch this to PCI Express.

00:05:13.380 --> 00:05:15.520
<v James Adams>Because it has all these advantages.

00:05:15.900 --> 00:05:19.320
<v James Adams>You can plug it into standards-based system.

00:05:20.120 --> 00:05:28.360
<v James Adams>We can plug it into, even before we had, because at this point, we knew the new Broadcom processes upcoming would have PCI Express.

00:05:29.100 --> 00:05:30.200
<v James Adams>And so we could see the future.

00:05:30.200 --> 00:05:38.720
<v James Adams>And we knew we could, even if we didn't have those yet, we could plug it into an x86 PC and do early development work, which is what we did.

00:05:39.500 --> 00:05:43.500
<v Chris Gammell>So does that mean on the, on the, so you mentioned like the MIPI and CSI DSI.

00:05:43.820 --> 00:05:49.700
<v Chris Gammell>You said those are kind of like extra channels that were on that, whatever the generation of the Raspberry Pi was.

00:05:49.780 --> 00:05:50.080
<v Chris Gammell>Is that right?

00:05:50.080 --> 00:05:53.060
<v James Adams>Yeah, so this is, this is the first, very first.

00:05:53.560 --> 00:05:56.720
<v James Adams>2035 actually had, has two camera two display.

00:05:57.200 --> 00:05:59.480
<v James Adams>And we only ever exposed one of each.

00:06:00.260 --> 00:06:01.260
<v James Adams>Oh, interesting.

00:06:01.380 --> 00:06:02.720
<v James Adams>On the original boards.

00:06:03.280 --> 00:06:09.480
<v James Adams>Now, the compute module subsequently exposed all of them, but we kind of had two spare rights.

00:06:09.720 --> 00:06:12.280
<v James Adams>So that, that was, that was the idea.

00:06:12.280 --> 00:06:12.600
<v James Adams>Okay.

00:06:12.840 --> 00:06:13.200
<v Chris Gammell>Okay.

00:06:13.260 --> 00:06:22.500
<v Chris Gammell>So those were, those were meant as, I thought, I, I, I, I, when I heard you saying, I thought you meant that they were kind of just spare general, general purpose type things.

00:06:22.500 --> 00:06:26.780
<v Chris Gammell>And then they were kind of repurposed as camera and display, but it was, it was a camera and display.

00:06:26.920 --> 00:06:27.080
<v Chris Gammell>Yeah.

00:06:27.140 --> 00:06:28.860
<v James Adams>They're quite specifically camera and display.

00:06:28.860 --> 00:06:37.260
<v James Adams>So you had to jump through hoops and decide, well, how on earth do we, we fake this thing to look like a camera and display, but actually just pump general purpose data through it.

00:06:37.260 --> 00:06:37.580
<v James Adams>Yeah.

00:06:37.680 --> 00:06:41.620
<v James Adams>So it was a bit, it was a bit hairy and a bit crazy and it did work, but yeah.

00:06:41.620 --> 00:06:50.780
<v Chris Gammell>I mean, that sounds like a lot of, I mean, like I've repurposed, you know, general, you know, like other GPIO or weird, weird peripherals, you know, use what you got.

00:06:50.780 --> 00:06:52.180
<v Chris Gammell>Sometimes you're just kind of pin limited.

00:06:52.500 --> 00:06:56.540
<v Chris Gammell>And in the case of Silicon, there's really no, there's no rework in that sort of thing.

00:06:56.540 --> 00:06:58.060
<v Chris Gammell>So that, that makes a lot of sense.

00:06:58.980 --> 00:07:01.880
<v Chris Gammell>Before we go into the RP one, I'd love to get a update.

00:07:01.880 --> 00:07:05.560
<v Chris Gammell>So like, like we mentioned at the beginning of this, we talked about two and a half years ago.

00:07:05.560 --> 00:07:07.120
<v Chris Gammell>It was mid pandemic.

00:07:07.380 --> 00:07:11.180
<v Chris Gammell>The RP2040 was the only thing that was really, I mean, it was one of the only chips that was available out there.

00:07:11.680 --> 00:07:13.220
<v Chris Gammell>I think it's seen a lot of success.

00:07:14.000 --> 00:07:16.280
<v Chris Gammell>What do you guys, what's, what's the status of it?

00:07:16.300 --> 00:07:17.280
<v Chris Gammell>Where, where do you see it?

00:07:17.440 --> 00:07:19.440
<v Chris Gammell>Where do you see it out in the world and how's it been going?

00:07:20.120 --> 00:07:21.140
<v James Adams>Who's going to answer that one?

00:07:21.160 --> 00:07:21.940
<v Liam Fraser>Liam, did you want to?

00:07:22.240 --> 00:07:22.960
<v Liam Fraser>Yeah, sure.

00:07:22.960 --> 00:07:36.360
<v Liam Fraser>I mean, I know we've got a lot of industrial customers using RP2040, but I think in particular, the thing that's really done RP2040 well is the, you know, the SDK.

00:07:36.360 --> 00:07:38.080
<v Liam Fraser>We spent so long on that.

00:07:38.520 --> 00:07:40.800
<v Liam Fraser>The MicroPython support out the box.

00:07:40.800 --> 00:07:49.120
<v Liam Fraser>You know, I, and then also Pico W, like, you know, I wanted to add wifi to some fan thing that I had lying around the other day.

00:07:49.200 --> 00:07:52.360
<v Liam Fraser>It was some noisy, uh, plinth heater thing.

00:07:52.560 --> 00:07:57.460
<v Liam Fraser>And I took that out, put some Noctua fans in and hooked up Pico W to it.

00:07:57.540 --> 00:08:04.700
<v Liam Fraser>And it literally is, you know, 10 minutes of MicroPython to, I can now talk to this thing over wifi and, uh, change the speed of the fans.

00:08:04.700 --> 00:08:17.440
<v Liam Fraser>So, uh, yeah, and actually it would be nice to touch on, uh, so the thing that RP2040 allowed us to do during the development to sort of bring RP1 back in a little bit.

00:08:17.600 --> 00:08:32.820
<v Liam Fraser>The first version of RP1, uh, that we started when we had the small chip team, all the RTL was written by hand and RP2040 is actually the chip that we use to develop our sort of scripting our tools on.

00:08:32.820 --> 00:08:40.560
<v Liam Fraser>Because I don't, I don't know when we started RP2040, but you know, it would be at least sort of three years or four years before, before release.

00:08:40.840 --> 00:08:53.380
<v Liam Fraser>And, and the idea was, you know, this is a reasonably small project and it will allow us to sort of prove out, prove out this kind of stuff and then sort of port it back in a, in a way to, to RP1.

00:08:53.380 --> 00:09:04.900
<v Liam Fraser>So sort of between the first version of RP1 and doing RP2040, then the sort of second version of RP1 was using all the scripting stuff that we developed on, on RP2040.

00:09:05.540 --> 00:09:12.280
<v Liam Fraser>And a lot of the RP2040 stuff was, you know, reusing RP1 components, but in a bit more scriptable way.

00:09:12.460 --> 00:09:15.480
<v Liam Fraser>So a lot of the groundwork had been done.

00:09:16.120 --> 00:09:21.800
<v Chris Gammell>And the scripting in that case, the scripting refers to like boot code and that sort of thing, or where, where is the scripting involved?

00:09:21.800 --> 00:09:25.060
<v Liam Fraser>So it's any sort of common components of the chip.

00:09:25.520 --> 00:09:33.420
<v Liam Fraser>So we've got a small team, RP1, for example, certainly the later versions, you know, there's probably like 20 different clocks on there.

00:09:33.540 --> 00:09:35.860
<v Liam Fraser>There's, you know, 50 odd GPIOs.

00:09:36.380 --> 00:09:39.380
<v Liam Fraser>The bus fabric's got like 90 different connections on it.

00:09:39.380 --> 00:09:49.300
<v Liam Fraser>And all of those things, we basically just stamp out several instances of the same thing, all configured in sort of, you know, in a scripting language.

00:09:49.620 --> 00:09:56.540
<v Liam Fraser>And it really, and this stuff also spits out, you know, your software headers, your documentation, all this kind of stuff.

00:09:57.220 --> 00:10:02.800
<v Chris Gammell>All the things that I assume is just magically working, but it's actually under the hood, it's very non-trivial, I'm guessing.

00:10:02.800 --> 00:10:06.180
<v Liam Fraser>Yeah, yeah, well, we're still developing it.

00:10:06.660 --> 00:10:12.380
<v Liam Fraser>And yeah, it's, you know, it's a sort of evolving thing as we decide we need new features.

00:10:12.900 --> 00:10:16.080
<v Liam Fraser>But yeah, that's how we're able to do quite a lot with a small team.

00:10:16.940 --> 00:10:25.480
<v James Adams>Yeah, I think it's driven by sort of us old hands, basically just being sick of all the typing required to stitch blocks of IP together in chips, right?

00:10:25.480 --> 00:10:25.600
<v James Adams>Yeah.

00:10:26.160 --> 00:10:37.280
<v James Adams>And every ASIC company has some form of scripting of doing this, but I don't know how many years ago it was, Liam, we sort of sat down and said, hey, we need to do some scripting because it's just such a pain to put all this stuff together.

00:10:37.280 --> 00:10:50.020
<v James Adams>And what Liam and Terry and the guys have produced is just like way above what I've ever seen in terms of its ability to, you can basically describe what the chip looks like in a high level kind of language.

00:10:50.280 --> 00:10:52.720
<v James Adams>And it literally goes and builds everything.

00:10:53.020 --> 00:10:53.660
<v James Adams>That's cool.

00:10:53.660 --> 00:11:07.020
<v James Adams>It reduces a ton of work and obviously reduces errors and bugs and it builds all the GPIO and the clocks and all these kind of things kind of out of sort of reusable basic components.

00:11:07.460 --> 00:11:12.480
<v Chris Gammell>It's starting to sound like FPJs, almost like halfway to FPJs.

00:11:12.900 --> 00:11:13.460
<v James Adams>Kind of.

00:11:13.560 --> 00:11:16.200
<v James Adams>I mean, there's still quite a lot of fettling of each chip, right?

00:11:16.340 --> 00:11:21.300
<v James Adams>I think probably, well, Liam, again, Liam is the expert because he's basically put this together.

00:11:21.300 --> 00:11:26.960
<v James Adams>So you can correct me if I'm wrong, but I think every time we do a chip, we do evolve it a bit.

00:11:27.120 --> 00:11:30.440
<v James Adams>And then that is part of the chip build process, right?

00:11:30.480 --> 00:11:35.600
<v James Adams>And then the next chip probably needs a little bit of customization, but it vastly reduces the amount of work.

00:11:35.660 --> 00:11:49.260
<v James Adams>And if you just want to, say, change a port on the bus or something and do something fairly trivial, it's five minutes work and a recompile rather than hacking a whole bunch of files to pull buses up and down the structure of the chip and all this kind of thing.

00:11:49.260 --> 00:11:58.360
<v James Adams>So it's like, we do try and, I think there's a general kind of approach that we have to engineering is where we, we, we're always slightly under-resourced.

00:11:58.360 --> 00:12:01.200
<v James Adams>So it kind of drives people to try and find ways to-

00:12:01.200 --> 00:12:00.960
<v Speaker ?>The hunger.

00:12:01.080 --> 00:12:01.600
<v Chris Gammell>It's the hunger.

00:12:01.820 --> 00:12:06.100
<v James Adams>Yeah, it's kind of like keep everyone hungry and it kind of drives people to find innovation.

00:12:06.100 --> 00:12:09.800
<v James Adams>So we like, we really like to find this way, ways of working smarter, not harder.

00:12:09.800 --> 00:12:16.140
<v James Adams>And I think this is one of the really nice ways that we've managed to do this on the chip design side.

00:12:16.960 --> 00:12:22.320
<v Chris Gammell>How does it get exposed then to something like MicroPython or, you know, other frameworks that are using this then?

00:12:22.680 --> 00:12:31.020
<v Chris Gammell>Is it like, does that call APIs that are then calling these scripts or like how much, how much is exposed to the, the software dummies like me?

00:12:31.020 --> 00:12:43.580
<v Liam Fraser>So the, the scripts aren't actually, what, what the scripts are doing is effectively writing out Verilog, which, which describes the, you know, which is the harder description language, which describes the chip.

00:12:43.700 --> 00:12:46.000
<v Liam Fraser>So by the time you're running software.

00:12:46.180 --> 00:12:46.420
<v Chris Gammell>Oh, okay.

00:12:46.420 --> 00:12:50.020
<v Chris Gammell>So this is, oh, this is all pre, pre Silicon being made.

00:12:50.140 --> 00:12:50.560
<v Chris Gammell>Oh, I see.

00:12:50.620 --> 00:12:50.920
<v Chris Gammell>Oh, wow.

00:12:50.960 --> 00:12:51.140
<v Chris Gammell>Okay.

00:12:51.600 --> 00:12:52.100
<v Chris Gammell>Oh, okay.

00:12:52.120 --> 00:12:54.820
<v Chris Gammell>I thought this was then configuration of the clock.

00:12:54.820 --> 00:13:04.500
<v Chris Gammell>So you were saying, so when you said like the 50 clocks and all these GPIOs and the fabric and stuff like that, this is to make a 2040 or example, like a 2041, right?

00:13:04.520 --> 00:13:09.980
<v Chris Gammell>If you were going to go from a 2040 to 2041, you would change the configuration file, rerun the scripts.

00:13:10.060 --> 00:13:11.820
<v Chris Gammell>And then you have like a new part.

00:13:11.900 --> 00:13:13.580
<v Chris Gammell>Is that, is that kind of the idea?

00:13:14.180 --> 00:13:14.460
<v Liam Fraser>Yeah.

00:13:14.820 --> 00:13:15.080
<v Liam Fraser>Yeah.

00:13:15.400 --> 00:13:15.620
<v Liam Fraser>Wow.

00:13:16.040 --> 00:13:21.040
<v Chris Gammell>And all that you got to do is just hand over some money to your, your fab and presto change it.

00:13:21.040 --> 00:13:21.600
<v Chris Gammell>What's the hard part?

00:13:21.680 --> 00:13:23.620
<v Chris Gammell>You know, like what, what's taking so long, right?

00:13:23.620 --> 00:13:24.060
<v Chris Gammell>Yeah.

00:13:24.280 --> 00:13:24.640
<v Liam Fraser>Yeah.

00:13:25.240 --> 00:13:28.180
<v Liam Fraser>Unfortunately, you've still got to test the output of what the script makes.

00:13:30.680 --> 00:13:35.800
<v James Adams>So yeah, this, this chip design is kind of like 20% design and then 80% test.

00:13:35.840 --> 00:13:39.440
<v James Adams>Actually, I think that's probably a lot of projects, right?

00:13:40.180 --> 00:13:44.100
<v Chris Gammell>Except at the beginning when you're planning, you're like, oh, test will be like, what?

00:13:44.200 --> 00:13:45.980
<v Chris Gammell>Like 30% overall.

00:13:46.200 --> 00:13:46.460
<v Chris Gammell>Right.

00:13:46.580 --> 00:13:48.540
<v Chris Gammell>And it's like, no, that's not how that works.

00:13:49.540 --> 00:13:49.900
<v James Adams>Yeah.

00:13:49.900 --> 00:13:54.000
<v James Adams>I mean, even with the tests, we have automated stuff to help with this as well.

00:13:54.520 --> 00:13:56.360
<v James Adams>So it's not all bad news.

00:13:56.360 --> 00:13:57.360
<v Chris Gammell>Hmm.

00:13:57.940 --> 00:14:03.320
<v Chris Gammell>So then, I mean, I don't dare to ask, uh, you know, what are the long-term plans, but

00:14:03.320 --> 00:14:04.920
<v Chris Gammell>what are the long-term plans?

00:14:05.020 --> 00:14:10.320
<v Chris Gammell>Like, I mean, so you have three Silicon pieces now that are, you know, released and public

00:14:10.320 --> 00:14:10.900
<v Chris Gammell>and stuff like that.

00:14:10.900 --> 00:14:16.740
<v Chris Gammell>But like, is there kind of like a overarching, like things that your team is trying to solve

00:14:16.740 --> 00:14:17.140
<v Chris Gammell>otherwise?

00:14:17.140 --> 00:14:24.160
<v Chris Gammell>I mean, because I don't see many blocks that are on Raspberry Pi-ified on an actual Raspberry

00:14:24.160 --> 00:14:24.580
<v Chris Gammell>Pi, right?

00:14:24.580 --> 00:14:30.100
<v Chris Gammell>You guys have captured a lot of, aside from the actual, the Broadcom chip, you know, you're

00:14:30.100 --> 00:14:32.280
<v Chris Gammell>doing a bunch of the other stuff that's on the board.

00:14:32.460 --> 00:14:34.020
<v Chris Gammell>So what's next?

00:14:35.280 --> 00:14:38.420
<v James Adams>Well, I mean, we have a chip team, so we're definitely doing more chip stuff.

00:14:38.420 --> 00:14:39.320
<v Chris Gammell>Mm-hmm.

00:14:39.720 --> 00:14:45.100
<v James Adams>I think it's been mentioned before that, you know, it's likely we'll have new generations

00:14:45.100 --> 00:14:47.220
<v James Adams>of microcontroller on the way at some point.

00:14:47.500 --> 00:14:50.340
<v James Adams>I don't think that's a particular secret, because, hey, we've done one.

00:14:50.540 --> 00:14:52.080
<v Chris Gammell>Because that's just how micro team works.

00:14:52.180 --> 00:14:52.280
<v Chris Gammell>Yeah.

00:14:52.320 --> 00:14:54.140
<v Chris Gammell>I mean, like, I want to see the new model here.

00:14:54.360 --> 00:14:57.320
<v James Adams>They'll crank out more chips, and I'm sure we'll do a spin of RP1.

00:14:57.940 --> 00:14:58.160
<v Chris Gammell>Yeah.

00:14:58.640 --> 00:14:59.820
<v James Adams>You know, it's still a small team.

00:14:59.960 --> 00:15:03.280
<v James Adams>So people have asked, you know, are you going to do your own main SoC?

00:15:03.820 --> 00:15:07.400
<v James Adams>I think we probably even touched, we even touched on this, I think, in the RP2040

00:15:07.400 --> 00:15:08.500
<v James Adams>discussion.

00:15:08.680 --> 00:15:12.060
<v James Adams>I've certainly, people have asked, and it's kind of like, that's a different order of

00:15:12.060 --> 00:15:12.740
<v James Adams>magnitude problem.

00:15:12.960 --> 00:15:13.480
<v James Adams>That, totally.

00:15:13.820 --> 00:15:13.980
<v Chris Gammell>Yeah.

00:15:14.120 --> 00:15:19.880
<v James Adams>You know, I think what we've done with RP1 is make it easier to make the big chip, like

00:15:19.880 --> 00:15:23.580
<v James Adams>at Brawcom, because they don't have to faff with adding all the analog Raspberry Pi stuff

00:15:23.580 --> 00:15:24.160
<v James Adams>to it, right?

00:15:24.560 --> 00:15:29.200
<v James Adams>And when I say analog, I mean the kind of like, the sort of medium speed stuff like USB,

00:15:29.760 --> 00:15:32.020
<v James Adams>GPIO, all this kind of stuff.

00:15:32.020 --> 00:15:36.320
<v James Adams>In general, they don't really want to, they kind of want to build chips that are a bit

00:15:36.320 --> 00:15:38.920
<v James Adams>more generic, and we want very specific Raspberry Pi things.

00:15:39.060 --> 00:15:42.800
<v James Adams>So it makes a lot of sense for us to put our specific Raspberry Pi things in a separate

00:15:42.800 --> 00:15:45.000
<v James Adams>bit of silicon, which we can control.

00:15:45.720 --> 00:15:48.260
<v James Adams>And then, you know, it just makes life a lot easier.

00:15:48.800 --> 00:15:49.960
<v James Adams>And of course, we can use it, right?

00:15:50.140 --> 00:15:50.840
<v Chris Gammell>Giving up duties, right?

00:15:50.860 --> 00:15:53.460
<v Chris Gammell>It's just who does what they do best, right?

00:15:53.620 --> 00:15:54.000
<v James Adams>That's right.

00:15:54.000 --> 00:16:01.360
<v James Adams>So on Pi 5, you know, we work with Brawcom to, on 2712, but it is a, you know, it is,

00:16:01.540 --> 00:16:03.040
<v James Adams>it isn't just our chip, right?

00:16:03.100 --> 00:16:06.440
<v James Adams>So it's, it's, we can kind of like work with them to add the features we want.

00:16:06.860 --> 00:16:09.920
<v James Adams>And then we work with Dialog Renegas to do the power management chip.

00:16:10.240 --> 00:16:11.660
<v James Adams>And that kind of thing works well, right?

00:16:12.000 --> 00:16:15.580
<v James Adams>You have to have a much, much bigger team to start attacking those kind of things.

00:16:16.060 --> 00:16:17.620
<v James Adams>And then, and then you've got three chips.

00:16:17.700 --> 00:16:20.580
<v James Adams>Then you've got the main SEC, the South Bridges and microcontrollers.

00:16:20.640 --> 00:16:21.280
<v James Adams>And it's like, oh my God.

00:16:21.280 --> 00:16:22.780
<v James Adams>And then you've got to keep going with all those.

00:16:22.780 --> 00:16:25.360
<v James Adams>That's just, that's just an immense amount of work, right?

00:16:25.440 --> 00:16:28.680
<v James Adams>So we kind of like to attack the things where we can make a big difference.

00:16:29.400 --> 00:16:31.700
<v James Adams>Hopefully that answers the question without kind of revealing.

00:16:31.820 --> 00:16:32.260
<v Chris Gammell>Yeah, no, no.

00:16:32.300 --> 00:16:34.400
<v Chris Gammell>I think that's, you know, that's good for me.

00:16:34.460 --> 00:16:35.760
<v Chris Gammell>I'm not sure if you revealed too much, but.

00:16:36.900 --> 00:16:39.400
<v James Adams>In some ways, a lot of the time we don't quite know.

00:16:39.540 --> 00:16:40.140
<v James Adams>We do experiments.

00:16:40.360 --> 00:16:45.320
<v James Adams>We do, we have different ideas and, you know, and actually the team right now are quite

00:16:45.320 --> 00:16:51.200
<v James Adams>focused on just like spinning up production of RP1, for example, and software on Pi 5.

00:16:51.200 --> 00:16:51.640
<v James Adams>Yeah.

00:16:51.780 --> 00:16:56.360
<v James Adams>And so, you know, it's, it's sort of a little bit of a lull in terms of really getting cracked

00:16:56.360 --> 00:16:57.180
<v James Adams>into any new stuff.

00:16:57.960 --> 00:16:58.120
<v Chris Gammell>Yeah.

00:16:58.260 --> 00:16:58.820
<v Chris Gammell>That makes sense.

00:16:59.180 --> 00:17:05.360
<v Chris Gammell>So on, on the 2040 specifically, I mean, I feel like that the PIOs have been a very interesting

00:17:05.360 --> 00:17:06.460
<v Chris Gammell>adopted thing.

00:17:06.500 --> 00:17:09.380
<v Chris Gammell>I think there's been a lot of people that are interested in using that sort of thing.

00:17:09.380 --> 00:17:15.180
<v Chris Gammell>I mean, are you seeing other, other use cases around the stuff that's in there or just kind

00:17:15.180 --> 00:17:18.620
<v Chris Gammell>of more as a general purpose chip, maybe other, other ecosystems as well.

00:17:18.660 --> 00:17:20.840
<v Chris Gammell>Like I know Zephyr just pulled in 2040 as well.

00:17:20.920 --> 00:17:22.060
<v Chris Gammell>I was excited about that.

00:17:22.540 --> 00:17:23.280
<v James Adams>I didn't see that one.

00:17:24.020 --> 00:17:24.240
<v James Adams>Oh yeah.

00:17:24.460 --> 00:17:24.600
<v Chris Gammell>Yeah.

00:17:24.600 --> 00:17:25.740
<v Chris Gammell>It's community support, but yeah.

00:17:25.900 --> 00:17:26.060
<v James Adams>Yeah.

00:17:26.440 --> 00:17:26.800
<v James Adams>Yeah.

00:17:26.840 --> 00:17:27.240
<v James Adams>Very good.

00:17:27.800 --> 00:17:28.260
<v James Adams>I don't know.

00:17:28.300 --> 00:17:29.140
<v James Adams>Liam, did you want to answer that one?

00:17:29.280 --> 00:17:30.800
<v Liam Fraser>I can, I can give an answer as well.

00:17:31.120 --> 00:17:31.480
<v Liam Fraser>Yeah.

00:17:31.480 --> 00:17:36.480
<v Liam Fraser>I'm just trying to think what projects I've seen recently using, using PIO.

00:17:36.760 --> 00:17:42.960
<v Liam Fraser>I actually saw someone who managed to root a Starlink satellite terminal using RP2040.

00:17:43.700 --> 00:17:44.220
<v Liam Fraser>Yeah.

00:17:44.480 --> 00:17:44.920
<v Liam Fraser>Yeah.

00:17:44.920 --> 00:17:50.260
<v Liam Fraser>And also, you know, it's, it's been, it's been sort of stretched to its limits in many

00:17:50.260 --> 00:17:56.560
<v Liam Fraser>ways with the DVI stuff and like the not HDMI video output that it can do.

00:17:57.400 --> 00:17:57.840
<v Liam Fraser>Yeah.

00:17:57.840 --> 00:17:58.000
<v Liam Fraser>Yeah.

00:17:58.060 --> 00:17:58.420
<v Liam Fraser>But yeah.

00:17:58.520 --> 00:18:00.300
<v Liam Fraser>James, do you want to, do you want to chip in?

00:18:00.300 --> 00:18:04.120
<v Chris Gammell>I can actually give an example if you guys would like, I, I, we're, we are doing video

00:18:04.120 --> 00:18:04.300
<v Chris Gammell>here.

00:18:04.380 --> 00:18:07.480
<v Chris Gammell>Obviously people won't see the one I'm showing here, but a board that I've, I've talked about

00:18:07.480 --> 00:18:11.400
<v Chris Gammell>publicly, I put a Pico onto like a display board and we use the PIO.

00:18:12.040 --> 00:18:18.160
<v Chris Gammell>We, Mike Stish, past guest and my coworker, he basically borrowed an I squared C listener,

00:18:18.460 --> 00:18:18.600
<v Chris Gammell>right?

00:18:18.620 --> 00:18:21.060
<v Chris Gammell>And that was a really good PIO thing.

00:18:21.060 --> 00:18:23.660
<v Chris Gammell>It's just sitting there cranking on I squared C incoming messages.

00:18:23.940 --> 00:18:25.340
<v Chris Gammell>And so it's an I squared C slave.

00:18:25.340 --> 00:18:27.220
<v Chris Gammell>And then we process the incoming messages.

00:18:27.220 --> 00:18:30.940
<v Chris Gammell>It's perfect application of that, of the PIO from my perspective.

00:18:31.040 --> 00:18:31.440
<v James Adams>Yeah.

00:18:31.940 --> 00:18:36.920
<v James Adams>We are seeing a lot more customers being as they, cause I think a lot of people originally

00:18:36.920 --> 00:18:38.180
<v James Adams>didn't really understand what it was.

00:18:38.600 --> 00:18:42.160
<v James Adams>And it's still a little bit of a hard thing to get your head around to.

00:18:42.220 --> 00:18:47.220
<v James Adams>You'd have to do a bit of work to understand it, but everyone who's put the effort in and

00:18:47.220 --> 00:18:48.440
<v James Adams>then sort of realized what it is.

00:18:48.440 --> 00:18:50.580
<v James Adams>They're like, Oh, Oh my God, this is really useful.

00:18:50.580 --> 00:18:55.620
<v James Adams>And we are now seeing a lot of interest in, in this stuff and a lot of cool, cool things.

00:18:55.800 --> 00:18:57.900
<v James Adams>I think the more people use it, the more they like it.

00:18:58.020 --> 00:19:01.120
<v James Adams>It's one of those, one of those things it's, it's a new feature.

00:19:01.320 --> 00:19:04.380
<v James Adams>So most people will ignore it until, until they see something cool.

00:19:04.580 --> 00:19:06.300
<v James Adams>And then they're like, Oh, what is that then?

00:19:06.840 --> 00:19:07.020
<v Chris Gammell>Yeah.

00:19:07.360 --> 00:19:07.540
<v Chris Gammell>Yeah.

00:19:07.540 --> 00:19:10.180
<v Chris Gammell>I remember with the BeagleBone as well, like BeagleBone had kind of like that same, like

00:19:10.180 --> 00:19:15.520
<v Chris Gammell>software defined, like Silicon, you know, it's high speed processing of GPIO and similar,

00:19:15.600 --> 00:19:15.720
<v Chris Gammell>right?

00:19:15.720 --> 00:19:21.140
<v Chris Gammell>Like that, it felt very similar to me, but on the RP2040 is a lot easier to use because

00:19:21.140 --> 00:19:22.600
<v Chris Gammell>it's not like its own compiled.

00:19:22.820 --> 00:19:24.540
<v Chris Gammell>I mean, it's not a whole nother processor basically.

00:19:24.540 --> 00:19:24.840
<v James Adams>Yeah.

00:19:24.940 --> 00:19:29.760
<v James Adams>I think we, the, the idea was always to make it as okay.

00:19:29.760 --> 00:19:33.860
<v James Adams>So I guess the design goal really was to be able to do real, real cycle by cycle, do lots

00:19:33.860 --> 00:19:34.720
<v James Adams>of stuff every cycle.

00:19:35.180 --> 00:19:38.340
<v James Adams>Cause you can attack this kind of problem with, I don't know, just use a general purpose

00:19:38.340 --> 00:19:41.180
<v James Adams>CPU, like another M zero plus that can bit bash.

00:19:41.560 --> 00:19:43.960
<v James Adams>But what you get with PIO is something much smaller.

00:19:43.960 --> 00:19:47.940
<v James Adams>There's actually much more capable of doing things, you know, every cycle, a lot of things,

00:19:48.020 --> 00:19:48.480
<v James Adams>every cycle.

00:19:48.740 --> 00:19:52.880
<v James Adams>So it, although it's a bit, you've got to wrap your head around the, the instruction

00:19:52.880 --> 00:19:55.380
<v James Adams>encoding and how the thing, how the machine actually works.

00:19:55.380 --> 00:19:57.460
<v James Adams>It's, it's much better for that kind of thing.

00:19:57.980 --> 00:19:59.160
<v James Adams>I don't know, Liam, if you've got any more.

00:19:59.420 --> 00:19:59.560
<v Liam Fraser>Yeah.

00:19:59.760 --> 00:20:05.660
<v Liam Fraser>There's a, there's a guy on Twitter whose name I can't remember, but he's had PIO talking

00:20:05.660 --> 00:20:07.320
<v Liam Fraser>to like fiber transceivers.

00:20:07.760 --> 00:20:10.480
<v Liam Fraser>So, you know, doing ethernet over, over fiber.

00:20:10.480 --> 00:20:16.180
<v Liam Fraser>And I think, you know, just being able to talk to arbitrary things and, and shift data

00:20:16.180 --> 00:20:20.660
<v Liam Fraser>to them, you know, as quickly as they require, especially with a clock that's, you know, for

00:20:20.660 --> 00:20:24.280
<v Liam Fraser>that kind of bulk data transfer stuff, it is, it is really helpful.

00:20:24.280 --> 00:20:29.700
<v Liam Fraser>And the, and the display kind of application is, is really an extension of that.

00:20:29.860 --> 00:20:37.200
<v Liam Fraser>And also someone did make a PIO logic analyzer that works with like some open source logic

00:20:37.200 --> 00:20:38.540
<v Liam Fraser>analyzers software.

00:20:38.540 --> 00:20:40.180
<v Liam Fraser>Was that Kumar Abhishek?

00:20:40.300 --> 00:20:41.000
<v Chris Gammell>I saw him.

00:20:41.380 --> 00:20:43.160
<v Chris Gammell>I can't remember what the software's called.

00:20:43.380 --> 00:20:46.340
<v Chris Gammell>I saw Kumar's giving a talk at, at, at Supercon.

00:20:46.540 --> 00:20:49.360
<v Chris Gammell>I was looking at all the Supercon talks that are coming up, but it's going to be there by

00:20:49.360 --> 00:20:49.660
<v Chris Gammell>chance.

00:20:49.880 --> 00:20:50.360
<v Chris Gammell>Superconfident?

00:20:50.560 --> 00:20:51.920
<v Chris Gammell>I'm not.

00:20:51.920 --> 00:20:52.600
<v James Adams>I'm not sure.

00:20:52.600 --> 00:20:53.380
<v James Adams>I'm not sure.

00:20:53.500 --> 00:20:54.760
<v James Adams>I'm not sure if anyone from Pi is.

00:20:54.980 --> 00:20:55.340
<v Chris Gammell>Maybe.

00:20:55.700 --> 00:20:55.880
<v Chris Gammell>Yeah.

00:20:56.260 --> 00:21:01.540
<v Chris Gammell>I'm sure that the 2040 will be prominently, prominently featured around, around the Supercon.

00:21:01.760 --> 00:21:01.940
<v Chris Gammell>Yeah.

00:21:01.940 --> 00:21:02.620
<v Chris Gammell>My favorite places.

00:21:02.620 --> 00:21:08.600
<v Liam Fraser>But yeah, it just hoovers up whatever pins you configure and puts the data over USB so

00:21:08.600 --> 00:21:10.700
<v Liam Fraser>you can see it in like a logic analyzer.

00:21:11.020 --> 00:21:11.760
<v Liam Fraser>It's really nice.

00:21:12.920 --> 00:21:13.480
<v Chris Gammell>That's cool.

00:21:13.780 --> 00:21:13.920
<v Chris Gammell>Yeah.

00:21:13.920 --> 00:21:15.200
<v Chris Gammell>That's, that's another, yeah.

00:21:15.200 --> 00:21:18.840
<v Chris Gammell>Another good example of something that's just like kind of sitting there crunching on inputs,

00:21:18.940 --> 00:21:20.120
<v Chris Gammell>outputs, that sort of thing.

00:21:20.380 --> 00:21:21.220
<v Chris Gammell>It's good.

00:21:21.300 --> 00:21:22.000
<v Chris Gammell>Good use case.

00:21:22.320 --> 00:21:25.100
<v Chris Gammell>I, you know, I think that's the kind of thing too, where it's like you guys said, it's,

00:21:25.500 --> 00:21:27.100
<v Chris Gammell>you don't know you need it until you do.

00:21:27.100 --> 00:21:31.140
<v Chris Gammell>And so I'm sure some of the, the coolest applications are still to come, which is great.

00:21:31.460 --> 00:21:38.900
<v Chris Gammell>And yeah, I mean, to talk a little bit about the, the sourcing side of things too, like,

00:21:39.400 --> 00:21:40.420
<v Chris Gammell>how do you guys keep it?

00:21:40.560 --> 00:21:42.020
<v Chris Gammell>Was it just like the timing was perfect?

00:21:42.120 --> 00:21:45.380
<v Chris Gammell>I mean, like, I mean, we talked about it and then it was just, I think it was like,

00:21:45.640 --> 00:21:51.540
<v Chris Gammell>you know, we talked in 2021, February, 2021, and like things got so, so bad in 21 into 22.

00:21:51.840 --> 00:21:54.720
<v Chris Gammell>And you guys were just like, we got tons of chips.

00:21:54.720 --> 00:21:55.900
<v Chris Gammell>So I just kept seeing them everywhere.

00:21:57.100 --> 00:21:57.340
<v James Adams>Yeah.

00:21:58.140 --> 00:22:02.860
<v James Adams>I mean, to be fair, we, I think we underestimated how bad it would get actually.

00:22:03.200 --> 00:22:06.660
<v James Adams>And obviously, you know, Raspberry Pis, the big, the big Pis, right.

00:22:06.900 --> 00:22:08.220
<v James Adams>Well, have been hard to come by.

00:22:08.620 --> 00:22:12.100
<v James Adams>However, that is massively improving now and will continue to improve.

00:22:12.440 --> 00:22:17.400
<v James Adams>But for RP2040, so we, we kind of bought a lot of wafers upfront anyway.

00:22:17.740 --> 00:22:22.940
<v James Adams>And actually because we're a kind of a small player, you know, we don't build that much silicon.

00:22:23.400 --> 00:22:26.760
<v James Adams>We could get offers on little bits of wafers here and there that we could take up.

00:22:26.760 --> 00:22:31.240
<v James Adams>Whereas someone like, I don't know, a Qualcomm or a big guy, they have like a wafer allocation

00:22:31.240 --> 00:22:33.060
<v James Adams>and that just got trimmed.

00:22:33.440 --> 00:22:36.600
<v James Adams>And it's like, they can't, you know, if they get an extra three wafers, it really doesn't

00:22:36.600 --> 00:22:37.160
<v James Adams>matter to them.

00:22:37.540 --> 00:22:38.980
<v James Adams>You know, they're not in that game.

00:22:38.980 --> 00:22:41.440
<v James Adams>Whereas for us, you know, there's 20,000 chips of wafer.

00:22:41.440 --> 00:22:46.080
<v James Adams>If some, if, if we get offered, oh, you know, we've got a few wafers spared, you, do you

00:22:46.080 --> 00:22:46.380
<v James Adams>want them?

00:22:46.440 --> 00:22:47.400
<v James Adams>We're like, oh yeah, we'll have them.

00:22:48.040 --> 00:22:53.060
<v James Adams>So actually the kind of just the scale and the fact that we actually did actually pre-buy

00:22:53.060 --> 00:22:54.280
<v James Adams>a fair quantity.

00:22:54.840 --> 00:22:55.160
<v James Adams>Yeah.

00:22:55.580 --> 00:22:58.220
<v James Adams>Meant that we, you know, we had some good stock.

00:22:58.680 --> 00:22:59.640
<v Chris Gammell>Could you explain that interaction?

00:22:59.820 --> 00:23:02.780
<v Chris Gammell>So like, James, are you the one that actually like, do you talk to the fabs?

00:23:02.780 --> 00:23:04.220
<v Chris Gammell>Like how does that conversation go?

00:23:04.660 --> 00:23:06.680
<v Chris Gammell>Like when you have three wafers, right?

00:23:06.680 --> 00:23:09.360
<v Chris Gammell>Or like, is it just like you have like, you get an email?

00:23:10.440 --> 00:23:15.780
<v James Adams>So we actually, we actually buy our wafers through iMEC, which is an aggregator.

00:23:16.300 --> 00:23:20.340
<v James Adams>And all we did, when you get to a certain scale, then you can talk to TSMC directly.

00:23:20.700 --> 00:23:23.480
<v James Adams>But you know, those guys are managing a bunch of customers.

00:23:23.780 --> 00:23:27.340
<v James Adams>And I think it's also true that TSMC occasionally have a few, a few spare here and there.

00:23:27.400 --> 00:23:29.420
<v James Adams>Someone maybe deallocates or decommit something.

00:23:29.700 --> 00:23:30.020
<v James Adams>Yeah.

00:23:30.080 --> 00:23:31.160
<v James Adams>And it's usually small numbers.

00:23:31.560 --> 00:23:35.780
<v James Adams>And basically, you know, we, we just get sort of an offer here and there.

00:23:35.780 --> 00:23:37.920
<v James Adams>And it's usually for a reasonable number.

00:23:38.100 --> 00:23:40.140
<v James Adams>And so we could take them.

00:23:40.240 --> 00:23:41.440
<v James Adams>And it's probably more.

00:23:41.720 --> 00:23:44.000
<v Chris Gammell>And when you say that we get a wafer, right?

00:23:44.060 --> 00:23:47.520
<v Chris Gammell>You don't, you mean there's a raw wafer available, raw piece of silicon.

00:23:47.700 --> 00:23:50.100
<v Chris Gammell>And then all of the process has already been defined.

00:23:50.100 --> 00:23:51.940
<v Chris Gammell>So they're like, we already know how to make this.

00:23:52.000 --> 00:23:53.280
<v Chris Gammell>We're just going to be able to run it through the line.

00:23:53.400 --> 00:23:54.160
<v Chris Gammell>Is that, is that kind of a...

00:23:54.160 --> 00:23:54.100
<v Speaker ?>Yeah.

00:23:54.100 --> 00:23:56.260
<v James Adams>So TSMC, TSMC have our mask.

00:23:56.380 --> 00:23:59.580
<v James Adams>They can make our chips and they have various fabs.

00:23:59.660 --> 00:24:02.720
<v James Adams>And each fab can do so many wafer starts a month.

00:24:02.840 --> 00:24:03.600
<v James Adams>They call it wafer starts.

00:24:03.920 --> 00:24:04.280
<v James Adams>Yep.

00:24:04.280 --> 00:24:07.440
<v James Adams>And, you know, those wafer starts will have been booked.

00:24:08.100 --> 00:24:11.460
<v James Adams>Like, you know, NVIDIA will book so much, so many, so many wafers a month because they

00:24:11.460 --> 00:24:14.120
<v James Adams>make zillions of chips and Qualcomm, Qualcomm, all these guys.

00:24:14.240 --> 00:24:16.000
<v James Adams>And that's all allocated and managed.

00:24:16.140 --> 00:24:20.280
<v James Adams>And in normal times, there's enough supply and demand and that's all well balanced.

00:24:20.420 --> 00:24:25.700
<v James Adams>And what happened in the chip shortages, you know, people kind of didn't expect the demand

00:24:25.700 --> 00:24:27.140
<v James Adams>for silicon to actually increase.

00:24:27.260 --> 00:24:28.500
<v James Adams>In fact, they expected to decrease.

00:24:28.500 --> 00:24:34.160
<v James Adams>And then because it's a bit like oil, I think, because there's quite a tight supply demand,

00:24:34.240 --> 00:24:35.320
<v James Adams>it's always managed fairly tightly.

00:24:35.320 --> 00:24:39.860
<v James Adams>All of a sudden, when it gets out of sync, it really gets out of sync.

00:24:40.860 --> 00:24:45.540
<v James Adams>And so the bigger silicon guys, they just kind of had to take a bit of a haircut on the quantity

00:24:45.540 --> 00:24:46.360
<v James Adams>that would be allocated.

00:24:46.360 --> 00:24:51.460
<v James Adams>But people like iMEC, who are an aggregator, who manage kind of multiple smaller accounts,

00:24:52.020 --> 00:24:54.080
<v James Adams>they maybe still have like 100 customers.

00:24:54.420 --> 00:24:57.960
<v James Adams>And one customer says, OK, I don't need these, you know, 10 wafers.

00:24:58.380 --> 00:25:01.220
<v James Adams>Because, yeah, I just, maybe I can't get silicon for my other project.

00:25:01.220 --> 00:25:02.920
<v James Adams>And therefore, I don't need it.

00:25:02.920 --> 00:25:08.360
<v James Adams>So there tends to be a small number of wafers floating around occasionally like this, right?

00:25:08.540 --> 00:25:15.620
<v James Adams>And because we're in the loop, we could sort of, you know, we just ask, you know, we'd said

00:25:15.620 --> 00:25:18.100
<v James Adams>to iMEC, if you have spare, give us a shout.

00:25:18.780 --> 00:25:21.980
<v James Adams>And we could pick up some of the stock.

00:25:22.560 --> 00:25:25.420
<v James Adams>And yeah, again, because RP2040 is a very small chip.

00:25:25.640 --> 00:25:26.980
<v James Adams>It's very economical with silicon.

00:25:27.380 --> 00:25:28.880
<v James Adams>A few wafers goes a long way for us.

00:25:29.760 --> 00:25:30.860
<v Chris Gammell>Yeah, 20,000 per wafers.

00:25:30.860 --> 00:25:33.320
<v Chris Gammell>There's, yeah, that's pretty crazy.

00:25:33.720 --> 00:25:40.240
<v Chris Gammell>I used to, so I used to work on, at Samsung in the 2006, the 8, and I was doing memory.

00:25:40.480 --> 00:25:43.920
<v Chris Gammell>And so they were, you know, not huge chips, but they weren't sized down yet.

00:25:43.980 --> 00:25:45.340
<v Chris Gammell>And they were much bigger than that.

00:25:45.340 --> 00:25:48.020
<v Chris Gammell>So like much fewer per wafer, that sort of thing.

00:25:48.100 --> 00:25:50.140
<v Chris Gammell>And it's also like 300 millimeter, right?

00:25:50.640 --> 00:25:50.940
<v James Adams>Yes.

00:25:51.120 --> 00:25:51.260
<v James Adams>Yeah.

00:25:51.560 --> 00:25:51.760
<v James Adams>Yeah.

00:25:52.320 --> 00:25:55.420
<v James Adams>So I think we're about, we sort of say headline two square mil.

00:25:55.440 --> 00:25:57.420
<v James Adams>I think we're just a bit bigger than two square millimeters.

00:25:57.420 --> 00:26:01.000
<v James Adams>But yeah, you still got a lot of diaper, a good diaper wafer.

00:26:01.200 --> 00:26:01.340
<v James Adams>Right.

00:26:01.560 --> 00:26:01.780
<v James Adams>Right.

00:26:02.140 --> 00:26:07.280
<v Chris Gammell>And then it's like, in terms of like yield, it's high enough that it's like, you just,

00:26:07.440 --> 00:26:08.820
<v Chris Gammell>you don't worry about it, that sort of thing.

00:26:08.880 --> 00:26:10.080
<v Chris Gammell>Like it's, it's dialed in enough.

00:26:10.180 --> 00:26:10.700
<v James Adams>Pretty much.

00:26:11.040 --> 00:26:11.220
<v James Adams>Yeah.

00:26:11.220 --> 00:26:19.100
<v James Adams>I think the small die on a, on a well, well known mature process, like TSMC40 yields very

00:26:19.100 --> 00:26:19.440
<v James Adams>well.

00:26:20.100 --> 00:26:22.060
<v James Adams>So yeah, we just package them up.

00:26:22.660 --> 00:26:23.720
<v James Adams>We don't test the wafer.

00:26:23.820 --> 00:26:25.560
<v James Adams>We just package them up and then we throw away the dead ones.

00:26:25.560 --> 00:26:30.220
<v James Adams>I think we're kind of like, you know, 95, 96% yield, something like that.

00:26:30.920 --> 00:26:31.020
<v Chris Gammell>Yeah.

00:26:31.320 --> 00:26:34.220
<v Chris Gammell>And yeah, there's benefits of being not on leading edge node too.

00:26:34.420 --> 00:26:35.360
<v Chris Gammell>That's, that's pretty interesting.

00:26:35.680 --> 00:26:35.840
<v Speaker ?>Yeah.

00:26:36.120 --> 00:26:36.320
<v Chris Gammell>Yeah.

00:26:36.680 --> 00:26:36.880
<v Chris Gammell>Huh.

00:26:37.660 --> 00:26:41.560
<v Chris Gammell>What does it take to get, so you said like a hundred people are at iMac, right?

00:26:42.500 --> 00:26:44.520
<v Chris Gammell>How big do you have to be to get to that point?

00:26:44.660 --> 00:26:45.040
<v Chris Gammell>I mean, like.

00:26:45.780 --> 00:26:47.120
<v James Adams>iMac helps startups, right?

00:26:47.120 --> 00:26:49.740
<v James Adams>So that's, that's kind of what they do.

00:26:49.820 --> 00:26:53.220
<v James Adams>They aggregate because big guys like TSMC, they don't want to, they don't want to talk

00:26:53.220 --> 00:26:55.560
<v James Adams>to the really small guys because there's so many of them and they can't, right?

00:26:55.620 --> 00:26:56.560
<v James Adams>They're, they're big.

00:26:57.160 --> 00:27:01.940
<v James Adams>So they talk to someone like iMac, which aggregates all these chip, these smaller chip companies.

00:27:02.560 --> 00:27:06.100
<v James Adams>And originally they helped us with a lot of the chip stuff.

00:27:06.560 --> 00:27:11.740
<v James Adams>As we've got bigger, we've gradually sort of done more and more of it ourselves, but they

00:27:11.740 --> 00:27:13.560
<v James Adams>also buy the wafers for you.

00:27:13.620 --> 00:27:13.760
<v James Adams>Yeah.

00:27:13.760 --> 00:27:14.680
<v James Adams>If you see, if you see what I mean.

00:27:14.680 --> 00:27:15.500
<v James Adams>Yeah.

00:27:15.800 --> 00:27:16.160
<v James Adams>Yeah.

00:27:16.160 --> 00:27:17.300
<v James Adams>Like cashflow type of stuff.

00:27:17.600 --> 00:27:17.640
<v James Adams>Yeah.

00:27:17.640 --> 00:27:20.700
<v James Adams>It's kind of like an, I guess, ASIC incubator type.

00:27:20.700 --> 00:27:22.720
<v Chris Gammell>Just distributor kind of thing almost.

00:27:23.120 --> 00:27:23.380
<v James Adams>Yeah.

00:27:23.660 --> 00:27:24.020
<v James Adams>Yeah.

00:27:24.020 --> 00:27:24.600
<v James Adams>Pretty much.

00:27:25.480 --> 00:27:29.540
<v James Adams>So that, you know, they were, they, they helped us bootstrap into the chip world.

00:27:29.540 --> 00:27:33.420
<v James Adams>As almost any chip startup will do, they'll, they'll go to one of these incubators.

00:27:33.520 --> 00:27:36.500
<v James Adams>Imex probably the, I think they're the biggest one and most well known.

00:27:36.860 --> 00:27:39.500
<v James Adams>And they kind of help, help you, help you get going and yeah.

00:27:39.500 --> 00:27:40.480
<v James Adams>Help supply your wafers.

00:27:40.480 --> 00:27:45.380
<v James Adams>Even if you only need a few, they can provide that kind of, that kind of service where you

00:27:45.380 --> 00:27:46.780
<v James Adams>couldn't just go and talk to TSMC.

00:27:46.780 --> 00:27:51.400
<v Liam Fraser>And I guess the other nice thing is as well, we didn't have to decide how to allocate our

00:27:51.400 --> 00:27:54.620
<v Liam Fraser>wafers because we only had one, you know, we only had RP2040.

00:27:54.900 --> 00:27:57.820
<v Liam Fraser>So we didn't have to divide up a smaller number.

00:27:58.880 --> 00:27:59.340
<v Liam Fraser>That's right.

00:27:59.500 --> 00:27:59.620
<v Liam Fraser>Yeah.

00:28:00.420 --> 00:28:05.440
<v Chris Gammell>Oh, you mean like if, if you were making the RP1 at the same time and then those had a

00:28:05.440 --> 00:28:06.320
<v Chris Gammell>lot of demand.

00:28:06.580 --> 00:28:06.880
<v Chris Gammell>Exactly.

00:28:07.020 --> 00:28:07.760
<v Chris Gammell>We'd have to pick.

00:28:08.160 --> 00:28:08.760
<v Chris Gammell>High five.

00:28:09.340 --> 00:28:09.580
<v Chris Gammell>Yeah.

00:28:09.760 --> 00:28:10.080
<v Chris Gammell>Right.

00:28:10.080 --> 00:28:10.520
<v Chris Gammell>Yeah.

00:28:10.640 --> 00:28:10.840
<v Chris Gammell>Yeah.

00:28:10.920 --> 00:28:11.940
<v Chris Gammell>And that does sound, yeah.

00:28:12.020 --> 00:28:17.880
<v Chris Gammell>Like I think about the companies that had like a TI that had like automotive, like breathing

00:28:17.880 --> 00:28:18.500
<v Chris Gammell>down their neck.

00:28:18.500 --> 00:28:21.540
<v Chris Gammell>It's like Chris Gammell is not getting any chips.

00:28:21.540 --> 00:28:27.040
<v Chris Gammell>Like I am buying from the bottom of the bottom of the barrel.

00:28:27.240 --> 00:28:29.640
<v Chris Gammell>Like I'm, I'm just the last in line.

00:28:29.880 --> 00:28:30.940
<v Chris Gammell>So it's fine.

00:28:30.980 --> 00:28:31.100
<v Chris Gammell>Right.

00:28:31.120 --> 00:28:32.240
<v Chris Gammell>I mean, like that's how it goes.

00:28:32.900 --> 00:28:37.100
<v James Adams>I mean, it's really great that we could supply people with silicon in a, you know, a really

00:28:37.100 --> 00:28:38.540
<v James Adams>probably the worst silicon shortage.

00:28:38.540 --> 00:28:40.140
<v James Adams>I think anyone has ever seen it in the industry.

00:28:40.680 --> 00:28:45.420
<v James Adams>So we have our commercial guy has been in the industry basically forever and it's the

00:28:45.420 --> 00:28:46.420
<v James Adams>worst he's ever seen.

00:28:46.660 --> 00:28:50.740
<v James Adams>I think if you go and talk to the really seasoned guys who've been here for a long time, yeah,

00:28:50.820 --> 00:28:52.340
<v James Adams>no one's ever seen anything like that.

00:28:53.260 --> 00:28:57.260
<v James Adams>But, you know, on the flip side, we're now aware that this kind of stuff can happen.

00:28:57.960 --> 00:29:02.220
<v James Adams>And I think people are, you know, the fact that they're aware of that kind of risk is

00:29:02.220 --> 00:29:02.640
<v James Adams>useful.

00:29:03.760 --> 00:29:03.980
<v Chris Gammell>Oh, sure.

00:29:04.180 --> 00:29:07.160
<v Chris Gammell>Until, until the people retire and then, then it all happens again.

00:29:07.160 --> 00:29:07.440
<v Chris Gammell>Right.

00:29:08.540 --> 00:29:08.980
<v Speaker ?>Yeah.

00:29:08.980 --> 00:29:09.300
<v James Adams>Yeah.

00:29:09.300 --> 00:29:10.680
<v James Adams>Hopefully we're kind of done with that.

00:29:11.100 --> 00:29:13.900
<v James Adams>I'm sure there'll be other, other things at some point in the future, but you know,

00:29:13.900 --> 00:29:16.140
<v James Adams>that was a really bad, that was a really bad situation.

00:29:16.280 --> 00:29:16.420
<v James Adams>Oh yeah.

00:29:16.880 --> 00:29:17.080
<v Chris Gammell>Yeah.

00:29:17.660 --> 00:29:18.060
<v Chris Gammell>Yes.

00:29:18.460 --> 00:29:23.340
<v Chris Gammell>My buddy who's in purchasing has told me stories that I, I just, that are somewhat

00:29:23.340 --> 00:29:24.280
<v Chris Gammell>unbelievable, but.

00:29:25.900 --> 00:29:30.840
<v James Adams>I think everyone on both sides, purchasing and supply have just had a really tough couple

00:29:30.840 --> 00:29:31.220
<v James Adams>of years.

00:29:31.600 --> 00:29:31.680
<v James Adams>So.

00:29:31.800 --> 00:29:31.920
<v James Adams>Yeah.

00:29:32.280 --> 00:29:32.560
<v James Adams>Yeah.

00:29:32.800 --> 00:29:33.000
<v James Adams>Yeah.

00:29:33.000 --> 00:29:34.980
<v Chris Gammell>Hopefully we get a break just to chill out.

00:29:34.980 --> 00:29:37.420
<v Chris Gammell>That's great.

00:29:37.480 --> 00:29:37.880
<v Chris Gammell>So, okay.

00:29:37.880 --> 00:29:40.520
<v Chris Gammell>So let's, let's go a little bit back into RP one now.

00:29:40.580 --> 00:29:42.540
<v Chris Gammell>So you, you had the 2040, you've learned all these things.

00:29:42.540 --> 00:29:47.540
<v Chris Gammell>You had the scripting, you have the, the learning from PIO and similar kinds of things.

00:29:47.540 --> 00:29:53.360
<v Chris Gammell>And now you're putting in kind of, you already had some stuff in there, but now you're putting

00:29:53.360 --> 00:29:59.240
<v Chris Gammell>in kind of a higher speed, higher complexity things like ethernet, like USB three, things

00:29:59.240 --> 00:29:59.680
<v Chris Gammell>like that.

00:29:59.800 --> 00:30:03.340
<v Chris Gammell>One question I had about that, you mentioned you started in 2015.

00:30:04.200 --> 00:30:05.920
<v Chris Gammell>Does IP get old?

00:30:06.180 --> 00:30:09.000
<v Chris Gammell>Like, do you have like a new version that you update to?

00:30:09.100 --> 00:30:14.460
<v Chris Gammell>So like, if you buy an IP block in 2015, there's probably a better IP block in 2022,

00:30:14.460 --> 00:30:15.300
<v Chris Gammell>right.

00:30:15.320 --> 00:30:17.080
<v Chris Gammell>Or whenever you update it.

00:30:17.080 --> 00:30:21.740
<v Chris Gammell>So like, how much does it change from one generation to the next?

00:30:22.860 --> 00:30:23.720
<v Liam Fraser>I can answer that.

00:30:24.000 --> 00:30:27.320
<v Liam Fraser>It really depends on sort of IP to IP.

00:30:27.540 --> 00:30:32.520
<v Liam Fraser>I think the IP vendors always recommend that when you do a tape out, you sort of pick the

00:30:32.520 --> 00:30:34.820
<v Liam Fraser>latest version of the IP.

00:30:35.540 --> 00:30:36.540
<v Chris Gammell>It's like a package update.

00:30:36.700 --> 00:30:41.140
<v Chris Gammell>It's like a, you do like an NPN, NPM update and then you get the newest USB three.

00:30:41.140 --> 00:30:41.500
<v Chris Gammell>Yeah.

00:30:41.640 --> 00:30:45.680
<v Liam Fraser>Well, once, obviously once we've sort of verified it, then, then we have to freeze the design

00:30:45.680 --> 00:30:49.120
<v Liam Fraser>and, and we won't, you know, once it's gone to TSMC, we won't, we won't be changing it

00:30:49.120 --> 00:30:51.880
<v Liam Fraser>unless it's a sort of a different version of the chip.

00:30:52.120 --> 00:30:57.620
<v Liam Fraser>But yeah, it's, you know, IP that you buy in still has bugs and they will sort of issue,

00:30:57.800 --> 00:31:00.660
<v Liam Fraser>you know, notifications of, of, of bugs.

00:31:00.660 --> 00:31:04.100
<v Liam Fraser>And then it's up to you to decide whether you want to take the bug fix or not.

00:31:04.560 --> 00:31:08.380
<v Liam Fraser>But yeah, some, sometimes we, we take a version change and nothing's changed.

00:31:08.380 --> 00:31:12.460
<v Liam Fraser>And sometimes you take a version change, do a diff and the whole thing's been rewritten.

00:31:12.680 --> 00:31:16.960
<v Liam Fraser>So it's a, you know, and, and if that does happen, then you're kind of like, oh, do we,

00:31:17.120 --> 00:31:19.820
<v Liam Fraser>do we want to take this, this new version?

00:31:19.960 --> 00:31:20.800
<v Liam Fraser>Because we're not sure.

00:31:21.320 --> 00:31:21.480
<v Chris Gammell>Got it.

00:31:21.920 --> 00:31:23.320
<v Chris Gammell>And so, and you're saying diffs as well.

00:31:23.400 --> 00:31:25.260
<v Chris Gammell>So is it being delivered as Verilog?

00:31:25.380 --> 00:31:28.340
<v Chris Gammell>I mean, or is it, is it actually more piled down to.

00:31:28.340 --> 00:31:34.420
<v Liam Fraser>So, so it tends to get delivered as Verilog and then we check all of our stuff into Git.

00:31:34.600 --> 00:31:39.400
<v Liam Fraser>So that allows us to, you know, diff it and see, see what the difference is.

00:31:39.860 --> 00:31:44.660
<v Liam Fraser>And yeah, every, you know, when we tape out a chip, we like tag releases of everything.

00:31:44.660 --> 00:31:47.400
<v Liam Fraser>So we know exactly what we've, what we've got.

00:31:47.580 --> 00:31:52.360
<v Liam Fraser>And the chip itself knows its chip ID, sort of its Git hash as it was taped out.

00:31:52.360 --> 00:31:55.500
<v Liam Fraser>So you can always go back and see what you made.

00:31:56.960 --> 00:31:57.900
<v Chris Gammell>Positively software-y.

00:31:58.060 --> 00:31:58.340
<v Chris Gammell>Yeah.

00:31:58.800 --> 00:32:02.280
<v Chris Gammell>I mean, like, I know it happens in software, but it still like feels weird to me like that,

00:32:02.360 --> 00:32:05.480
<v Chris Gammell>to like hear it like that, you know, like, of course it's all software, but like, yeah.

00:32:05.900 --> 00:32:06.120
<v Chris Gammell>Yeah.

00:32:06.140 --> 00:32:10.860
<v Chris Gammell>In my head, I'm like, well, no, someone's like moving this transistor from A to B, right?

00:32:10.980 --> 00:32:12.100
<v Chris Gammell>It's like, no, that's not how it works.

00:32:12.480 --> 00:32:12.640
<v Chris Gammell>Yeah.

00:32:13.340 --> 00:32:15.260
<v James Adams>Oh, we have a little tiny bit of that.

00:32:15.360 --> 00:32:19.140
<v James Adams>Maybe not the transistor level, but when you get to the fettling of the chip before you kick

00:32:19.140 --> 00:32:23.660
<v James Adams>it out the door, the physical design guys, often what we call, who we call the backend

00:32:23.660 --> 00:32:25.380
<v James Adams>guys, you know, they are looking at the wires.

00:32:25.500 --> 00:32:29.120
<v James Adams>They're looking at the physical transistors and the, and the different blocks and moving

00:32:29.120 --> 00:32:31.360
<v James Adams>them slightly and connecting wires and things.

00:32:31.420 --> 00:32:35.640
<v James Adams>So there's a bit of that, uh, looking at the ESD structures and, you know, there's quite

00:32:35.640 --> 00:32:38.780
<v James Adams>a bit of hand fettling at the end for sure.

00:32:38.860 --> 00:32:39.080
<v Chris Gammell>Yeah.

00:32:39.080 --> 00:32:42.140
<v James Adams>Cause you've got to kind of broadly tell the tools, you know, where does, where does this

00:32:42.140 --> 00:32:42.940
<v James Adams>stuff go as well?

00:32:43.100 --> 00:32:44.440
<v James Adams>You know, how's the layout going to be?

00:32:44.480 --> 00:32:45.100
<v James Adams>Where are all the pins?

00:32:45.400 --> 00:32:45.520
<v James Adams>Right.

00:32:45.880 --> 00:32:46.040
<v James Adams>Yeah.

00:32:46.040 --> 00:32:47.300
<v James Adams>Kind of placing all the blocks.

00:32:47.920 --> 00:32:48.040
<v James Adams>Yeah.

00:32:48.040 --> 00:32:49.640
<v James Adams>Silicon, Silicon auto routers, right?

00:32:51.460 --> 00:32:53.040
<v James Adams>Well, I mean, yeah, most of it is right.

00:32:53.140 --> 00:32:53.280
<v Chris Gammell>Yeah.

00:32:53.400 --> 00:32:53.820
<v James Adams>Placing route.

00:32:53.940 --> 00:32:54.080
<v James Adams>Yeah.

00:32:54.220 --> 00:32:54.400
<v James Adams>Totally.

00:32:55.040 --> 00:32:59.040
<v James Adams>Describe the chip in Verilog, which is a sort of a higher level description language, and

00:32:59.040 --> 00:33:03.400
<v James Adams>then compile it down into gates and then pull the gates into the, into the box in the chip.

00:33:03.440 --> 00:33:08.320
<v James Adams>And then the, basically the tool kind of goes and goes and puts them in, in places and

00:33:08.320 --> 00:33:08.680
<v James Adams>routes them.

00:33:09.600 --> 00:33:12.660
<v James Adams>But then when you get to the, when you get to kind of the edges of the box, then you do

00:33:12.660 --> 00:33:13.980
<v James Adams>have to do a bit more manual work.

00:33:13.980 --> 00:33:14.300
<v Chris Gammell>Right.

00:33:14.640 --> 00:33:14.920
<v Chris Gammell>Yep.

00:33:15.100 --> 00:33:15.420
<v Chris Gammell>Yep.

00:33:15.420 --> 00:33:21.940
<v Chris Gammell>I guess I always think about like, obviously there's a lot of high value IP that's in there.

00:33:21.940 --> 00:33:22.220
<v Chris Gammell>Right.

00:33:22.220 --> 00:33:24.700
<v Chris Gammell>So like not cheap things, right.

00:33:24.700 --> 00:33:28.880
<v Chris Gammell>You guys are dealing with IP that's expensive because it was very expensive to manufacture

00:33:28.880 --> 00:33:29.460
<v Chris Gammell>and stuff like that.

00:33:29.580 --> 00:33:32.560
<v Chris Gammell>Is there any interesting process around that?

00:33:32.560 --> 00:33:37.180
<v Chris Gammell>Like in my mind, I think of like, like a golden USB key that has walked through the door with,

00:33:37.300 --> 00:33:39.380
<v Chris Gammell>with, you know, with monks in robes.

00:33:39.380 --> 00:33:42.920
<v Chris Gammell>But it, or is it just like literally like, oh, here's, we'll add you to the Git repo.

00:33:42.920 --> 00:33:46.600
<v Liam Fraser>Oh, I can probably talk about this as well.

00:33:46.720 --> 00:33:51.620
<v Liam Fraser>Uh, so I also, well, uh, one of my many hats, uh, raspberry pie is sysadmin.

00:33:51.620 --> 00:33:57.300
<v Liam Fraser>So I, uh, when I first started, I sort of, uh, had a sysadmin background and then I've

00:33:57.300 --> 00:33:58.820
<v Liam Fraser>ended up doing a lot of other stuff as well.

00:33:58.820 --> 00:34:03.060
<v Liam Fraser>So we keep all of our servers on for ASIC on site.

00:34:03.220 --> 00:34:03.640
<v Liam Fraser>Cool.

00:34:03.900 --> 00:34:04.020
<v Liam Fraser>Yeah.

00:34:04.120 --> 00:34:04.540
<v Liam Fraser>Partially.

00:34:04.680 --> 00:34:09.960
<v Liam Fraser>So we know where they are and partially because we need a fair amount of compute.

00:34:10.200 --> 00:34:17.120
<v Liam Fraser>So, uh, I think each, each server's got like 40 calls, 80 threads, you know, 5, 12 gig of

00:34:17.120 --> 00:34:17.420
<v Liam Fraser>RAM.

00:34:17.420 --> 00:34:22.140
<v Liam Fraser>And there's like six of those and that, that kind of performance in AWS.

00:34:22.140 --> 00:34:26.680
<v Liam Fraser>It's a lot cheaper to buy the physical thing and plug it in.

00:34:27.380 --> 00:34:27.860
<v Chris Gammell>Yep.

00:34:27.860 --> 00:34:28.500
<v Liam Fraser>Yeah.

00:34:28.580 --> 00:34:32.340
<v Liam Fraser>So we know, you know, where our servers are, it's all encrypted.

00:34:32.880 --> 00:34:35.740
<v Liam Fraser>You know, you need an SH key to get onto it, all that kind of stuff.

00:34:35.920 --> 00:34:36.200
<v Chris Gammell>Yeah.

00:34:36.420 --> 00:34:36.600
<v Liam Fraser>Yeah.

00:34:36.640 --> 00:34:36.980
<v Liam Fraser>That makes sense.

00:34:37.080 --> 00:34:38.760
<v Liam Fraser>And everyone logs into them.

00:34:39.460 --> 00:34:43.760
<v Liam Fraser>So, uh, so, you know, you, you can't like copy stuff to your laptop, that, that kind of

00:34:43.760 --> 00:34:44.000
<v Liam Fraser>thing.

00:34:44.100 --> 00:34:44.500
<v Liam Fraser>Ah, yeah.

00:34:44.500 --> 00:34:49.780
<v Liam Fraser>And then also our backups are in a known, you know, with a known company and a known

00:34:49.780 --> 00:34:50.360
<v Liam Fraser>data center.

00:34:50.560 --> 00:34:56.540
<v Liam Fraser>So we sort of know, know where all the stuff is, uh, and yeah, try and, uh, try and control

00:34:56.540 --> 00:34:57.460
<v Liam Fraser>access to it.

00:34:57.460 --> 00:34:57.920
<v Liam Fraser>Got it.

00:34:57.980 --> 00:34:58.780
<v Liam Fraser>And pretty tightly.

00:34:59.060 --> 00:34:59.240
<v Liam Fraser>Yeah.

00:34:59.360 --> 00:35:01.240
<v Chris Gammell>I wouldn't call that paranoid, but that's like responsible.

00:35:01.300 --> 00:35:03.220
<v Chris Gammell>Like it's just not, it's not super webby.

00:35:03.220 --> 00:35:03.480
<v Chris Gammell>Right.

00:35:03.480 --> 00:35:07.660
<v Chris Gammell>I mean, that's, that's kind of interesting because you, I would kind of imagine AWS and

00:35:07.660 --> 00:35:12.600
<v Chris Gammell>Azure and all the, the big processors, they love to talk about like instant compute availability,

00:35:12.600 --> 00:35:15.200
<v Chris Gammell>but, but they do love to charge for it too.

00:35:15.200 --> 00:35:15.540
<v Chris Gammell>Don't they?

00:35:15.720 --> 00:35:16.040
<v Chris Gammell>Oh yeah.

00:35:18.280 --> 00:35:19.360
<v Chris Gammell>It's like, okay guys.

00:35:19.440 --> 00:35:22.220
<v Chris Gammell>Well, there is some, there is some crossover point where it's like, no, no, no, we'll just

00:35:22.220 --> 00:35:23.060
<v Chris Gammell>we'll buy our own computers.

00:35:23.200 --> 00:35:23.720
<v Chris Gammell>Thank you very much.

00:35:23.800 --> 00:35:23.960
<v Chris Gammell>You know?

00:35:24.340 --> 00:35:24.660
<v Chris Gammell>Yeah.

00:35:24.660 --> 00:35:25.420
<v Chris Gammell>That makes a lot of sense.

00:35:26.700 --> 00:35:27.020
<v Chris Gammell>Yeah.

00:35:27.020 --> 00:35:28.200
<v Chris Gammell>That's cool.

00:35:28.340 --> 00:35:28.520
<v Chris Gammell>Okay.

00:35:28.620 --> 00:35:28.800
<v Chris Gammell>Yeah.

00:35:28.800 --> 00:35:29.220
<v Chris Gammell>That's great.

00:35:29.340 --> 00:35:33.060
<v Chris Gammell>So, but no, but no golden keys and monks and stuff like that.

00:35:33.080 --> 00:35:33.200
<v Chris Gammell>Huh?

00:35:34.280 --> 00:35:35.480
<v Chris Gammell>I mean, you guys are in Cambridge too.

00:35:35.540 --> 00:35:39.000
<v Chris Gammell>So I just imagine all the pomp and circumstance that the university has.

00:35:39.000 --> 00:35:43.700
<v James Adams>That's just when we go to the, we just go to the pub and enjoy that.

00:35:44.100 --> 00:35:44.380
<v James Adams>Yeah.

00:35:44.800 --> 00:35:45.200
<v James Adams>Yeah.

00:35:45.360 --> 00:35:50.600
<v James Adams>We just, as Liam says, you know, actually one of the things, our IT is quite simple.

00:35:50.600 --> 00:35:55.960
<v James Adams>It's basically a bunch of encrypted disk and servers that we keep on site and backup encrypted

00:35:55.960 --> 00:35:57.800
<v James Adams>to a known site.

00:35:57.940 --> 00:36:01.640
<v James Adams>It's actually not particularly complicated and that's, that's nice because it's easy to

00:36:01.640 --> 00:36:01.960
<v James Adams>monitor.

00:36:02.360 --> 00:36:07.140
<v James Adams>And yeah, if you, yeah, you have to be a known good Raspberry Pi employee to get an SHH

00:36:07.140 --> 00:36:07.500
<v James Adams>key.

00:36:07.660 --> 00:36:08.060
<v James Adams>Yeah.

00:36:08.280 --> 00:36:08.580
<v James Adams>That's good.

00:36:08.580 --> 00:36:09.660
<v James Adams>To work on this stuff.

00:36:10.560 --> 00:36:10.960
<v Chris Gammell>Yeah.

00:36:11.060 --> 00:36:15.300
<v Chris Gammell>I think, I think it kind of comes, comes back to that idea of like, it is software, right?

00:36:15.300 --> 00:36:19.680
<v Chris Gammell>You guys are working in software to make this very complex end result.

00:36:20.220 --> 00:36:22.800
<v Chris Gammell>And so many software things apply, right?

00:36:22.800 --> 00:36:25.500
<v Chris Gammell>Like you're saying SSH keys and all that, like it, it all makes sense.

00:36:25.560 --> 00:36:28.300
<v Chris Gammell>It is, you know, an IT organization type of thing.

00:36:28.300 --> 00:36:29.540
<v Chris Gammell>So like, yeah, it makes a lot of sense.

00:36:29.740 --> 00:36:30.580
<v Chris Gammell>It's, it's interesting.

00:36:30.740 --> 00:36:30.900
<v Chris Gammell>Yeah.

00:36:31.480 --> 00:36:31.820
<v Chris Gammell>Cool.

00:36:32.140 --> 00:36:38.540
<v Chris Gammell>So then what does it take to, so you, you buy the IP for the USB and the ethernet type.

00:36:38.540 --> 00:36:44.840
<v Chris Gammell>Like I, I, I guess I don't have a good feeling for like, so maybe, maybe this, if, if, if

00:36:44.840 --> 00:36:50.520
<v Chris Gammell>Chris was going to go start a chip tomorrow and I had gobs and bunny and maybe I hire some

00:36:50.520 --> 00:36:54.020
<v Chris Gammell>smart folks like you to come and sit in my IT center, which you guys are peering into

00:36:54.020 --> 00:36:57.920
<v Chris Gammell>my, my office right now, but let's pretend this is an actual legit operation.

00:36:58.520 --> 00:37:04.460
<v Chris Gammell>I could go out and buy IP for modern ethernet controllers, modern USB controllers.

00:37:04.560 --> 00:37:05.200
<v Chris Gammell>Is that, that the idea?

00:37:06.200 --> 00:37:06.440
<v Chris Gammell>Yeah.

00:37:07.360 --> 00:37:07.720
<v James Adams>Yep.

00:37:08.320 --> 00:37:10.760
<v James Adams>So we, I mean, there's no secret.

00:37:10.760 --> 00:37:13.960
<v James Adams>Cause I think most of us in the day, we buy a lot of stuff from Synopsys, for example.

00:37:14.300 --> 00:37:14.520
<v James Adams>Okay.

00:37:14.640 --> 00:37:14.820
<v James Adams>Sure.

00:37:14.820 --> 00:37:20.160
<v James Adams>And they supply the IT, sorry, the IP, you know, you've got to pay them money and support

00:37:20.160 --> 00:37:22.540
<v James Adams>money and they, they'd sort of deliver it to you.

00:37:22.860 --> 00:37:24.520
<v James Adams>Got to keep it secret as we've talked about.

00:37:25.060 --> 00:37:28.180
<v James Adams>And then it comes with a manual of how to use it.

00:37:28.180 --> 00:37:32.140
<v James Adams>You know, usually with a lot of pages that you have to read through and it sort of tells

00:37:32.140 --> 00:37:33.440
<v James Adams>you how you, how to hook it up.

00:37:33.520 --> 00:37:33.700
<v James Adams>Right.

00:37:34.060 --> 00:37:39.660
<v James Adams>So things like USB, you buy the controller, which is the digital bit, which is written in

00:37:39.660 --> 00:37:40.380
<v James Adams>Verilog usually.

00:37:40.380 --> 00:37:44.900
<v James Adams>And the FI, which is this, the SIRDES, the high speed SIRDES, you know, the analog bit,

00:37:44.960 --> 00:37:46.060
<v James Adams>we call it the analog bit.

00:37:46.320 --> 00:37:48.080
<v James Adams>So it's because the high speed analog.

00:37:48.760 --> 00:37:54.240
<v James Adams>Similar for the, you know, the MIPI CSI DSI cameras and USB, that's basically the model.

00:37:54.600 --> 00:37:57.980
<v James Adams>I guess ethernet's a bit different because it's, it's kind of a digital only thing.

00:37:57.980 --> 00:38:01.820
<v James Adams>And then you just buy some fast IO pads to, to run the RGMI.

00:38:02.300 --> 00:38:04.020
<v James Adams>That's kind of, yeah, that's kind of it.

00:38:04.100 --> 00:38:10.180
<v James Adams>And then what we do is, I mean, we design some little bits ourselves, but broadly,

00:38:10.400 --> 00:38:12.740
<v James Adams>for something like RP1, it's stitching it all together.

00:38:12.960 --> 00:38:16.900
<v James Adams>But there is a lot of work still in the, in the stitching, the cloths for each thing.

00:38:16.920 --> 00:38:17.140
<v James Adams>Totally.

00:38:17.620 --> 00:38:20.540
<v James Adams>The control, the GPIO, you know, all of that kind of stuff.

00:38:20.600 --> 00:38:22.780
<v James Adams>I mean, maybe Liam could speak a bit more about that.

00:38:22.940 --> 00:38:23.380
<v Liam Fraser>Yeah.

00:38:23.620 --> 00:38:28.780
<v Liam Fraser>And I actually wanted to touch on prototyping for FPG, on FPGAs as well.

00:38:29.140 --> 00:38:33.780
<v Liam Fraser>So like James said, this sort of chip version, what we call the ASIC version.

00:38:33.780 --> 00:38:38.960
<v Liam Fraser>So there'll be a controller and then the ASIC version will be like the controller and the

00:38:38.960 --> 00:38:39.300
<v Liam Fraser>PHY.

00:38:39.560 --> 00:38:46.540
<v Liam Fraser>So the analog side, but then you'll also have a sort of FPGA version, which is the controller.

00:38:46.540 --> 00:38:52.240
<v Liam Fraser>And then that will expose the sort of PHY interface on a bunch of signals that will come out of

00:38:52.240 --> 00:38:52.860
<v Liam Fraser>the FPGA.

00:38:52.860 --> 00:38:52.900
<v Liam Fraser>Okay.

00:38:53.520 --> 00:38:57.100
<v Liam Fraser>So for RP1, we used a pro FPGA system.

00:38:57.720 --> 00:39:01.000
<v Liam Fraser>So that's, it sort of has loads of daughter cards.

00:39:01.140 --> 00:39:05.240
<v Liam Fraser>So you can get a daughter card, which has an ethernet PHY on it and an ethernet jack, or

00:39:05.240 --> 00:39:08.680
<v Liam Fraser>one that has like USB three ports on it, et cetera, et cetera.

00:39:08.680 --> 00:39:16.300
<v Liam Fraser>So it allows us to test the controller with, with a real PHY and know that we've got at

00:39:16.300 --> 00:39:17.360
<v Liam Fraser>least that side, right.

00:39:17.620 --> 00:39:22.460
<v Liam Fraser>The ASIC PHY, you just have to simulate and then trust that, you know, if they've got it

00:39:22.460 --> 00:39:25.340
<v Liam Fraser>right and you'll find out if you've got it right when the chip comes back.

00:39:25.680 --> 00:39:25.780
<v Chris Gammell>Yeah.

00:39:26.440 --> 00:39:32.060
<v Liam Fraser>I think the, you can never run the FPGAs at like full speed.

00:39:32.060 --> 00:39:37.840
<v Liam Fraser>So I think the RP1 system clock is probably 200 megahertz and then on FPGA it will be,

00:39:38.120 --> 00:39:41.380
<v Liam Fraser>I don't know, 30 or something like that, but it's still enough.

00:39:41.760 --> 00:39:45.900
<v Liam Fraser>You can still run small bits fast, which is enough to run like the high speed interface

00:39:45.900 --> 00:39:48.400
<v Liam Fraser>of the USB three of the ethernet quickly.

00:39:48.920 --> 00:39:54.160
<v Liam Fraser>And also FPGA is a great for software prototyping because you can attach debugger.

00:39:54.280 --> 00:39:56.880
<v Liam Fraser>You can, you know, all, all that kind of stuff.

00:39:56.880 --> 00:40:02.880
<v Liam Fraser>Whereas simulations are very, very slow and you get, you know, sort of one millisecond

00:40:02.880 --> 00:40:07.680
<v Liam Fraser>will be one millisecond of SIM time might take, you know, five minutes, 10 minutes, depends

00:40:07.680 --> 00:40:10.240
<v Liam Fraser>on how big the design is, but it takes a long time.

00:40:10.800 --> 00:40:15.140
<v Chris Gammell>And I always remember like looking, my FPGA knowledge is very 2004.

00:40:15.540 --> 00:40:16.740
<v Chris Gammell>So please excuse that.

00:40:16.880 --> 00:40:22.480
<v Chris Gammell>But I just remember like looking at the, the logic outputs basically from a simulator and

00:40:22.480 --> 00:40:24.520
<v Chris Gammell>just being like, it's useful.

00:40:24.520 --> 00:40:28.900
<v Chris Gammell>I know it's telling me stuff that is or isn't working, but it's just like, it's not, it

00:40:28.900 --> 00:40:29.680
<v Chris Gammell>doesn't feel real.

00:40:29.800 --> 00:40:34.520
<v Chris Gammell>You know, it doesn't feel like it's a, you know, it is, it is like, it's telling me important

00:40:34.520 --> 00:40:36.300
<v Chris Gammell>things, but it doesn't feel real.

00:40:36.380 --> 00:40:38.200
<v Chris Gammell>I know it's feels over real.

00:40:38.420 --> 00:40:43.740
<v James Adams>I mean, on the digital logic side, simulators are perfectly accurate pretty much.

00:40:44.260 --> 00:40:44.400
<v James Adams>Yeah.

00:40:44.520 --> 00:40:48.320
<v James Adams>As Liam says, when you've got things like high speed fires, you can only kind of approximate

00:40:48.320 --> 00:40:55.160
<v James Adams>that, and then you take these big Xilinx FPGAs and you compile your design onto one of those,

00:40:55.160 --> 00:41:00.220
<v James Adams>run it slower, but at least then you have real digits, real IOs twiddling, talking to

00:41:00.220 --> 00:41:02.180
<v James Adams>chips that are kind of fire emulation.

00:41:02.180 --> 00:41:04.300
<v James Adams>And then you can actually plug real USB threes in.

00:41:04.520 --> 00:41:10.840
<v James Adams>So we, you know, we had RP one on FPGA plugged into a, an Intel box over PCI express and

00:41:10.840 --> 00:41:15.380
<v James Adams>had, you know, USB in a camera and a display kind of running in prototype.

00:41:15.380 --> 00:41:17.720
<v James Adams>This is before we'd actually taped the chip out.

00:41:17.760 --> 00:41:17.940
<v James Adams>Right.

00:41:18.000 --> 00:41:21.420
<v James Adams>But it's basically running exactly, pretty much exactly as the chip would run.

00:41:21.700 --> 00:41:22.920
<v Chris Gammell>Do you guys have any photos of that setup?

00:41:23.020 --> 00:41:27.740
<v Chris Gammell>I love those like old, like the, did you ever see the original like iPhone, like all on one

00:41:27.740 --> 00:41:28.960
<v Chris Gammell>huge board thing?

00:41:29.540 --> 00:41:31.120
<v Chris Gammell>I haven't seen that, but I can imagine.

00:41:31.540 --> 00:41:31.620
<v Chris Gammell>Yeah.

00:41:31.780 --> 00:41:35.780
<v Chris Gammell>It was just all dev boards, basically all, it was like a, it was like a full panel of

00:41:35.780 --> 00:41:38.140
<v Chris Gammell>a PCB and they just kind of shoved everything together.

00:41:38.320 --> 00:41:40.160
<v Chris Gammell>Like I love those old prototype photos.

00:41:40.200 --> 00:41:40.980
<v Chris Gammell>I think they're so fun.

00:41:41.240 --> 00:41:42.260
<v James Adams>We must have some somewhere.

00:41:42.820 --> 00:41:43.060
<v Liam Fraser>Yeah.

00:41:43.060 --> 00:41:44.160
<v Liam Fraser>I'm sure I've got some.

00:41:44.600 --> 00:41:45.300
<v Chris Gammell>Someone's phone probably.

00:41:46.020 --> 00:41:46.260
<v Chris Gammell>Yeah.

00:41:47.260 --> 00:41:51.420
<v James Adams>Trouble is this stuff takes a long time to, you know, how many, really a long time ago

00:41:51.420 --> 00:41:52.060
<v James Adams>we took those pictures.

00:41:52.220 --> 00:41:53.480
<v James Adams>I'm sure I've got some somewhere as well.

00:41:53.760 --> 00:41:53.940
<v Chris Gammell>Yeah.

00:41:54.080 --> 00:41:54.260
<v Chris Gammell>Yeah.

00:41:54.880 --> 00:41:56.040
<v Chris Gammell>Google photos is how I do it.

00:41:56.080 --> 00:41:58.120
<v Chris Gammell>I just search for electronics and then Google photos.

00:41:58.180 --> 00:41:59.340
<v Chris Gammell>Like this is probably electronics.

00:41:59.420 --> 00:42:00.440
<v Chris Gammell>And I'm like, oh yeah, 2015.

00:42:01.720 --> 00:42:02.780
<v Chris Gammell>Sorry if I find my old stuff.

00:42:02.780 --> 00:42:04.460
<v Chris Gammell>I really like that.

00:42:04.460 --> 00:42:07.780
<v Chris Gammell>You guys have the, so like I'm looking, I'm looking at the data sheet right now and it

00:42:07.780 --> 00:42:09.340
<v Chris Gammell>says like each SSI.

00:42:09.460 --> 00:42:10.180
<v Chris Gammell>So this is for spy.

00:42:10.300 --> 00:42:16.620
<v Chris Gammell>Each SSI controller is based on a configuration of the synopsis DW APB SSI IP version 4.0.2.

00:42:16.620 --> 00:42:18.040
<v Chris Gammell>A, right?

00:42:18.100 --> 00:42:19.280
<v Chris Gammell>Like that's, that's great.

00:42:19.340 --> 00:42:22.720
<v Chris Gammell>That's like almost like traceable back to like the actual IP.

00:42:23.260 --> 00:42:26.740
<v Chris Gammell>Not that it really matters, but like, is there any concern about like, well, someone could

00:42:26.740 --> 00:42:27.780
<v Chris Gammell>just replicate this whole design?

00:42:27.840 --> 00:42:30.160
<v Chris Gammell>Is that, is that like of concern or not really?

00:42:30.740 --> 00:42:31.540
<v James Adams>It's a good question.

00:42:31.660 --> 00:42:37.580
<v James Adams>I mean, I think it would be hard to really replicate it exactly, but you know, you could

00:42:37.580 --> 00:42:41.240
<v James Adams>with a, with a, with a design team and enough money and enough time.

00:42:41.360 --> 00:42:45.000
<v James Adams>But I guess by the time you've done that, we'd probably be onto gen two or something.

00:42:45.000 --> 00:42:45.360
<v James Adams>Yeah.

00:42:45.980 --> 00:42:49.700
<v James Adams>I mean, you know, it's kind of like, you know, you could go build an Intel processor if

00:42:49.700 --> 00:42:51.220
<v James Adams>you really wanted to with enough time and money.

00:42:52.100 --> 00:42:53.160
<v James Adams>How much is it worth?

00:42:53.480 --> 00:42:57.300
<v Chris Gammell>I think it also like, it kind of sidesteps the whole, like the value prop of something

00:42:57.300 --> 00:43:01.520
<v Chris Gammell>like the, you know, obviously your company and the RP 2040 is this finished thing that

00:43:01.520 --> 00:43:02.500
<v Chris Gammell>gets delivered to me.

00:43:02.740 --> 00:43:05.560
<v Chris Gammell>The user is like, is that it's done and it's tested.

00:43:05.660 --> 00:43:09.100
<v Chris Gammell>Like all of the other stuff that goes into it is the, like the software and, you know,

00:43:09.120 --> 00:43:13.100
<v Chris Gammell>the putting it all together, the, the optimizations that you're doing around this stuff.

00:43:13.440 --> 00:43:17.240
<v Chris Gammell>And yeah, there, you know, there is legitimate like competition out there.

00:43:17.240 --> 00:43:18.660
<v Chris Gammell>So I'm just getting curious about that.

00:43:18.860 --> 00:43:23.060
<v Chris Gammell>I think you guys probably have more scale than maybe some upstart might have, despite you

00:43:23.060 --> 00:43:25.500
<v Chris Gammell>also being upstarts, you know, it's all relative scale, right?

00:43:25.500 --> 00:43:27.160
<v Chris Gammell>You versus NVIDIA.

00:43:27.300 --> 00:43:28.700
<v Chris Gammell>It's like, oh, they're like, oh, okay.

00:43:29.200 --> 00:43:30.700
<v Chris Gammell>Ant, ant of a company, right?

00:43:30.800 --> 00:43:34.040
<v Chris Gammell>But, but like at one point NVIDIA was an ant of a company too, right?

00:43:34.080 --> 00:43:35.220
<v Chris Gammell>So like, that's the.

00:43:35.820 --> 00:43:36.220
<v James Adams>That's true.

00:43:36.400 --> 00:43:39.900
<v James Adams>I mean, you know, we do like to make this stuff open, right?

00:43:39.920 --> 00:43:44.480
<v James Adams>Because, you know, it's just where we can, we like to, right?

00:43:44.540 --> 00:43:49.640
<v James Adams>It's, you know, we can't open up more than, for example, in the controllers, what the

00:43:49.640 --> 00:43:52.780
<v James Adams>synopsis will let us, for example, and we can't open certain things from the Broadcom

00:43:52.780 --> 00:43:57.860
<v James Adams>stuff because it's Broadcom IP, but where we have our own IP, most of the time we really

00:43:57.860 --> 00:43:59.100
<v James Adams>like to, to make it open.

00:43:59.980 --> 00:44:00.080
<v James Adams>Yeah.

00:44:00.120 --> 00:44:05.980
<v James Adams>People, people do kind of like, I think unreasonably say, oh, Raspberry Pi, they say they're open,

00:44:06.060 --> 00:44:09.580
<v James Adams>but they're not open because yeah, the Broadcom thing, there's still, you know, there's still

00:44:09.580 --> 00:44:11.560
<v James Adams>stuff there that we can't see and et cetera, et cetera.

00:44:11.560 --> 00:44:15.640
<v James Adams>But over the years, we have really tried hard to make the platform more open.

00:44:15.780 --> 00:44:19.540
<v James Adams>If you look at the firmware now, the, the firmware blobs really reduced down to just kind

00:44:19.540 --> 00:44:23.620
<v James Adams>of some clocks and power stuff, you know, all the GPU stuff is, is open.

00:44:23.940 --> 00:44:27.320
<v James Adams>You know, we're doing our best to push, push openness where we really can.

00:44:27.580 --> 00:44:30.280
<v James Adams>And, you know, more stuff is upstream in the kernel now.

00:44:30.620 --> 00:44:34.240
<v Chris Gammell>I'm sure in the 13 years of the amp hour, we have also complained about that at some point.

00:44:35.460 --> 00:44:39.560
<v Chris Gammell>And, but like at a certain point, you know, any layer of abstraction eventually runs into

00:44:39.560 --> 00:44:40.860
<v Chris Gammell>some kind of opaqueness, right?

00:44:41.160 --> 00:44:41.800
<v Chris Gammell>I think so.

00:44:41.860 --> 00:44:42.040
<v Chris Gammell>Yeah.

00:44:42.120 --> 00:44:42.860
<v Chris Gammell>Through the stack, right?

00:44:42.880 --> 00:44:46.140
<v Chris Gammell>It's just like, I don't know how Silicon, you know, I don't ultimately know what the process

00:44:46.140 --> 00:44:46.500
<v Chris Gammell>node is.

00:44:46.500 --> 00:44:49.820
<v Chris Gammell>Like I'm not going to get access to that, even if I have all of the transistor layout,

00:44:49.900 --> 00:44:50.060
<v Chris Gammell>whatever.

00:44:50.180 --> 00:44:54.380
<v Chris Gammell>And like the truth is at a certain point, it's just like performative, some of the openness,

00:44:54.620 --> 00:44:54.780
<v Chris Gammell>right?

00:44:54.780 --> 00:44:58.640
<v Chris Gammell>So doing the best you can and you're making good over, over, over strides overall.

00:44:58.640 --> 00:45:03.920
<v James Adams>I think where it is actually valuable to people, then we'll always try and do it right.

00:45:04.080 --> 00:45:08.400
<v James Adams>I mean, you know, making stuff on RP1 open, what's the value in that?

00:45:08.520 --> 00:45:12.460
<v James Adams>Well, I mean, you know, we provide drivers for it and stuff, but at least you can now go and

00:45:12.460 --> 00:45:13.480
<v James Adams>have a look inside this chip.

00:45:13.480 --> 00:45:16.980
<v James Adams>And if you've got some weird, weird thing going on, maybe you can go and then hack on

00:45:16.980 --> 00:45:17.560
<v James Adams>it and have a look.

00:45:17.700 --> 00:45:21.220
<v James Adams>And if you want to do something funky with it, yep, off you go.

00:45:21.460 --> 00:45:25.060
<v James Adams>You've got some data there and it'll be interesting to see what people do do with it.

00:45:25.800 --> 00:45:25.900
<v James Adams>Yeah.

00:45:26.980 --> 00:45:27.300
<v James Adams>Competition.

00:45:27.580 --> 00:45:28.460
<v James Adams>There's always competition.

00:45:28.880 --> 00:45:31.980
<v James Adams>I think we just like to, yeah, it's a balance, right?

00:45:32.140 --> 00:45:34.120
<v James Adams>We don't ever really withhold stuff.

00:45:35.060 --> 00:45:38.840
<v James Adams>Philosophically, we don't really want to withhold stuff for that kind of reason.

00:45:39.260 --> 00:45:40.040
<v James Adams>If you see what I mean?

00:45:40.040 --> 00:45:40.280
<v Chris Gammell>Right.

00:45:40.660 --> 00:45:43.620
<v Chris Gammell>Sometimes the IP agreement, so it's very big to know contract, I'm sure.

00:45:43.820 --> 00:45:45.460
<v James Adams>Well, yeah, it's mostly just IP agreements.

00:45:45.640 --> 00:45:49.280
<v James Adams>We can't release this stuff because it's not our stuff to release.

00:45:49.760 --> 00:45:52.440
<v James Adams>But anyway, hopefully, yeah, RP1.

00:45:52.840 --> 00:45:54.300
<v James Adams>And the documentation is not quite finished.

00:45:54.540 --> 00:45:58.640
<v James Adams>I don't know if you've noticed, but yeah, it says draft all over it and we're still writing

00:45:58.640 --> 00:45:58.900
<v James Adams>it.

00:45:59.160 --> 00:46:02.200
<v James Adams>But it has got a lot of good information in there now, but we're sort of still fettling

00:46:02.200 --> 00:46:02.420
<v James Adams>it.

00:46:02.420 --> 00:46:06.940
<v James Adams>Just the whole Raspberry Pi release thing, there's just such a huge amount of work, peripherals,

00:46:07.080 --> 00:46:08.360
<v James Adams>the documentation.

00:46:09.460 --> 00:46:14.800
<v James Adams>You know, so we're, you know, the usual thing where we're kind of like peddling really fast

00:46:14.800 --> 00:46:16.700
<v Chris Gammell>still to get it all going.

00:46:16.700 --> 00:46:20.280
<v Chris Gammell>I remember from the last time you guys were here too, just like the scale and the, I

00:46:20.280 --> 00:46:23.460
<v Chris Gammell>remember you talking about the cost of things and just like some of the design decisions

00:46:23.460 --> 00:46:25.240
<v Chris Gammell>that came around, like the Pico, I think it was.

00:46:25.300 --> 00:46:29.660
<v Chris Gammell>And just like, do we put a, there was some like button that you didn't have a certain button

00:46:29.660 --> 00:46:33.740
<v Chris Gammell>on there or you didn't have a header on there or something that was like a cost savings.

00:46:34.520 --> 00:46:36.220
<v Liam Fraser>The run pin probably.

00:46:37.220 --> 00:46:37.540
<v Liam Fraser>What was it?

00:46:37.580 --> 00:46:38.840
<v Liam Fraser>The run pin probably.

00:46:39.100 --> 00:46:39.580
<v Chris Gammell>The run pin.

00:46:39.940 --> 00:46:40.300
<v Chris Gammell>Yeah.

00:46:40.300 --> 00:46:41.460
<v Chris Gammell>Yeah.

00:46:41.460 --> 00:46:41.500
<v Chris Gammell>Yeah.

00:46:41.580 --> 00:46:43.560
<v James Adams>People put on a reset button.

00:46:43.660 --> 00:46:44.700
<v James Adams>I think that's so unreasonable.

00:46:47.280 --> 00:46:48.340
<v Chris Gammell>Get a wire, folks.

00:46:49.800 --> 00:46:54.080
<v James Adams>I think what you need to realize when you build these products, there's sort of a certain low,

00:46:54.260 --> 00:46:58.540
<v James Adams>there's a low level of fixed costs that you can't really, you know, it's not a linear scale,

00:46:58.620 --> 00:46:58.840
<v James Adams>right?

00:46:59.140 --> 00:47:01.840
<v James Adams>It starts from a certain fixed costs that you can never get away from.

00:47:02.040 --> 00:47:07.180
<v James Adams>So when you build these really low cost products, if you want them to be still good quality,

00:47:07.180 --> 00:47:12.420
<v James Adams>like they, Picos have a hundred percent test, you know, they're built at Sony, they're built

00:47:12.420 --> 00:47:17.300
<v James Adams>to a very high standard, that stuff, even though the materials going into it and the

00:47:17.300 --> 00:47:20.820
<v James Adams>number of components is kind of, you know, it's not a lot, there's still that kind of

00:47:20.820 --> 00:47:21.360
<v James Adams>fixed cost.

00:47:21.480 --> 00:47:25.520
<v James Adams>And when you're trying to build a $4 product, you know, you really, you really have to engineer

00:47:25.520 --> 00:47:27.560
<v James Adams>out as much cost as you possibly can.

00:47:27.880 --> 00:47:30.560
<v James Adams>And yeah, people complain about like, why didn't the Pico have USB-C?

00:47:30.660 --> 00:47:31.460
<v James Adams>That's a common one.

00:47:31.460 --> 00:47:35.900
<v James Adams>Well, actually, USB-C connectors are twice the price of the micro USB.

00:47:36.660 --> 00:47:38.780
<v James Adams>And, you know, that makes a difference, actually.

00:47:39.100 --> 00:47:41.060
<v James Adams>And also, you know, we buy a lot of micro USB connectors.

00:47:41.540 --> 00:47:42.500
<v James Adams>We had the supply there.

00:47:42.500 --> 00:47:44.320
<v Chris Gammell>I think because you guys probably can't say it, I can say it.

00:47:44.560 --> 00:47:46.660
<v Chris Gammell>The right answer is, shut up, it's $4, right?

00:47:47.580 --> 00:47:48.580
<v James Adams>Pretty much, right?

00:47:48.880 --> 00:47:49.060
<v James Adams>Yeah.

00:47:49.640 --> 00:47:53.280
<v Chris Gammell>I mean, like, people will always ask for more, no matter what price you are.

00:47:53.320 --> 00:47:56.200
<v Chris Gammell>You could offer it to them at 99 cents, they say, why is it at 98 cents, right?

00:47:56.260 --> 00:47:56.700
<v Chris Gammell>I mean, like...

00:47:56.700 --> 00:47:57.780
<v James Adams>I get it, right?

00:47:57.780 --> 00:47:58.860
<v James Adams>I think we all get it.

00:47:58.920 --> 00:48:03.180
<v James Adams>It's like, as an engineer, especially, you're like, oh, God, if only they'd just done that,

00:48:03.200 --> 00:48:04.220
<v James Adams>it would be just so awesome.

00:48:04.300 --> 00:48:06.280
<v James Adams>And it's like, yeah, but we couldn't.

00:48:06.380 --> 00:48:08.780
<v James Adams>Like, we did think about it, but kind of just didn't.

00:48:09.060 --> 00:48:10.460
<v James Adams>You know, we've always got to make these trade-offs.

00:48:10.480 --> 00:48:11.940
<v James Adams>And that is engineering in a nutshell, right?

00:48:11.940 --> 00:48:14.380
<v James Adams>You're always making costs, you know, time.

00:48:14.380 --> 00:48:16.340
<v James Adams>All this stuff is all a trade-off.

00:48:16.980 --> 00:48:21.140
<v James Adams>And so, yeah, hopefully, you know, maybe in the future, USB-C gets cheaper.

00:48:21.140 --> 00:48:26.600
<v Chris Gammell>So instead of optimizing that next thing, we went and just worked on the new thing to

00:48:26.600 --> 00:48:28.020
<v Chris Gammell>also optimize that for you, folks.

00:48:28.220 --> 00:48:29.060
<v Chris Gammell>Exactly, yeah.

00:48:30.620 --> 00:48:33.860
<v James Adams>But so, you know, we kind of take some of the feedback.

00:48:34.020 --> 00:48:34.920
<v James Adams>You know, we take the feedback, right?

00:48:34.920 --> 00:48:35.580
<v James Adams>We listen to people.

00:48:35.580 --> 00:48:39.500
<v James Adams>But we also, you know, we have to build this stuff and make it work.

00:48:40.120 --> 00:48:46.580
<v Chris Gammell>I know we're much into the recording here now, but like the RP1 that lives on the Raspberry

00:48:46.580 --> 00:48:48.720
<v Chris Gammell>Pi 5, what is it for?

00:48:48.720 --> 00:48:50.840
<v Chris Gammell>I guess we probably should have led with that.

00:48:51.240 --> 00:48:55.060
<v Chris Gammell>It's a companionship to the Broadcom part that's on there that does all the Linux lifting,

00:48:55.180 --> 00:48:55.300
<v Chris Gammell>right?

00:48:55.360 --> 00:48:58.380
<v Chris Gammell>But what is the RP1 kind of generally meant for?

00:48:58.940 --> 00:49:01.900
<v James Adams>Ooh, who wants to have a go at that?

00:49:02.920 --> 00:49:03.740
<v James Adams>What is it for?

00:49:04.000 --> 00:49:06.280
<v James Adams>I think we kind of already, we sort of spoke about it.

00:49:06.360 --> 00:49:08.680
<v James Adams>It's too, what's the best way to answer this?

00:49:09.360 --> 00:49:10.500
<v Chris Gammell>You had said, how about this?

00:49:10.520 --> 00:49:11.660
<v Chris Gammell>You said the word Southbridge.

00:49:11.920 --> 00:49:12.180
<v James Adams>Yeah.

00:49:12.440 --> 00:49:15.500
<v Chris Gammell>And that's like a term from like the Intel stuff, I think.

00:49:15.660 --> 00:49:15.940
<v James Adams>Yeah.

00:49:16.060 --> 00:49:17.480
<v Chris Gammell>I never really understood what that is.

00:49:17.480 --> 00:49:18.600
<v Chris Gammell>So maybe what is a Southbridge?

00:49:18.720 --> 00:49:20.040
<v James Adams>So that's kind of the IO hub, right?

00:49:20.260 --> 00:49:25.540
<v James Adams>So in past times, your IO hub would live off your processor, right?

00:49:25.660 --> 00:49:27.240
<v James Adams>And that was called a Southbridge.

00:49:27.760 --> 00:49:33.300
<v James Adams>And you had Northbridge, which tended to be the sort of cache and memory system, which that

00:49:33.300 --> 00:49:37.440
<v James Adams>was the first thing that got kind of subsumed into these all-in-one these days, Intel and

00:49:37.440 --> 00:49:38.140
<v James Adams>AMD processors.

00:49:38.140 --> 00:49:40.560
<v James Adams>And so Southbridge was the kind of IO hub.

00:49:40.560 --> 00:49:42.940
<v James Adams>Northbridge was the memory controller and cache.

00:49:43.400 --> 00:49:44.300
<v James Adams>So yeah, it's a Southbridge.

00:49:44.980 --> 00:49:48.820
<v James Adams>It's the natural evolution of the kind of architecture of the Pi, I think, really.

00:49:49.040 --> 00:49:53.280
<v James Adams>It's taking the Pi-ness and putting it into a chip that is kind of stable.

00:49:53.420 --> 00:49:54.480
<v James Adams>It's well-engineered.

00:49:54.700 --> 00:49:56.080
<v James Adams>It's just the stuff we need.

00:49:56.080 --> 00:49:58.620
<v James Adams>It takes the load off designing the main processor.

00:49:59.180 --> 00:50:07.380
<v James Adams>And as you shift nodes down, as you go down from, you know, 2711 on Pi-4 is a 28 nanometer

00:50:07.380 --> 00:50:07.700
<v James Adams>chip.

00:50:08.460 --> 00:50:11.820
<v James Adams>2712 on Pi-5 is 16 nanometer.

00:50:12.340 --> 00:50:16.840
<v James Adams>The analog side, so there's all the PHYs, the USB 3 PHYs, the PCI Express, all this kind

00:50:16.840 --> 00:50:20.300
<v James Adams>of stuff doesn't scale very well when you go down the small nodes, right?

00:50:20.540 --> 00:50:22.860
<v James Adams>And as you go down the nodes, the silicon gets more expensive.

00:50:22.860 --> 00:50:24.460
<v James Adams>So you've got kind of two wins there.

00:50:24.560 --> 00:50:29.420
<v James Adams>One is you've got a stable thing that you've farmed out all this development work and you

00:50:29.420 --> 00:50:30.140
<v James Adams>know it just works.

00:50:30.600 --> 00:50:36.420
<v James Adams>And then you've also got the fact that you're not putting a USB PHY on this chip that's still

00:50:36.420 --> 00:50:40.220
<v James Adams>the same size as it was on 28, but now your silicon's twice the price, for example.

00:50:40.840 --> 00:50:42.460
<v James Adams>So it has a kind of a double benefit.

00:50:43.320 --> 00:50:46.240
<v James Adams>And I guess people are re-realizing this.

00:50:46.320 --> 00:50:50.000
<v James Adams>I'm not sure if it's complete revelation, but basically this is what chiplets are, right?

00:50:50.080 --> 00:50:51.900
<v James Adams>So we kind of got a chiplet here.

00:50:51.900 --> 00:50:53.240
<v James Adams>This is basically it, right?

00:50:53.360 --> 00:50:55.260
<v James Adams>So we've done a chiplet architecture.

00:50:55.600 --> 00:50:57.520
<v James Adams>We've used PCI Express, which is a standard.

00:50:57.800 --> 00:51:02.520
<v James Adams>I mean, I think chiplets now, they're coming up with lower, you know, different standards,

00:51:02.680 --> 00:51:06.740
<v James Adams>similar kind of serial standards, but, you know, lower power and designed to put these

00:51:06.740 --> 00:51:08.440
<v James Adams>things next to each other on a chip substrate.

00:51:08.640 --> 00:51:12.420
<v James Adams>But it's basically a chiplet architecture and done for the kind of the similar reasons,

00:51:12.500 --> 00:51:12.700
<v James Adams>right?

00:51:12.700 --> 00:51:19.780
<v Chris Gammell>I also think about the amount that I've abused Raspberry Pi pins in the past, you know, 5-volt

00:51:19.780 --> 00:51:23.560
<v Chris Gammell>tolerance, 3-volt tolerance, and just like, you know, you're moving down to 16 nanometers.

00:51:23.720 --> 00:51:25.720
<v Chris Gammell>Like, what is the core voltage on a 16 nanometer process?

00:51:25.900 --> 00:51:27.260
<v Chris Gammell>Like 0.9 volts?

00:51:27.920 --> 00:51:29.880
<v James Adams>Yeah, it's like 0.9, 0.95.

00:51:30.240 --> 00:51:32.200
<v James Adams>I think 1-volt absolute max.

00:51:32.820 --> 00:51:32.980
<v James Adams>Yeah.

00:51:32.980 --> 00:51:35.780
<v James Adams>So, yeah, the transistors, yeah, the transistors get smaller.

00:51:36.460 --> 00:51:36.940
<v James Adams>Exactly.

00:51:37.300 --> 00:51:39.340
<v James Adams>They're less stable.

00:51:39.620 --> 00:51:45.140
<v James Adams>Actually, our GPIO pins on RP1 are hardened for ESD as well.

00:51:45.180 --> 00:51:50.400
<v James Adams>So we actually did some custom engineering on those to make them more hardened to ESD strikes.

00:51:50.980 --> 00:51:51.080
<v James Adams>Okay.

00:51:51.180 --> 00:51:55.020
<v James Adams>So, again, it's just another improvement that can kind of help the thing be more robust.

00:51:55.580 --> 00:51:55.820
<v Chris Gammell>Right.

00:51:55.820 --> 00:52:00.660
<v James Adams>So, yeah, we can add all that kind of good stuff, and we can stop having to put it on

00:52:00.660 --> 00:52:01.500
<v James Adams>the expensive silicon.

00:52:02.480 --> 00:52:02.800
<v James Adams>Yeah.

00:52:03.480 --> 00:52:04.140
<v James Adams>That's interesting.

00:52:04.500 --> 00:52:05.440
<v Chris Gammell>You know, I'll just slip this in.

00:52:05.480 --> 00:52:07.300
<v Chris Gammell>Are there any Easter eggs we should know about in the RP1?

00:52:08.740 --> 00:52:09.100
<v Liam Fraser>Liam?

00:52:10.220 --> 00:52:12.760
<v Liam Fraser>Well, I think I'm sure it's in the date sheet somewhere.

00:52:13.120 --> 00:52:16.980
<v Liam Fraser>It's definitely on Twitter and on the blog that there's PIOs in there.

00:52:17.440 --> 00:52:17.640
<v Liam Fraser>Ah.

00:52:18.160 --> 00:52:20.880
<v Liam Fraser>But there are, oh, sorry, there's only one PIO.

00:52:21.240 --> 00:52:22.540
<v Liam Fraser>I mean, yeah, you got to do it, right?

00:52:22.540 --> 00:52:26.240
<v Liam Fraser>But it's not accessible over PCI Express.

00:52:27.160 --> 00:52:35.500
<v Liam Fraser>I mean, I think that's sort of key architectural difference between RP1 and RP2040 is that RP2040

00:52:35.500 --> 00:52:39.720
<v Liam Fraser>has got a really small bus fabric that's designed for, like, low latency.

00:52:40.000 --> 00:52:45.600
<v Liam Fraser>So if you want to go and read a peripheral, that will take one or two clock cycles on RP2040.

00:52:45.600 --> 00:52:51.720
<v Liam Fraser>Whereas because you've got PCI Express and you've got sort of big hungry busmasters like USB

00:52:51.720 --> 00:52:59.580
<v Liam Fraser>and, like, you know, cameras and displays, it's a higher latency but higher throughput fabric.

00:53:00.240 --> 00:53:05.900
<v Liam Fraser>So you might go and want to talk to a PIO or you might want to talk to a UART, let's say.

00:53:06.120 --> 00:53:09.980
<v Liam Fraser>That might take you 10 clock cycles, which is, you know, quite...

00:53:09.980 --> 00:53:14.700
<v Liam Fraser>It's perfectly fine when you're on Linux because it's not that long.

00:53:15.380 --> 00:53:15.540
<v Liam Fraser>Yeah.

00:53:15.720 --> 00:53:20.640
<v Liam Fraser>So RP1, it's not really a microcontroller.

00:53:20.780 --> 00:53:26.080
<v Liam Fraser>It does have some small ARM processes in it to manage it, but it's quite different.

00:53:26.760 --> 00:53:33.520
<v Liam Fraser>So to expose PIO, we effectively need a way for Linux to talk to the small ARM cores

00:53:33.520 --> 00:53:38.760
<v Liam Fraser>in the, like, little microcontrollery management bit to then talk to PIO on its behalf.

00:53:38.880 --> 00:53:42.300
<v Liam Fraser>So it's not quite as simple as just saying there's a PIO at this address.

00:53:42.980 --> 00:53:43.100
<v Liam Fraser>Yeah.

00:53:43.300 --> 00:53:43.580
<v Liam Fraser>Yeah.

00:53:43.880 --> 00:53:44.260
<v James Adams>Makes sense.

00:53:44.960 --> 00:53:45.240
<v James Adams>Yeah.

00:53:45.340 --> 00:53:48.500
<v James Adams>And we are working on how to do that nicely so people can use it.

00:53:49.260 --> 00:53:49.660
<v Liam Fraser>Yeah.

00:53:49.900 --> 00:53:51.240
<v James Adams>So it's coming.

00:53:51.500 --> 00:53:52.520
<v James Adams>I think we're just trying to...

00:53:53.040 --> 00:53:55.540
<v James Adams>Yeah, we just want to make a nice way so that people can use it.

00:53:56.100 --> 00:53:56.240
<v James Adams>Yeah.

00:53:56.600 --> 00:54:00.840
<v James Adams>If we just expose how to hack at it, then we'll probably get 10 different ways

00:54:00.840 --> 00:54:02.420
<v James Adams>and none of them work very well.

00:54:02.420 --> 00:54:06.860
<v James Adams>But try and come up with a nice solution for people to load code into it and run stuff.

00:54:07.400 --> 00:54:10.860
<v James Adams>I mean, actually, it's a good point that Liam brings up is because you're remoting

00:54:10.860 --> 00:54:13.740
<v James Adams>the stuff over PCI Express, even just...

00:54:13.740 --> 00:54:17.160
<v James Adams>Because a lot of people do use Raspberry Pis and bit bang the GPIO from the processor,

00:54:17.300 --> 00:54:21.260
<v James Adams>which is still a relatively high latency thing, even on, like, a Pi 4.

00:54:21.380 --> 00:54:23.060
<v James Adams>But it's much worse on a Pi 5, right?

00:54:23.060 --> 00:54:24.500
<v James Adams>Because you've got this PCI Express bus.

00:54:24.580 --> 00:54:27.260
<v James Adams>You've got to send your messages over before it can go and twiddle the pin.

00:54:27.940 --> 00:54:31.340
<v James Adams>So there will be some cases where people are kind of doing this

00:54:31.340 --> 00:54:33.980
<v James Adams>and it will be worse on Pi 5.

00:54:34.980 --> 00:54:37.780
<v James Adams>Just bit banging an IO and expecting it to...

00:54:37.780 --> 00:54:39.780
<v James Adams>It's sort of for something like WS2810...

00:54:40.660 --> 00:54:40.900
<v James Adams>What are they?

00:54:40.980 --> 00:54:42.160
<v James Adams>2810B or what are they called?

00:54:42.280 --> 00:54:42.780
<v James Adams>The little...

00:54:42.780 --> 00:54:44.220
<v James Adams>The LEDs.

00:54:44.740 --> 00:54:45.000
<v James Adams>That's right.

00:54:45.060 --> 00:54:45.160
<v James Adams>Yeah.

00:54:45.200 --> 00:54:45.400
<v James Adams>Yeah.

00:54:45.400 --> 00:54:46.540
<v Chris Gammell>Those have, like, crazy...

00:54:46.540 --> 00:54:46.780
<v Chris Gammell>It's like...

00:54:47.360 --> 00:54:48.500
<v Chris Gammell>It's like the timing is...

00:54:48.500 --> 00:54:48.920
<v Chris Gammell>Yeah, specific time windows.

00:54:49.460 --> 00:54:49.860
<v James Adams>Exactly.

00:54:50.320 --> 00:54:50.500
<v James Adams>Yeah.

00:54:50.500 --> 00:54:55.860
<v James Adams>That kind of just about works on older Pis because, you know, there's...

00:54:55.860 --> 00:54:57.180
<v James Adams>It's generally fast enough.

00:54:57.320 --> 00:54:58.880
<v James Adams>I mean, I still wouldn't recommend it, but people do.

00:54:59.280 --> 00:55:04.020
<v James Adams>But on a Pi 5, that's almost certainly not going to work because your latency is higher

00:55:04.020 --> 00:55:05.620
<v James Adams>to the GPIO and it's also more variable.

00:55:05.780 --> 00:55:10.000
<v James Adams>Because, like, if USB bus traffic or a camera bus traffic gets in there, it's going to...

00:55:10.000 --> 00:55:11.420
<v James Adams>You know, it's much more variable.

00:55:11.940 --> 00:55:12.300
<v James Adams>Right.

00:55:12.400 --> 00:55:13.440
<v James Adams>So, yeah, that is a thing.

00:55:13.520 --> 00:55:16.000
<v James Adams>And, again, the PIO can solve some of these.

00:55:16.000 --> 00:55:18.400
<v James Adams>So, we are going to try and make it usable.

00:55:19.180 --> 00:55:20.260
<v James Adams>But it's coming.

00:55:20.460 --> 00:55:22.080
<v James Adams>It's just not quite revealed yet.

00:55:22.540 --> 00:55:22.740
<v Chris Gammell>Okay.

00:55:22.940 --> 00:55:23.360
<v Chris Gammell>That's cool.

00:55:23.460 --> 00:55:25.240
<v Chris Gammell>I mean, honestly, on...

00:55:25.240 --> 00:55:28.180
<v Chris Gammell>You know, when I'm using Raspberry Pis, I'm using other people's software.

00:55:28.540 --> 00:55:29.160
<v Chris Gammell>I'm using...

00:55:29.160 --> 00:55:30.060
<v Chris Gammell>You know, I'm just...

00:55:30.060 --> 00:55:31.560
<v Chris Gammell>I'm using off-the-shelf type of stuff.

00:55:31.720 --> 00:55:36.780
<v Chris Gammell>Again, as a hardware person, if I want to go in two little bits, I'm going down embedded

00:55:36.780 --> 00:55:38.940
<v Chris Gammell>because I mostly know how to do it.

00:55:39.080 --> 00:55:41.500
<v Chris Gammell>And, like, that's just what I know what to do.

00:55:41.580 --> 00:55:45.140
<v Chris Gammell>I feel like one of the challenges is that because Raspberry Pi is so...

00:55:46.000 --> 00:55:47.400
<v Chris Gammell>The Raspberry Pi board, right?

00:55:47.420 --> 00:55:55.020
<v Chris Gammell>The 4, 5, 6, whatever, is so broadly accessible and powerful and, you know, out there.

00:55:56.020 --> 00:55:59.100
<v Chris Gammell>People use it for all kinds of things because they don't know any different.

00:55:59.240 --> 00:56:03.000
<v Chris Gammell>I mean, like, I remember seeing the discussion on the RP2040 when it first came out and people

00:56:03.000 --> 00:56:04.800
<v Chris Gammell>were like, oh, this is never going to run Linux.

00:56:04.900 --> 00:56:06.440
<v Chris Gammell>It's like, no shit.

00:56:08.820 --> 00:56:10.960
<v Chris Gammell>But, you know, that's the context a lot of people come in with.

00:56:11.080 --> 00:56:11.760
<v Chris Gammell>And I was like, that's fine.

00:56:11.760 --> 00:56:18.080
<v Chris Gammell>But I feel like that's also, like, the best case scenario would be when you write to the

00:56:18.080 --> 00:56:19.880
<v Chris Gammell>slash sys, whatever, you know, GPIO.

00:56:20.620 --> 00:56:24.780
<v Chris Gammell>It sends a little message back on the terminal that says, have you heard about the RP2040?

00:56:25.060 --> 00:56:27.620
<v Chris Gammell>Or try MicroPython today.

00:56:27.800 --> 00:56:30.620
<v Chris Gammell>You know, like, C is your friend, you know?

00:56:31.060 --> 00:56:32.100
<v Chris Gammell>So, yeah.

00:56:33.240 --> 00:56:33.460
<v Liam Fraser>Yeah.

00:56:33.580 --> 00:56:36.000
<v Chris Gammell>The opportunity to educate.

00:56:36.000 --> 00:56:36.280
<v Chris Gammell>Yeah.

00:56:36.940 --> 00:56:43.760
<v Liam Fraser>And we have, you know, where people are using standard Linux interfaces for, you know, for

00:56:43.760 --> 00:56:47.480
<v Liam Fraser>GPIO, for cameras, for displays, all that kind of stuff.

00:56:47.880 --> 00:56:53.440
<v Liam Fraser>We have implemented that for, you know, for RP1 and Pi5.

00:56:54.480 --> 00:57:00.660
<v Liam Fraser>You know, when we were sort of prototyping on FPGA, you might have a Linux driver that can,

00:57:00.660 --> 00:57:05.800
<v Liam Fraser>you know, shove a frame out over to a DSI display, for example.

00:57:06.320 --> 00:57:12.320
<v Liam Fraser>But then there's probably a year's work to go from that to I can support any DSI display

00:57:12.320 --> 00:57:16.760
<v Liam Fraser>at any resolution and present it as a video device in video for Linux, you know?

00:57:17.020 --> 00:57:20.540
<v Liam Fraser>And it's the same story for a camera and all that kind of stuff.

00:57:21.080 --> 00:57:21.800
<v Liam Fraser>So there's been...

00:57:21.800 --> 00:57:24.200
<v Chris Gammell>And it's all upstream and, like, well-tested.

00:57:24.420 --> 00:57:24.600
<v Chris Gammell>Yeah.

00:57:24.680 --> 00:57:24.860
<v Chris Gammell>Right.

00:57:24.940 --> 00:57:26.620
<v Chris Gammell>Built without errors.

00:57:26.620 --> 00:57:26.960
<v Chris Gammell>And, yeah.

00:57:26.980 --> 00:57:27.880
<v Chris Gammell>I mean, like, there's so many...

00:57:27.880 --> 00:57:29.380
<v Chris Gammell>Like, that's one of the things that I think about, too.

00:57:29.440 --> 00:57:33.500
<v Chris Gammell>Like, you know, I am the piddliest little Linux user of the world.

00:57:33.500 --> 00:57:38.820
<v Chris Gammell>But, like, just the number of people that, like, are involved in the whole thing to make

00:57:38.820 --> 00:57:41.340
<v Chris Gammell>my RetroPie work, right?

00:57:41.420 --> 00:57:43.280
<v Chris Gammell>Like, I love RetroPie.

00:57:43.360 --> 00:57:43.860
<v Chris Gammell>It's so great.

00:57:43.980 --> 00:57:49.020
<v Chris Gammell>But, like, just the number of people in that software value chain that gets down to me

00:57:49.020 --> 00:57:50.440
<v Chris Gammell>so I could play, you know, Zelda.

00:57:51.580 --> 00:57:52.260
<v Chris Gammell>It's just...

00:57:52.260 --> 00:57:54.880
<v Chris Gammell>It's really crazy how many people are there.

00:57:54.880 --> 00:57:56.540
<v Chris Gammell>So it starts with you guys.

00:57:56.580 --> 00:57:57.000
<v Chris Gammell>That's great.

00:57:57.000 --> 00:57:58.360
<v James Adams>Yeah.

00:57:58.540 --> 00:58:04.440
<v James Adams>Well, I mean, it's amazing that we can provide the kind of hardware that, you know, allows

00:58:04.440 --> 00:58:06.000
<v James Adams>people to do all this cool stuff as well.

00:58:06.160 --> 00:58:08.820
<v James Adams>So, yeah, it's been a fun 10-plus years, actually.

00:58:09.100 --> 00:58:09.740
<v Chris Gammell>10 years, yeah.

00:58:09.960 --> 00:58:10.880
<v James Adams>See how far we've come.

00:58:11.140 --> 00:58:11.400
<v James Adams>Yeah.

00:58:11.860 --> 00:58:12.100
<v James Adams>Yeah.

00:58:12.520 --> 00:58:14.720
<v James Adams>I mean, for me, I wasn't a founder.

00:58:15.540 --> 00:58:18.120
<v James Adams>But I joined, basically, as soon as...

00:58:18.120 --> 00:58:21.880
<v James Adams>It was obvious that it needed some permanent people to sort of run it.

00:58:21.880 --> 00:58:24.080
<v James Adams>But I started in the beginning of 2013.

00:58:24.560 --> 00:58:25.740
<v James Adams>So, yeah, we're...

00:58:25.740 --> 00:58:27.860
<v James Adams>For me, it's over 10 years now.

00:58:28.080 --> 00:58:28.920
<v James Adams>But it's been fun, right?

00:58:28.960 --> 00:58:29.880
<v James Adams>When we started, we didn't...

00:58:30.420 --> 00:58:31.340
<v James Adams>We didn't even have an office.

00:58:31.740 --> 00:58:37.500
<v James Adams>And now we're at, like, 120 sort of people on the trading, sort of on the tech company

00:58:37.500 --> 00:58:38.180
<v James Adams>side, if you like.

00:58:38.280 --> 00:58:38.440
<v James Adams>Yeah.

00:58:38.440 --> 00:58:42.240
<v James Adams>And then another at least 120 on the sort of the charity side.

00:58:42.400 --> 00:58:44.040
<v James Adams>So, it's become quite big.

00:58:44.740 --> 00:58:44.820
<v James Adams>Yeah.

00:58:45.020 --> 00:58:47.360
<v Chris Gammell>That's a silicon company right there, right?

00:58:48.660 --> 00:58:49.060
<v Chris Gammell>Yeah.

00:58:49.740 --> 00:58:51.180
<v Chris Gammell>A very small silicon company.

00:58:51.820 --> 00:58:52.560
<v Chris Gammell>Yeah, right, right.

00:58:52.580 --> 00:58:52.860
<v Chris Gammell>Of course.

00:58:52.940 --> 00:58:53.520
<v Chris Gammell>Yeah, yeah, exactly.

00:58:53.560 --> 00:58:54.260
<v Chris Gammell>It's all relative, right?

00:58:54.720 --> 00:58:55.040
<v Chris Gammell>Okay.

00:58:55.100 --> 00:58:59.920
<v Chris Gammell>So, last time, I'm looking at notes from our last show in 2021.

00:59:00.920 --> 00:59:02.440
<v Chris Gammell>I asked about using...

00:59:02.440 --> 00:59:04.960
<v Chris Gammell>I asked on the Raspberry Pi 5 if you said...

00:59:04.960 --> 00:59:08.780
<v Chris Gammell>If you were going to use the 2040 on the Raspberry Pi 5.

00:59:08.920 --> 00:59:10.460
<v Chris Gammell>And you said no, because of cost.

00:59:11.240 --> 00:59:14.380
<v Chris Gammell>So, is RP1 actually a cost-down version?

00:59:14.460 --> 00:59:17.940
<v Chris Gammell>Or is it just because it would be in addition to other things at the time?

00:59:18.660 --> 00:59:19.460
<v James Adams>I mean, I'll tell you what.

00:59:19.480 --> 00:59:24.140
<v James Adams>If we had our time again with RP1, we would have put a 2040...

00:59:24.140 --> 00:59:25.720
<v James Adams>Kind of an actual 2040 in it.

00:59:25.940 --> 00:59:28.900
<v James Adams>We've sort of got a proto-ish 2040 type thing.

00:59:29.220 --> 00:59:29.560
<v James Adams>Uh-huh.

00:59:29.560 --> 00:59:34.200
<v James Adams>As you can see, you know, there's a couple of ARM cores and a bit of PIO and some memory.

00:59:34.760 --> 00:59:38.780
<v James Adams>But, yeah, I mean, it would have been really lovely to have had the whole thing synchronized

00:59:38.780 --> 00:59:41.000
<v James Adams>so that we could actually put an RP2040 in it.

00:59:41.380 --> 00:59:41.780
<v James Adams>Got it.

00:59:42.580 --> 00:59:46.460
<v James Adams>But putting a physical one on the board, I mean, you know, it's a 7x7 QFN chip.

00:59:46.540 --> 00:59:47.260
<v James Adams>Where would we put it?

00:59:47.300 --> 00:59:47.940
<v James Adams>What would it do?

00:59:48.480 --> 00:59:54.300
<v James Adams>As we've gone down the generations in Raspberry Pi, we've always tried to hoover up more and more of the function

00:59:54.300 --> 00:59:58.240
<v James Adams>into the main bits of silicon, either the power management chip or the main chip

00:59:58.240 --> 01:00:00.340
<v James Adams>or the I.O. controller now.

01:00:00.940 --> 01:00:04.600
<v James Adams>And so what you see on something like Pi 5 is it's just got those big bits of silicon

01:00:04.600 --> 01:00:07.100
<v James Adams>and a memory, of course, and Wi-Fi.

01:00:07.700 --> 01:00:11.240
<v James Adams>And the rest of it's just kind of like connectors and capacitors and resistors.

01:00:11.960 --> 01:00:17.240
<v James Adams>We've taken away a lot of control, sort of external control silicon type stuff.

01:00:17.880 --> 01:00:18.860
<v James Adams>That's good for two reasons.

01:00:18.860 --> 01:00:22.380
<v James Adams>One is it lowers the cost in general.

01:00:23.000 --> 01:00:24.820
<v James Adams>And it also makes your supply chain a bit simpler.

01:00:25.640 --> 01:00:26.040
<v James Adams>Yeah.

01:00:26.760 --> 01:00:28.360
<v James Adams>So it's kind of a good thing to do.

01:00:28.680 --> 01:00:32.440
<v James Adams>And also, yeah, board space is always actually now really is an issue for us.

01:00:32.560 --> 01:00:37.200
<v James Adams>You know, trying to cram everything onto a credit card is getting trickier and trickier.

01:00:38.280 --> 01:00:38.500
<v Chris Gammell>Yeah.

01:00:38.980 --> 01:00:42.100
<v Chris Gammell>Well, that's Eben for choosing that form factor to start with.

01:00:42.220 --> 01:00:44.260
<v Chris Gammell>That's, you know, you just got to give him a hard time, right?

01:00:44.340 --> 01:00:44.700
<v Chris Gammell>Come on, man.

01:00:44.960 --> 01:00:45.840
<v Chris Gammell>Why credit card?

01:00:45.840 --> 01:00:47.620
<v Chris Gammell>Why not like passport sized or whatever?

01:00:48.780 --> 01:00:52.060
<v James Adams>I did have a prototype of Pi 4, which was three millimeters longer.

01:00:52.420 --> 01:00:54.000
<v James Adams>And yeah, he...

01:00:54.000 --> 01:00:55.360
<v James Adams>Hard no, huh?

01:00:55.440 --> 01:00:56.380
<v James Adams>Basically, yeah.

01:00:56.400 --> 01:00:58.800
<v James Adams>Basically just like poked me until it was the right size again.

01:00:59.760 --> 01:01:00.740
<v James Adams>Didn't let me rest.

01:01:01.160 --> 01:01:01.400
<v James Adams>Yeah.

01:01:01.600 --> 01:01:02.700
<v James Adams>That was just not acceptable.

01:01:03.180 --> 01:01:04.400
<v James Adams>And fair enough, right?

01:01:04.420 --> 01:01:05.740
<v James Adams>We managed to squeeze it back again.

01:01:06.200 --> 01:01:06.560
<v James Adams>Yeah.

01:01:06.560 --> 01:01:06.620
<v Chris Gammell>Yeah.

01:01:06.620 --> 01:01:06.840
<v Chris Gammell>I remember.

01:01:06.960 --> 01:01:08.400
<v Chris Gammell>So I saw some of the stuff.

01:01:08.600 --> 01:01:12.760
<v Chris Gammell>I saw some people complaining online, which is like basically like a universal, right?

01:01:12.860 --> 01:01:14.120
<v Chris Gammell>All things complaining, right?

01:01:14.120 --> 01:01:15.340
<v Chris Gammell>But like the...

01:01:15.340 --> 01:01:20.660
<v Chris Gammell>I saw the ports layout change back, but it wasn't anything extreme.

01:01:20.660 --> 01:01:25.400
<v Chris Gammell>It's just like, yeah, you have to buy a new $5 plastic case for this new one.

01:01:25.480 --> 01:01:26.060
<v Chris Gammell>Okay, fine.

01:01:26.100 --> 01:01:26.260
<v Chris Gammell>Right.

01:01:26.360 --> 01:01:30.640
<v Chris Gammell>But it did change back again because of, I'm guessing, the geometry of the layout and stuff

01:01:30.640 --> 01:01:31.080
<v Chris Gammell>like that, right?

01:01:31.520 --> 01:01:32.160
<v James Adams>Yeah, pretty much.

01:01:32.160 --> 01:01:35.160
<v James Adams>So we decided on the how the...

01:01:35.780 --> 01:01:37.160
<v James Adams>So it's another...

01:01:37.700 --> 01:01:42.400
<v James Adams>If you look at Pi 5, if you look at the chips on it and you look at the routing, every chip

01:01:42.400 --> 01:01:47.920
<v James Adams>has been co-designed or designed to have the pins and the wires coming out in exactly the

01:01:47.920 --> 01:01:48.320
<v James Adams>right place.

01:01:48.960 --> 01:01:55.300
<v James Adams>I guess the classic, the best example on Pi 5 is the PCI Express bus between RP1 and 2712.

01:01:55.300 --> 01:02:01.420
<v James Adams>There's a big, just a big multi-lane motorway of differential pairs and coupling caps.

01:02:02.300 --> 01:02:06.600
<v James Adams>And if any of those crossed over, you blow up your PCB board tech.

01:02:06.940 --> 01:02:07.000
<v James Adams>Yeah.

01:02:07.000 --> 01:02:09.520
<v James Adams>So each of these things has been designed to be exactly...

01:02:10.520 --> 01:02:12.500
<v James Adams>The pinout has been engineered carefully.

01:02:13.020 --> 01:02:15.200
<v Chris Gammell>Another benefit to doing your own silicon, honestly, right?

01:02:15.260 --> 01:02:15.740
<v Chris Gammell>I mean, like...

01:02:15.740 --> 01:02:16.360
<v James Adams>Absolutely, right.

01:02:16.360 --> 01:02:18.200
<v Chris Gammell>I don't get to do that when I'm just buying off the shelf.

01:02:18.280 --> 01:02:20.260
<v Chris Gammell>I'm like, okay, well, I guess this one goes all the way around here.

01:02:20.260 --> 01:02:20.500
<v Chris Gammell>You know?

01:02:21.960 --> 01:02:25.000
<v James Adams>So that whole full stack thing really makes a difference.

01:02:25.100 --> 01:02:29.460
<v James Adams>And yeah, we can get away with a six-layer board exactly the same as the Pi 4, the same

01:02:29.460 --> 01:02:30.000
<v James Adams>PCP tech.

01:02:30.420 --> 01:02:35.060
<v James Adams>It does have blind and buried wires, but it's still a relatively cheap PCB.

01:02:35.720 --> 01:02:38.780
<v James Adams>And the PCB is quite a reasonable part of the cost of the Pi.

01:02:38.920 --> 01:02:40.960
<v James Adams>So, you know, that really makes a difference, that kind of...

01:02:40.960 --> 01:02:42.980
<v Chris Gammell>Well, eventually you get the...

01:02:42.980 --> 01:02:44.720
<v Chris Gammell>You do have some benefit of volume.

01:02:45.080 --> 01:02:45.860
<v Chris Gammell>I feel like...

01:02:45.860 --> 01:02:46.160
<v Chris Gammell>Oh, we do.

01:02:46.160 --> 01:02:46.860
<v Chris Gammell>...the blind and buried.

01:02:47.220 --> 01:02:48.300
<v Chris Gammell>Yeah, that might...

01:02:48.300 --> 01:02:50.580
<v Chris Gammell>You know, eventually they're like, well, we'll give that to you for free, right?

01:02:50.660 --> 01:02:51.780
<v Chris Gammell>You know, you're making enough boards.

01:02:52.100 --> 01:02:54.060
<v James Adams>Oh, no, we don't get anything for free.

01:02:54.560 --> 01:02:55.740
<v James Adams>I know.

01:02:55.800 --> 01:02:56.260
<v James Adams>But no, no, no.

01:02:56.280 --> 01:02:59.840
<v James Adams>I mean, you know, we get relatively better pricing because of the volume.

01:03:00.060 --> 01:03:03.920
<v James Adams>But, you know, each of those things is an extra process step and you do pay for it.

01:03:04.120 --> 01:03:06.580
<v Chris Gammell>Five cents can impact the end result, right?

01:03:06.600 --> 01:03:07.560
<v Chris Gammell>Everything is cost-based.

01:03:07.720 --> 01:03:08.080
<v Chris Gammell>Yeah, totally.

01:03:08.320 --> 01:03:08.440
<v Chris Gammell>Yeah.

01:03:09.140 --> 01:03:15.180
<v Chris Gammell>Like I asked last time, I'll ask again, where's RISC-V in the Raspberry Pi?

01:03:15.180 --> 01:03:19.220
<v Chris Gammell>If I say, if I walked into the Raspberry Pi office and I said RISC-V, do I get pushed out the door?

01:03:19.780 --> 01:03:21.040
<v Chris Gammell>Or do I get welcomed in with it?

01:03:21.280 --> 01:03:22.180
<v James Adams>No, I don't think so.

01:03:22.580 --> 01:03:23.560
<v James Adams>We like RISC-V.

01:03:23.720 --> 01:03:24.800
<v James Adams>I think we keep an eye on it.

01:03:25.060 --> 01:03:25.220
<v James Adams>Yeah.

01:03:25.640 --> 01:03:27.780
<v James Adams>Well, we have a kind of members program.

01:03:28.180 --> 01:03:29.400
<v James Adams>I'm sorry, I forgot exactly what it's called.

01:03:29.400 --> 01:03:32.320
<v Liam Fraser>I think it's the RISC-V Foundation, something like that.

01:03:32.660 --> 01:03:32.780
<v James Adams>Yeah.

01:03:32.780 --> 01:03:33.840
<v James Adams>Yeah, RISC-V Foundation.

01:03:34.060 --> 01:03:35.840
<v James Adams>Pay our fees to keep an eye on it.

01:03:35.840 --> 01:03:38.300
<v James Adams>I mean, it could be a thing in the future.

01:03:38.860 --> 01:03:41.380
<v James Adams>It's just, you know, maturity of things like ARM.

01:03:41.640 --> 01:03:47.400
<v James Adams>I guess when you're engineering these chips, ARM, and actually RISC-V have helped in this regard,

01:03:47.580 --> 01:03:52.840
<v James Adams>is they've made ARM a bit keener and they've changed some of their processes to make getting their IP a bit easier.

01:03:53.320 --> 01:03:54.140
<v James Adams>So that's nice.

01:03:54.260 --> 01:03:56.620
<v James Adams>But in general, it's very easy to get hold of ARM IP.

01:03:57.020 --> 01:03:57.740
<v James Adams>It works very well.

01:03:57.780 --> 01:03:58.280
<v James Adams>It's well tested.

01:03:58.380 --> 01:03:59.040
<v James Adams>It's well documented.

01:03:59.600 --> 01:04:02.700
<v James Adams>The ARM cores, you know, they have very high performance cores if that's what you want.

01:04:02.700 --> 01:04:05.820
<v James Adams>And, of course, that's what we kind of use.

01:04:06.260 --> 01:04:07.900
<v James Adams>There isn't really any RISC-V equivalent.

01:04:08.440 --> 01:04:10.700
<v James Adams>And then there's the kind of maturity of the software tools.

01:04:10.940 --> 01:04:15.080
<v James Adams>You know, I think in the stack of maturity, x86 is way, way up here.

01:04:15.360 --> 01:04:18.900
<v James Adams>And ARM's had a fair bit of, you know, attention down here.

01:04:18.980 --> 01:04:21.360
<v James Adams>And then RISC-V has kind of got a long way to go, I think.

01:04:21.760 --> 01:04:21.860
<v Chris Gammell>Yeah.

01:04:22.200 --> 01:04:28.380
<v James Adams>And it's all of those things together just makes ARM more compelling at the moment, right?

01:04:28.380 --> 01:04:29.660
<v James Adams>It's just the right answer.

01:04:29.660 --> 01:04:37.260
<v James Adams>For us guys who have a small team and, you know, we need this stuff, you know, a lot of stuff in the RISC-V world just doesn't exist that we would need.

01:04:37.460 --> 01:04:37.600
<v Chris Gammell>Right.

01:04:37.920 --> 01:04:38.560
<v Chris Gammell>I was wondering about that.

01:04:38.620 --> 01:04:46.960
<v Chris Gammell>Like you said, with all of the IP blocks that are, like, listed here, too, like, is it would be, like, you'd have to go and build your own SPI controller IP at that point then?

01:04:47.780 --> 01:04:51.100
<v James Adams>That's got, you know, that's an independent thing to RISC-V, so that's not a problem.

01:04:51.100 --> 01:04:52.640
<v James Adams>But it's more the process of calls, right?

01:04:52.760 --> 01:04:56.360
<v James Adams>And then the software tools and debuggers and all of that kind of stack, right?

01:04:56.440 --> 01:04:59.240
<v James Adams>The ecosystem isn't as mature as ARM.

01:04:59.640 --> 01:05:01.500
<v James Adams>And, you know, we kind of, you know, we like ARM.

01:05:01.880 --> 01:05:04.000
<v James Adams>They're a Cambridge company like us.

01:05:04.520 --> 01:05:05.860
<v James Adams>And we have a great relationship with them.

01:05:06.020 --> 01:05:09.620
<v James Adams>So it's just that, really, more than anything.

01:05:10.080 --> 01:05:12.580
<v James Adams>So we're definitely not anti-RISC-V, for sure.

01:05:12.800 --> 01:05:14.260
<v James Adams>We're definitely interested in it.

01:05:14.740 --> 01:05:16.260
<v James Adams>But we also like ARM.

01:05:16.480 --> 01:05:16.560
<v James Adams>Sure.

01:05:16.840 --> 01:05:16.980
<v James Adams>Okay.

01:05:17.120 --> 01:05:17.580
<v James Adams>That makes sense.

01:05:17.660 --> 01:05:17.760
<v Chris Gammell>Yeah.

01:05:17.880 --> 01:05:18.900
<v James Adams>They build good stuff.

01:05:19.360 --> 01:05:19.500
<v James Adams>Yeah.

01:05:19.760 --> 01:05:20.260
<v Chris Gammell>Yeah, that's great.

01:05:20.260 --> 01:05:20.660
<v Chris Gammell>That's great.

01:05:21.440 --> 01:05:29.660
<v Chris Gammell>Anything you'd love the audience to know or things to check out on the new Raspberry Pi 5 or the RP1 when they're looking at the Raspberry Pi 5 that they buy?

01:05:30.640 --> 01:05:33.280
<v Liam Fraser>There's your signature hiding on the board somewhere, James.

01:05:33.840 --> 01:05:34.660
<v James Adams>Oh, yes.

01:05:35.440 --> 01:05:36.300
<v James Adams>Is that on every board?

01:05:36.400 --> 01:05:36.720
<v James Adams>Is that why?

01:05:37.840 --> 01:05:40.740
<v James Adams>It's under the USB 3 connector, I think.

01:05:40.920 --> 01:05:41.060
<v James Adams>Yes.

01:05:41.520 --> 01:05:46.820
<v James Adams>I think from Pi 4, because it's such a, yeah, I still do the layouts for these.

01:05:46.820 --> 01:05:50.740
<v James Adams>It's become such a big, you know, it's a long-term project, right?

01:05:50.780 --> 01:05:57.420
<v James Adams>Because we're, you know, we're designing these things for a long time because you're doing chips with the pinouts and all the rest of it.

01:05:57.420 --> 01:06:00.480
<v James Adams>So to my mind, this is now kind of a bit of digital art.

01:06:00.580 --> 01:06:01.620
<v James Adams>That's why I like to think about it.

01:06:01.660 --> 01:06:03.880
<v James Adams>And I take a long time to make them look nice.

01:06:04.120 --> 01:06:06.240
<v James Adams>So I just decided I'll just start signing them.

01:06:06.460 --> 01:06:09.380
<v James Adams>And it's hopefully not gratuitous because it's then hidden by the connector.

01:06:09.380 --> 01:06:15.580
<v James Adams>But, yeah, I kind of, like, put a lot of effort into it.

01:06:15.680 --> 01:06:18.520
<v James Adams>So it feels like it might be a good thing to sign.

01:06:19.640 --> 01:06:19.980
<v Chris Gammell>That's great.

01:06:20.340 --> 01:06:29.940
<v Chris Gammell>No, I think that is, I think Easter eggs and, you know, signatures and things like that, those are the things, too, where, like, that's just, like, trivia into the future, too.

01:06:30.080 --> 01:06:35.700
<v Chris Gammell>You know, like seeing old signatures under, like, the case of, like, early computers and stuff like that.

01:06:36.340 --> 01:06:38.420
<v Chris Gammell>You know, it's like we have to make our own mythologies, basically.

01:06:38.420 --> 01:06:39.180
<v Chris Gammell>It's fine.

01:06:39.440 --> 01:06:40.660
<v James Adams>So I like it.

01:06:41.940 --> 01:06:43.760
<v James Adams>Is there anything else to say about Pi 5?

01:06:44.080 --> 01:06:47.500
<v James Adams>I mean, it's, I think we are, it's the most Raspberry Pi, Raspberry Pi.

01:06:47.660 --> 01:06:54.220
<v James Adams>I think you've kind of already alluded to this, the fact that we've kind of touched every bit of silicon pretty much to make it work.

01:06:54.640 --> 01:06:59.660
<v James Adams>I think it's, you know, kind of just advertising for Raspberry Pi now, but it's a great platform.

01:06:59.840 --> 01:07:01.240
<v James Adams>You know, we enjoy using it internally.

01:07:01.720 --> 01:07:02.080
<v James Adams>Right.

01:07:02.480 --> 01:07:02.660
<v James Adams>Yeah.

01:07:02.820 --> 01:07:07.540
<v James Adams>It's really fast, and it's going to get better as we evolve the software, and, you know, it's early days.

01:07:07.540 --> 01:07:10.180
<v James Adams>So it's just a product we're very proud of.

01:07:10.980 --> 01:07:11.040
<v Chris Gammell>Yeah.

01:07:11.040 --> 01:07:22.820
<v Chris Gammell>I was watching Jeff's, Jeff Geerling's review of it, and I was, like, looking at it, I'm like, that, I mean, that could, that could be something you just take with you and just play around with and have, you know, like, the display and the processing and all this stuff.

01:07:22.820 --> 01:07:24.780
<v Chris Gammell>Like, it's not a laptop, of course, right?

01:07:24.780 --> 01:07:26.600
<v Chris Gammell>But it's not, not bad.

01:07:26.760 --> 01:07:27.840
<v Chris Gammell>Definitely better than my old laptops.

01:07:28.920 --> 01:07:37.580
<v Chris Gammell>So, like, I definitely use slower and, like, you know, like, obviously the software is just so ubiquitous and useful.

01:07:37.840 --> 01:07:38.600
<v Chris Gammell>So, yeah.

01:07:39.260 --> 01:07:39.540
<v James Adams>Yeah.

01:07:39.540 --> 01:07:48.700
<v James Adams>I mean, if you want to use something like Office 365 or Teams or Google Meet, all this stuff now works really well on the Pi 5, whereas in Pi 4 is kind of a bit hit and miss.

01:07:49.220 --> 01:07:50.320
<v James Adams>Nobody wants to use Teams.

01:07:50.420 --> 01:07:51.880
<v James Adams>In the history of the world, no one has wanted to use Teams.

01:07:51.880 --> 01:07:52.680
<v James Adams>Well, yeah.

01:07:52.780 --> 01:07:58.200
<v James Adams>Sometimes we get forced to do this, but even Teams works very well.

01:07:58.200 --> 01:07:59.380
<v Chris Gammell>Yeah, that's, that's interesting.

01:07:59.460 --> 01:08:05.540
<v Chris Gammell>Yeah, and I think that does actually open up other product ideas and, you know, just, like, accessibility to the underlying tech.

01:08:05.580 --> 01:08:07.540
<v Chris Gammell>And it's, like, it's not some bespoke thing.

01:08:07.540 --> 01:08:12.820
<v Chris Gammell>It's, like, an off-the-shelf piece of hardware that you can deploy for lots of different things now.

01:08:12.940 --> 01:08:13.500
<v Chris Gammell>It's great.

01:08:13.700 --> 01:08:14.960
<v Chris Gammell>It's really, really good.

01:08:15.080 --> 01:08:17.120
<v Chris Gammell>And I'm excited to have it.

01:08:18.000 --> 01:08:19.000
<v Chris Gammell>Guys, thanks for coming back on.

01:08:19.280 --> 01:08:21.140
<v Chris Gammell>I look forward to Raspberry Pi 6.

01:08:21.220 --> 01:08:21.820
<v Chris Gammell>So, get started.

01:08:22.060 --> 01:08:25.660
<v Chris Gammell>Get started on the RP, what are you, probably, 4, 5, 6?

01:08:25.660 --> 01:08:27.320
<v Chris Gammell>RP, 4, 5, 6, whatever's up there.

01:08:27.340 --> 01:08:28.300
<v Chris Gammell>And then Raspberry Pi 6.

01:08:28.540 --> 01:08:30.620
<v Chris Gammell>I'm sure there's no shortage of things to do.

01:08:30.880 --> 01:08:34.480
<v Chris Gammell>And, you know, or whatever, the RP, 20, 40, whatever the next thing of that is.

01:08:34.620 --> 01:08:36.680
<v Chris Gammell>Well, we'd love to talk more about that in the future.

01:08:37.240 --> 01:08:37.940
<v Chris Gammell>So, thanks for being here today.

01:08:38.360 --> 01:08:38.820
<v James Adams>Been a pleasure.

01:08:38.940 --> 01:08:39.660
<v James Adams>Thank you very much, Chris.

01:08:39.980 --> 01:08:40.740
<v Chris Gammell>Thanks for having us.

01:08:40.900 --> 01:08:41.100
<v Chris Gammell>Cheers.

01:08:55.660 --> 01:08:57.420
<v Speaker ?>Cheers.

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:08:57.420
<v Speaker ?>

01:08:57.420 --> 01:09:01.980
<v Speaker ?>administered in administered

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:01.980
<v Speaker ?>

01:09:01.980 --> 01:09:31.960
<v Speaker ?>Thank you.
