Oscar's new PDP-10 replica (and PDP-8 and PDP-11 too)
[Marc] Hello, so today we have Oscar - how do you say your last name? [Oscar] Vermeulen [Marc] Where are you from? [Oscar] Holland originally, although Switzerland these days. [Marc] Okay, and you brought all your PDP replicas? [Oscar] Unfortunately not all intact! [Marc] You were just having some some slight issues... What happened? [Oscar] So it looks like three airline flights were too much. But apparently you have M3 nuts and bolts in your collection? [Marc] Yes, I used to live in Europe, I still understand the metric system, although...
[Oscar] It makes a lot of sense. [Marc] ...it's very very hard to think about it. We also have some people that know about DEC machines, because I don't know anything. So we have Lyle, he restored the PDP-1.
And hiding in the other corner... [Otto] I'm Otto. [Marc] And you're Oscar's sidekick? [Otto] I'm Oscar's sidekick since 1988. [Marc] Well, hello and welcome back. Today Oscar has brought us some of his beautiful
reproductions of early Digital Equipment Corporation computers, or DEC as the company became known. They are replicas of some of the most popular PDP computer models, with meticulously reproduced front panels. They run software emulations of the original computers on a Rasberry Pi hidden behind the panel, so he calls them PiDP computers.
We are particularly excited about the newest member of the family, the PiDP-10. And if you follow the channel, you have probably recognized Lyle Bickley, who gave us a tour of the PDP-1 he restored at the Computer History Museum. The PDP-1 was DEC’s first computer, introduced in 1959. PDP stood for Programmed Data Processor, a name that was chosen to circumvent a government rule that prohibited the purchase of computers at the time. But since you asked for
a Programmed Data Processor, not a computer, the order went right through. Problem solved. So, DEC went on to make many PDP computer models in the 1960s and 1970s, from the very small to the quite large. One of their most successful early model was the tiny (for the time) PDP-8, introduced 1965, a minimalistic 12-bit mini-computer that retailed for below $20,000. This was a breakthrough in computer affordability and size at the time. An even more successful model was the PDP-11, a full-fledged 16-bit minicomputer that supported multi-user time sharing. It was one of the most popular minicomputers of the 1970s,
once again an industry leader in the performance to price ratio. But foremost, we want to talk about the PDP-10, or more exactly, Oscar’s latest recreation, the PiDP-10 as he calls it. The PDP-10 was a large 36-bit mainframe machine, an affordable super computer of sorts, introduced in 1967. It found its way to MIT’s AI lab shortly after its introduction, and was the machine behind many a computer breakthrough in computer science, as you’ll see in a minute.
[Marc] So, tell me! So, this is, on the left - oh that's written on it: PDP-11. [Oscar] Yeah, that's the 11/70. [Marc] It's a shrink-ified version of it? [Oscar] It's ⅔. Scale two to three. [Marc] Those are available as kits? You sell those, right? [Oscar] Yeah, as a kit for the last few years, and now we also have them assembled, if people don't like the soldering. [Marc] And your newest one is what we wanted to show off.
[Oscar] That's the dearly beloved PDP-10! We worked for seven years on this. And actually the externals, the physical replica, is one thing. But the real project, in some way, is Lars Brinkhoff's ITS reconstruction project. [Marc] And the switches, they feel so incredibly good! They feel like the old switches, the good ones when they weren't flimsy.
So we have laid all the beautiful replicas Oscar has made of DEC machines. So how did you start this whole thing? [Oscar] Oh, just because I wanted a PDP-8, and my living room was too small for a real one, so... [Marc] Oh, who doesn't want a PDP-8 in their living room? [Oscar] Well, my wife! And so, I discovered also that the original machines are quite expensive. So
I made a replica for myself, and I decided to go full OCD and make a nice sort of instrument panel. And the factory that could make them said you have to order 50 of them in one go. So then I had to look for 49 fellow idiots that wanted a PDP-8. But it turned out, there were many more than this. So I ended starting this sort of kit making thing, and I started making these things at the dinner table. But after a while,
of course, the desire to do an 11 became quite strong, right? Because while the 8 is sort of the minimalist computer, the 11 is the birthplace of Unix, so it deserves attention. That got me into injection molding, that's an injection molded case. [Marc] And how did you do the buttons? [Oscar] Oh, because the original switches that I used for the PDP-8, I found on AliExpress. And they turned out to be New Old Stock from the Chinese Army. So I couldn't buy anymore. And then,
amazingly, the switch manufacturer to pity on me, and said, okay if you pay $1,500, send us a 3D model of the switch you want to have, and we put that cap on a standard switch. [Marc] Oh! [Oscar] And when I found out that trick, then of course I thought, well, then I can also do the iconic PDP-11 switches. [Marc] Of course! [Oscar] Because they are the most interesting shape of all.
Then if you have a replica panel of the PDP-11 with replica switches, you must must have the white bezel, right? There's no way you can put this into a square box... [Marc] Of course! [Oscar] ...it hurts! So that brought me into injection molding of larger things. So I made a lot
of them. And then of course I met Lars Brinkhoff who was reconstructing the ITS operating system on the PDP-10. And that was such a cool project that I thought this must be encapsulated in hardware as well. But this was like 54 cm wide. It is very tricky, you get all
sorts of manufacturing problems, right? Cooling down, shrinking, warping. You need to know about injection molding. That's where Otto got involved. [Marc] It's an art, and then tooling is not for the faint-hearted. [Oscar] No! [Marc] Wow, the result is stunning. Those are fully working emulations of the real thing? [Oscar] Yes.
[Marc] So, you can run the original software, paddle your things in? [Oscar] Yeah! The real fun is to make these things into a sort of history capsule, where you have all the operating systems, all the old software reconstructed and working. You can boot up the PDP-8 with a multi-user TSS8 operating system, right? [Marc] You can do eight users on a PDP-8?! [Oscar] And it works! I mean you can do BASIC, FOCAL. OS8 is the normal boot option, because everybody loves OS8. If you want full pain, you can start it up in a hardware configuration where you have nothing but the front panel. And you toggle in your program on the front panel. [Marc] So, bare metal. [Oscar] Bare metal. You load the paper tape boot loader from the instructions that are
here. And then it will load the bootstrap from a USB stick that it thinks is a paper tape. [Marc] Okay. But the PDP-10, that's a more monstrous machine, right? [Oscar] Yeah. And for me, the late stage DEC, or pre-VAX DEC perhap,s I have complete now: the 8,
the 11 and the 10. So the temptation will be to go one generation before, with the PDP-1 [Marc] Ah! We have somebody that knows something about the PDP-1. [Oscar] Yes, indeed! [Marc] And you guys are working together, I understand? [Lyle] That's right. [Marc] So we have a PDP-1 at the CAHM that Lyle restored lovingly. It's actually my favorite video of you explaining the PD-1.
[Oscar] Yes, mine too. [Marc] And then so, we are going to have one that we can put in our living room, once we get rid of our wives? [Oscar] Yes, the same scale two to three as the others. [Marc] Excellent, all right! Okay, so let's get the PDP-1 through its paces. It's run by a Raspberry Pi, right? [Oscar] Yes. [Marc] The PDP 10 is already running? [Oscar] Yeah. By default, if I set all the switches to zero, it just runs a little demo program, that is sort of highly educational, to the front panel.
[Marc] Oh, so, by default it runs the blinkenlight program. [Oscar] Yeah. [Lyle] Plus you can use the switches to make it change pattern. [Marc] Oh, yeah! [Oscar] So, make it faster...
[Marc] So, how many switches, how many lights? [Oscar] 126, although maybe we settle for 128. [Marc] Lights? [Oscar] Lights. [Marc] And the switches? [Oscar] 74. [Lyle] 74, yeah. [Marc] Ah, you don't realize it's that many! [Oscar] Let me reconfigure the hardware for ITS booting. [Marc] ITS? So you have to... For me not being a DEC guy, what is ITS.
[Oscar] The Incompatible Timesharing System. [Marc] Okay. [Oscar] The better operating system. [Marc] That must be an academic thing, with a title like this.
[Oscar] Yes, right. You had a Compatible Timesharing System, and the people at the MIT AI lab felt that was a bit bureaucratic, and not all how they wanted it to be. So they wrote their own operating system. Which clearly was the Incompatible Timesharing System. Incompatible, because better. [Marc] It's just, poke in the eye, Mr corporate guy. [Oscar] Indeed! [Marc] So let’s talk about ITS for a minute. One of the great attractions of the machines
at the AI lab is that they were run like personal computers, where you could have the whole machine for yourself and run anything you pleased, including doing things that the regular operating systems would not support nor allow. Actually, that was the whole point. This is best described in the book Hackers by Steve Levy, where hacking is actually a good thing, the very underpinning of computer science progress at the AI lab. But this single user model became untenable as larger machines like the PDP-6, then the PDP-10 arrived. As recounted in the book, there was utmost consternation at the prospect of running the machines under the time sharing OS being developed at the time, which was called the Compatible Timesharing System or CTS. The first thing this would do is introduce all kind of security nannies to prevent average users to access the machine hardware directly and do mischief inadvertently. But of course, these were not average users, and direct access to the machine was primordial.
After a large upheaval, best recounted in the books, a compromise was reached between the users and the administration: the students would develop their own time sharing OS, which would be entirely under their control and not undermine the great plans they had for their new machines. This was of course rebelliously called the Incompatible Timesharing System, but of course this was not true, as the ITS was much more clever and way more compatible with any previous programs than the nannyfied CTS was. The entire ITS environment from the AI lab has been recreated from the original tape backups by Lars Brinkhoff, and runs on the PiDP-10. So the setup transports you back in time as if you were in the MIT AI lab, rubbing shoulders with computer luminaries and their user accounts that you can delve into. [Marc] You had a few things to say about ITS? [Lyle] Yeah. It is a really interesting operating system, because of the fact that, not only it's
got a funny name by Incompatible Timesharing System, but it has absolutely no security. And that's a big plus! Remember, this is for hackers. [Marc] Right! [Lyle] It's one of the big selling points of the PDP-10 and all the software the Lars did, is the fact that you have access to everything. So, when you start looking at the software, you can log in as Lars, and you can look at the games that he has up in his account.
[Marc] And you don't need his password? [Lyle] And you don't need a password to log into any account on the system. [Marc] Then you have the whole archive of the disk and everything, live in there. [Lyle] Exactly. [Marc] And you can explore it without any security.
[Lyle] Explore without any security. Now the way that they handled that, by the way, at the University, was that if you crash the system, it announces to everybody that you crashed the system with your username. So, it was a way to scare people into being careful. So the security, if anything, was minimal, but it was kind of effective because of the fact that it would have said you did it.
[Marc] The fingers would point at you and people would come to lynch you. [Lyle] That's right! [Oscar] So, let me start up the teletype. Because if you think you're firing up the old PDP-10, mainframe class machine, of course the first thing that you do, is there's a teletype attached to the machine where you do the bootstrap. [Marc] Oh, complete with sounds! [Oscar] Good. So now the system is in a boot-up state. I have to bring the machine to a halt with stop. [Marc] Okay.
[Oscar] And then, when I push 'read in', that actually triggers the paper tape to be read in. And once it is read in, to execute the last instruction on the paper tape. [Marc] And all those things - the paper tape, the teletype - that's all emulated inside the the Pi computer at the back of your PDP-1? [Oscar] Yeah! There's way more than that. [Marc] So, it's way more than just a PDP-10 emulation. [Oscar] Yeah, yeah. This is basically the MIT AI lab, right? You have to imagine that there is this sort of computer science lab at MIT. They're also very close to DEC, Digital Equipment Corporation.
So they get this machine inside their computer science lab, with the message: "this is kind of way ahead of anything else, super computer that we have at the moment. You go and play with it". This is an overview of all the hardware that was in that AI lab that we have recreated as a software simulation. You have the PDP-10 mainframe system, that's the heart of the whole thing. On the top left, there's a lot of storage devices attached to it: magnetic tape, hard drives, paper tape, all that kind of thing. Here on the left, you see that also the old PDP 6 was still connected.
[Marc] Yeah. And this was even connected to the internet, right? It has it has an IMP somewhere? [Oscar] Yeah, exactly. So at the bottom left is the IMP, the Internet Message Processor, I think. So at the moment, we use it to do TCP/IP, and get the machine on the internet.
[Marc] And so to do the PDP-10 justice, you re-implemented the entire environment. So you can run all the demos, all the things, connect to the internet. [Oscar] Yep! [Marc] Let's play with it, because you have the... There's the famous round screen,
right? You have an emulation of that? All right, oh there it is. [Oscar] And there it is. [Marc] It's at the bottom.
[Oscar] So that's the type 340 graphical display, as they call it. [Marc] 1024 x 1024, right? [Oscar] Yes! [Marc] High res. [Oscar] So let me quickly log in here. And then we're going to use the type 340 display that you saw. So the round tube is now a square
panel here on this display. Let me make it full screen, it's nicer. [Marc] Yeah, I was joking that we should have done a cardboard round circle. [Oscar] Making a replica type 340 is one of my next projects. [Lyle] Ah, there you go! [Marc] Coming, secret plan! [Oscar] I think Minskytron is a nice example. Minskytron was already famous from your from your PDP-1 times, right? [Marc] Yeah, it's my favorite program. From Professor Minsky, right? [Lyle] Yes, Marvin Minsky.
[Marc] At the MIT lab. [Lyle] He was early AI, a professor. [Marc] Yeah. Giant in the world of computing, right?
[Lyle] Exactly. [Marc] Oh oh oh oh oh, here, we are Minskytron-ing! [Oscar] So it just depends, you can play with the front panel toggles, which the program uses as input. And you can get to modify the Minskytron patterns, from more or less white noise, all the way to fairly complex patterns. [Marc] So, it's a chaotic simulation...
[Oscar] ...out of which some regular patterns emerge. [Marc] This is just... those MIT guys, I tell you. [Oscar] And as much as you like teletypes, they were not always efficient. So Tom Knight, one of the junior staff there - he arrived at MIT at age 14 I believe - developed this whole system built around the PDP-11. That was basically a graphics terminal, 16 graphics terminals at the same time, onto this main frame. And that would basically do in 1972 what you could do on a home computer maybe 20-25 years later. So let me bring up one of his terminals.
[Marc] Oh so, that's a terminal with special capabilities that they had in the lab? [Oscar] Totally custom made, yeah. So, to give you an idea, let me log in. So one of the major applications, of course, is to play Spacewar! [Marc] Okay. And the Spacewar! would be on on the round screen of course. [Oscar] And stupidly enough, I forgot to bring my joysticks with me. You can have up to four joysticks. [Marc] What the heck, it has three stars in it! [Oscar] You don't like that? Because when we demoed this...
[Marc] I didn't even know that existed! I'm accustomed to the PDP-1 with just one star. [Oscar] This because when we demoed this to Richard Greenblatt, he protested and said: yeah, but this is a later version of Spacewar!. Spacewar! has one sun, two spaceships. And then the beauty of the it operating system is, actually, the debugger is the command line, the command line is the debugger. So you can basically say: okay, you don't like the three stars, let me just load the symbol table and change the program as it is. So, let me break the program, control-K, to load the symbol table. And now, I can say, how many stars do you want? Let's say one. It is three at the moment, let's change it to one. And let's change the number of ships,
and set it to two. It will now run for a little bit further, to continue the current loop. But the next time it comes into its demo cycle, you will see that this historical inaccuracy has been fixed. Because now, you have one sun... [Marc] There you go, Spacewar! as we know it. [Oscar] ...and you have two spaceships.
[Marc] All right. [Lyle] Now on the PDP-1, there was a reason there was only one. And that is because it ran out of machine capability! It just didn't have the ability to do it. [Marc] It's so amazing it could even do two ships and a star field and a blinking star. [Oscar] Maybe it's nice to show chess? [Marc] Yes! [Oscar] This gives you the very first chess game that actually had a human rating - what you call it? - an ELO rating.
[Marc] Oh! [Oscar] And it was also the first chess program to beat a rated human. It was written by Richard Greenblatt. But you got to think, maybe this doesn't look spectacular today. But think this is 1972. The rest of the world is just moving from teletypes to VT-52, if they're lucky. And these guys have, you know, 16 graphical terminals hooked up to a mainframe. [Marc] So that's running on the PDP-10? [Oscar] That's on the 10, yeah. Shall we
do a bit of artificial intelligence? Because that's a typical application. [Marc] Yes, I need some. My intelligence is not artificial enough. Go for it. [Oscar] The program was called Shrdlu. This came out in 1969-1967. [Marc] Called what? [Oscar] Shrdlu.
[Lyle] S-H-R-D-L-U. And where that came from, from a row of the characters on a linotype machine. Has nothing to do with AI. But it's S-H-R-D-L-U. [Marc] Shrdlu. Okay, fine.
[Oscar] So here's the Shrdlu demo on the Type 340. [Marc] Ooh! [Oscar] And, a bit of background. So, this is an AI, now we're talking a proper AI, that lives in a small 3D world: a table with objects on it, cubes, boxes pyramids. They have colors, but the colors are written in text. So this is a blue pyramid, and this is a green box. [Marc] Hey, that will do. [Oscar] But basically the idea is, it has an English language parser. So you
give it complex commands. It will actually manipulate that 3D world in a complex way. [Marc] Go for it, give us PDP-10 intelligence. [Oscar] Pick up a big red block. [Marc] It did.
[Oscar] It did. That's relatively simple. [Marc] Okay. [Oscar] Grasp the pyramid. "I don't understand which pyramid you mean". [Marc] So, you have to tell it the color? [Oscar] There's two pyramids, right? [Marc] Oh, that's a demo, it self-runs? "Find the block which is taller than the one you are holding, and put it into the box". [Oscar] And then the PDP-10 shows its intelligence.
[Marc] Gee-wee! I would be perplexed with that command myself. And it did put something into a box! [Oscar] So, "What does the box contain?". [Marc] "A blue pyramid and a blue block". Very smart! [Oscar] Now, the original Shrdlu was actually interactive, so you you basically typed your your own questions in on the top line. But Winograd who programmed this, was quite open about it, the purpose was to demonstrate the power of AI, not be... [Marc] Right, right, what it could be in the near future.
[Lyle] By way this was Terry's PhD thesis. So, it was pretty amazing to think that this was done in the era it was. And he did his PhD thesis on this. This was AI back in early 70s. So that's pretty amazing.
[Marc] Pretty smart people at this MIT lab. Well, it's an impressive reconstitution of just... [Lyle] It's truly amazing. [Marc] Not just the machine, but the whole environment, the whole programs, the devices. [Lyle] Yes. It's the whole environment. [Oscar] Because in this demo you haven't seen a fraction of what's out there. I mean, there's so much stuff there.
One thing, we didn't show is that they have a color scope, right? A color version of the type 340 display. [Lyle] Okay. [Oscar] And there's some new graphics algorithm that Lars converted to PDP-10 assembler. [Marc] Oh, ah! [Lyle] Oh look at that! [Oscar] Isn't it amazing, that with just seven pages of PDP-10 assembly, you get this stuff? [Lyle] Right! [Oscar] 'Cause it's a new graphics algorithm that some Japanese guy posted on Twitter. Nobody really knows his name. [Marc] Oh, that's modern programming for the PDP-10? [Lyle] I love it! [Oscar] Yeah, yeah, six month old.
[Marc] Six months old! [Lyle] Oh, that's cool. [Marc] Excellent, okay. Well it's a world of exploration open to you in the - how do you call it? the P-I-DP-10? [Oscar] PiDP-10. [Marc] PiDP-10, all right. So if you feel like getting one of these replicas, go to obsolescence.dev, link in the doodly-doo, and order one for yourself.
It’s very old fashioned, you just send an email to Oscar and he’ll get one to you. So, happy retro-hacking, and see you in the next episode...
2024-06-05 06:15
