Advent of Computing - Episode 165 - LSI-11

Episode Date: September 21, 2025

This episode we continue my series on the PDP-11 by examining how DEC adapted to the advent of the microprocessor. Along the way we will see how the PDP-11 inspired new generations of computers, and the surprising connection to early digital hobbyists.  

When it comes to technology, I think there are certain facts that can act as thought-terminating cliches. And also, I realize how pretentious that sentence sounds. Please forgive me for a moment, if you will. This is something I've noticed over the years, and it came up pretty blatantly when I was reading about the IAS machine and its relatives. There was a paper about one of those related computers that opened along the lines of, Machine X is inspired by the IAS machine with a few notable differences. The authors will only be discussing those differences. That's a perfectly reasonable approach, right?

I mean, I'm lazy, and that kind of approach saves space and time. You're basically just saying, hey, you know most of this already. Let's just look at the changes. That relationship is the crucial fact here. This technology is like that other technology. You may get into the differences. You may even summarize what's similar, but that one fact kind of ends a big part of the discussion.

I've ran into this a number of times over the years, and it's starting to really bug me. You may be able to tell. You can find many primary sources that follow this pattern. And don't get me wrong, it's really nice to have contemporary sources that explain technology, especially when those explanations

put things in the larger context. The issue, however, is what that fact of relation stops. The 68,000 was inspired by the PDP 11. Here are some of the ways it's similar and some ways it's different. What's missing there? It's the why.

If you go cruising, you'll find very few sources that will say, why one technology was based off another technology. Yet you'll find plenty of papers that explain that there is, in fact, a connection, and here's how the new technology is slightly different. It's actually very innovative if you look at the details. So, why are things this way?

Well, I think it has to do with, well, that one nice word, context. In the 1970s, the PDP 11 was one of the better, design computers out there. If you were designing a new machine, well, of course, you'd be inspired. Why would you need to write that down? Everyone knows the PDP 11 is a great design. In the 1980s, the PC was the most successful computer out there.

So, of course, if you wanted to make a new computer, you'd make a PC compatible. Why would you need to explain that to your audience? Why would a reader ever need to know that context? I mean, they're only going to be reading this page. for the next few months, right? Welcome back to Advent of Computing. I'm your host, Sean Hass,

and this is episode 165, the LSI 11. And I'm back. Like, I'm back back now. I survived the lead-up to my wedding, my wedding, and actually also just learned a big backpacking trip. The two are unrelated, except temporally. So if you've been trying to get in touch with me, I swear I have seen your emails and messages. I just haven't had a spare minute to

reply. So I'll be going through my inbox very soon. But I'm back. The show's back, and we're back to the PDP 11 series. Today we're going to be continuing by examining how the 11 adapted to changes in technology in the early 70s. This is of particular interest for two reasons. The first is the deck, the company that made the PDP 11, claimed it was a machine, quote, immune to obsolescence. The basis of this claim was the machine's bus, the unibus. The entire system was composed of cards plugged into that bus.

So as technology changed, you could stay up to date by simply replacing cards. But did that claim pan out? How did the PDP 11 fare as new technology emerged? I think this is especially important given the time period. In 1971, Intel releases the 4-004, the first commercial microprocessor. This is the first entire computer compressed onto a single chip, or at least all its processing circuits. As early as 1969, four-face systems had released a similar product, a processor spread across four different chips. The PDP 11 enters the market right between these two steps, so it's immediately thrust into the flow of progress.

The first model of the new machine, the PDP 11-20, didn't use any of this new technology. Instead, it was all TTL logic, meaning it was built for. packages of transistors. The actual processor took up a pile of cards on the unibus. That means that the 11 was well positioned to take advantage of some new technology, at least in theory. The other reason I want to talk about the LSI 11 comes down to the matter of influence. The LSI 11 comes to market in 1975. By that time, there are a number of successful microprocessors in the wild. Some have even been claimed to be heavily patterned off the PDP 11. What I want to figure out is what influenced Deck to make the LSI 11. In my mind, there's

the possibility that Deck saw those 11-like microprocessors running around and thought, hey, that could be us. That should be our market. But that's just kind of the story I want. That's my little fun fanfic. I want to figure out what the actual motivation was. It could be that these inspired chips had nothing to do with it. By going down that path, we can also examine just how influential the PDP 11 was, and at what points its influence actually mattered. That's a rough plan for today. I guess I should give this disclaimer before I get much further, though. Since this is part of a series, you probably want to go back and listen to last episode, which covered the early days of the PDP 11 project.

This episode, as with most of my series, is still standalone. It'll make sense on its own, but you'll have a lot more context if you've heard last episode. With that, let's get started. Last time, we only really got up to the PDP 11's release and its immediate amazing sales figures. That's cool and all, but it doesn't answer the question of influence.

A product can do very well, but not leave a lasting technological or cultural impact. You can still buy slinkies everywhere, but slinky technology hasn't really made its way into some next-generation movement-based toy. Specifically, I want to work out early influence. Here I'm talking about the period between 1970 and 1975, from the release of the PDP 1120 to the release of the first LSI 11.

That should give us a good understanding of how impactful the PDP 11 was back in its day and give us a foundation to answer some other questions. I want to start us off with what should be the clearest example, or at least what's often propped up as the clearest example. In 1971, Motorola started efforts to develop their first first. microprocessor. That would lead to the Motorola 6800, released in 1974. In an oral history, Jeff Lavelle, engineering director of the 6800 project, explicitly states that this chip was inspired by the PDP 11. To quote, that was the time of the PDP 11, and it was the great high

technology computer at the time, so a little mini thing with a lot of switches, and so that was a neat architecture, and we liked that a lot. So we started with that architecture and built some variations on it, and the original product definition for the 6800 family was a 16-bit machine. And then we talked to Tom and he said, go away! And then we evolved into the real world of what we could build. And so we did it over a couple of years, end quote. The final processor would be much simplified and it would come in at eight bits wide. Now, the story is about as clear as we're going to get. The design team wanted to build a PDP11-like processor. They even made it 16-bit, but due to some technological realities, they had to shrink their designs. The result is a smaller PDP-11-inspired processor.

But how close to the 11 is Motorola's first chip? Well, uh, That's where things get a little off. Looking through the 6,800's designs, I can't find a resemblance, and I've been trying to. From the programmer's perspective, the hallmarks of the PDP 11 are pretty clear. Lots of general purpose registers, orthogonal instructions like move that are very complex,

and external devices are mapped into memory over the unibus. The 6,800 doesn't really work that way. The Motorola chip had two accumulators, an index register, and then a stack and instruction pointer. That's all special purpose. You can't, for instance, add a value to the indexed register, and you can't reference an address in memory using the value of the accumulator. Those operations are tied to specific registers.

However, there is a little register crossover. You can increment and increment the index register, which, does require some math circuits, and that is similar to one of the addressing modes on the PDP 11. You can also store and load the index register, but in general, the 6,800 looks and feels like a fancy load store computer. That's something the PDP 11 very much was not. Now, it would be unfair if I didn't err other arguments. I don't personally see the resemblance here, but some will argue that the 6,800, is clearly taking cues from the PDP 11.

The core of that argument comes down to the instruction set. Proponents say that the 6800 provides many of the same instructions that the PDP 11 does just in different ways, as in on a smaller scale with different addressing modes, that sort of thing. This kind of bears out. The 6800 does have a lot of instructions for such a small chip. Many of those instructions have equivalence on the PDP 11, but they're set up differently. The closest thing to move on the 6800 is actually load.

To make load almost general purpose, the 6800 actually has eight different load instructions, but they all use the same mnemonic. To load accumulator A with the contents of memory, you use Operation No. 86. To load Accumulator B, you would use Operation C6. Contrast that to the PDP11's Move. That one instruction works for everything, and just takes arguments to specify what the source and destination are. Move is only one operation. It's not a sneaky set of operations with the same name.

But there is something else here. If you look through both computers' instruction lists, you'll find there are no instructions for handling input and output to external devices. That's because the PDP11 uses memory mapped I.O to handle that job. Every external device is just treated like a special region of memory. The 6800 does the same thing. That is an undeniable influence. So what we get here is a bit of a mixed bag. The 6800 has some similarities to the PDP 11, but the result Symblance is vague at best. There's just like one or two similar ideas.

It also doesn't help matters that when folk say there is a relation, they don't really explain their reasoning. They just end it there. For instance, take this example from Paul Seruzi's A History of Modern Computing, quote, The microprocessor phenomenon passed the PDP 11 by, even though elements of its architecture turned up in microprocessor design, especially the Motorola 6800, end quote.

The claim there is in parentheses. And here's the best part. There is a citation at the end of the sentence. Siruzi cites a conversation he had with C. Gordon Bell, an architect who worked at deck and not at Motorola. That's not exactly the strongest argument in the world. So, again, the 6,800 is a bit of a mixed bag here. But what about other early microprocessors?

Early on, we don't really get much else. Intel's chips in this period follow their own lineage. That's pretty true of all other chips on the market as well. I'm sure some are taking a few cues from larger machines, but none are wholesale implementing a PDP 11 on a chip. And there's a simple reason for that, complexity. Up to 1975, microprocessors use 4,000 or fewer transistors. That's not a whole lot of power to work with.

As a result, they're pretty simple affairs. Like Lavelle said, they may have wanted a PDP 11 on a chip, but it simply was too complicated for the technology of the day. It wasn't realistic. We can see this bear out as we reach the end of the 1970s. In 1979, two important processors come out, the Motorola's 68,000, and the Intel 8086. Both have large instruction sets full of very complex instructions.

They use memory mapped I.O. They have lots and lots of general purpose registers that are truly general purpose and even move instructions. The resemblances here are clear. In the case of the 68,000, we have oral histories where designers start up by saying they were inspired by the PDP 11. The quotes are kind of rambly, so I'm not going to repeat them here. You can find them, though.

What's weird with the 8086 is I don't have a smoking gun for the connection. I just know the chip looks very, very, very similar to the PDP 11, with some weird legacy features. This is especially weird to me because the 8086 was kind of designed by a single guy at Intel. Recall that it was meant as a quick and dirty job to retain customers while another project wrapped. But there's nothing that says, oh yeah, I made that chip like the PDP 11 because I like the PDP 11. It just kind of is that way. The larger meta reason for these 11 likes is, again, complexity. The 8086 uses 29,000 transistors.

The 68,000 used, and get this, 68,000 transistors. Pretty good naming convention, huh? The PDP 11 has a whole lot of different hardware implementations over the years, but eventually the first LSI 11s used 23,000 transistors. What I'm getting at is that once microprocessors hit a certain level of complexity, they're actually able to take design ideas from the PDP 11 and implement them. Prior to that, we see something more like design hints, I guess. Putting this all together, we can answer one of our first questions.

The LSI 11 isn't really spurred on by the development of PDP 11 like chips. That's not a contributing factor worth examining, because there weren't PDP-11-like chips earlier than the LSI 11. So then, what did Inspire Deck to make the LSI 11? The PDP 11 was a machine immune to obsolescence. So goes the ad copy. This is also, broadly speaking, true. The PDP 11 would survive on for an impressively long time, in part because Deck was always willing to adapt

the machine to new technologies. This was possible, in large part, because of the Unibus. It was also helped along by the simple fact of the family. PDP11s only had to be software compatible. Anything that can run a PDP11 binary program is, by deck definition, a PDP11. That means that on the hardware level, there is a wild amount of flexibility. This can be seen almost immediately. The PDP 1120, the first machine to market, is all TTL logic. That means all the computer's logic is described in physical circuits. Every PDP11 model that followed used microcode.

Hardwired logic is replaced with a more simplified circuit that can read software, defined instructions. The PDP 11 is very much a software machine. Its instruction set and architecture is designed to be slick for programmers. Its I.O. is built in such a way that it's easy to control a software. And outside of the 1120, these computers are defined by software. This was, at least in part, intentional. To quote, see Gordon Bell in the PDP 11 family after three generations, quote,

Because of the many pressures on the design, the planning was asynchronous and diffuse. Development was distributed throughout the company. This sort of decentralized design organization provides a system of checks and balances, but often at the expense of perfect hardware compatibility. This compatibility can hopefully be provided in the software and at lower cost to the user, end quote. Each machine in the family was designed using whatever hardware was made, most reasonable, whatever its designers wanted to work with. High-end machines got cutting-edge parts. Cheaper machines got cost-cut parts. All the differences of hardware were ironed out

in software. This came in the form of microcode, drivers, or operating system patches. Deck viewed the PDP 11 as a full range of computers. This ranged from high-end to mid-tier and low-end. But around 74, they started contemplating something new, a low, low-end computer. Initial memos identified this as a, quote, computer on a board, as in a PDP 11 that could live on a single circuit board. That idea is very close to a microcomputer. The core concept is to make a computer that's as cheap as possible. A difference here is the compatibility aspect. This new single-board computer had to be a PDP-11.

It had to be compatible. That meant that Deck couldn't use off-the-shelf components. Recall, there was no microprocessor out there as of yet that was compatible with the PDP-11. This is where we get to a bit of an issue. Deck was very much a component kind of house. Their first products were laboratory modules, simple circuit boards populated with semiconductor components.

Deck then turned around and used those components to build their own computers. Deck would keep up a certain level of vertical integration throughout their history, but that would only go so deep. Deck's integration would stop at around the silicon level. They didn't make their own semiconductors or integrated circuits. That had worked out just fine for other PDP11 systems. Those used custom circuit boards with off-the-shelf ICs soldered in. But to make an entire PDP-11 on a board, they would need much smaller components.

They would need custom ICs. Before this, deck turned to Western Digital. Now, I know whenever I hear Western Digital or WD, I think about hard drives. But back in the day, WD was a Simco. They made calculator chips as well as some custom work. Why specifically work with Western Digital on this project? Well, it turns out that was actually a question at the time. It was asked in a 1974 article in Electronics Magazine.

Andrew Knowles, who helped manage the LSI 11 project on the deck side, gave two reasons. I'm paraphrasing here because the Electronics Magazine article is kind of poorly worded, But here's what the answer was. First, Western Digital could fabricate in-channel MOS integrated circuits. For the time, those were the fastest type of integrated circuits. So the technology was right. And second, the team at WD was familiar with the PDP-11 architecture. The documents around that particular claim aren't forthcoming here,

but I'm betting that familiarity is due to the fact that the PDP 11 was everywhere by 1974. WD probably had at least one in their main office. The contract details here are actually pretty interesting. Deck paid WD $6.3 million to deliver a custom chip set. Deck would get to use those chips for the LSI 11, and Western Digital would be allowed to sell that chip. to other vendors. How exactly does that work, though?

Is Western Digital going to be selling PDP-11s on chips to anyone? Well, no, not at all. And that's the neat part. The chipset that Western Digital produced isn't really a PDP-11. Instead, it's a microcode engine. The chip set is more accurately called the MCP-1600, the multi-chip processor. Again, this isn't a single chip.

Instead, we get three. One is a control chip, which is where microcode gets executed. Another is a combination of register space and arithmetic logic circuits. And the final is the all-powerful mic ROM. It's a ROM where microcode is stored. That's all it is. Together, that's all you need to make a processor, just broken up into a few different chips. different chips. This wasn't a really uncommon setup for the period. The Intel 4-0-4 was the first

single-chip microprocessor on the market. That wasn't released until 71. But crucially, the 4-0-4 is a very, very simple processor. It's 4-bit and not even sophisticated enough to use microcode. I mean, the thing only had 2,300 transistors. Prior to the 4-0-4, there was multi-chip processors. Now, of course, the language here gets a little messy since the term microprocessor was just being created in the period. One of these multi-chip systems, the AL-1, was released all the way back in 1969. So Western Digital isn't breaking any ground here. And remember, Deck doesn't even necessarily want a single-chip processor. The All-Electronic Corp is trying to optimize for

board cost. That includes, of course, cost of development. Trying to cram an entire PDP 11 into a single chip using 1974 technology that may not have even been possible. If it was, it would

probably take some huge leaps and bounds, which would be expensive. There's also the matter of form factor to consider. This is something that always sounds kind of funny to me, but it does come up quite often in older documents. You see, there are some practical limits to how you can package integrated circuits. In the 70s, circuits were all packed using through-hole

components, and usually as dual in-line packages. If you think of the classic little black microchip with the two rows of teeth, then you're thinking of a dual in-line package, a dip chip, if you will. In that form factor, there is a limit to how many pins you can have before things get, well, a little ridiculous. You can't have a 200-pin dip chip. The LSI 11 ended up using three chips, and each of those were 40-pin packages. Those were pretty big packages for the time. Now, if you had compressed everything down to a single chip, you'd probably be able to drop some pins, but still you'd end up with some weird unwieldy size that you probably couldn't buy off-the-shelf packing for. It would be difficult to mass produce or maybe even need custom

machines to manufacture. Again, that would blow up the price. There's another level of savvy going on here that I really dig. So, listener, if you will, dig this. Western Digital designed the chips, the microcode engine, the ROM, the combination ALU register. So then who wrote the microcode? Well, deck. That was the division of labor. And recall that deck was already making microcoded PDP 11s. Every model, except for the very first PDP 11, used microcode. The only difference with the LSI 11 is that Deck isn't making all of the hardware. From

just a cost perspective, that's really cool. Deck would play to their strong suit and use a little money to make up for where they were lacking internally. This also explains how WD was able to sell the same chipset

without selling a PDP11. The key difference was microcode. Switch out the microm and you could have an LSI 11, or, well, you could have something vastly different. Western Digital would actually sell machines that ran Pascal on this same chip set. This entire setup is actually strikingly similar to something going on at IBM in this exact same time period.

Have you ever heard of the IBM Palm processor? It stands for Put All Logic in Microcode. It's another microcode engine developed in the middle of the 1970s. It was a chip that read microcode and then executed it. Specifically, IBM used a palm in their 5100. It was a small portable computer, big quotes around portable, that could run System 360 code. It's also a computer that was supposedly so sophisticated,

a time traveler from the far future came back to procure one. If you know, you know. So we have a small microcode machine that acts like a larger and older computer. The only real difference here is that IBM, with its vast integration, made the palm chip in-house. Why do I bring up the similarity here? Well, there may have been some weird microcode trend going on here. But more importantly, I think this shows how the PDP 11 really was. keeping up with the times.

Work being done at Deck is mirroring work done at IBM. Deck is staying on the cutting edge, or at least they're really close to it. In 1975, Deck would use this new chip set. That year, they released the PDP 11-03, the first LSI-11 power machine.

It could run any PDP-11 software, but with a catch. The 11-03 didn't use the unibus. And that's right. Perhaps the ultimate expression of the PDP 11's power. This new microcode-based LSI computer has no unibus. On the surface, that sounds like a huge deal, right? The unibus was one of the reasons Decc was able to say the PDP 11 was immune to obsolescence.

But fear not, because the LSI 11 did have a bus. This was initially called the LSI 11 bus that's serviceable but not the best name. It's better known as the Q Bus. Here's the funny thing. The Q Bus is essentially a cost-reduced Unibus. It uses fewer pins and drops some of the more sophisticated signaling, but it does the same thing as Unibus. It can sound like the Q-Bus is a departure from some core standard, but that's not exactly the case.

The unibus was a standard with big quotes around it. Some PDP 11s used a pretty vanilla unibus. Higher Performance 11s used a variation of the unibus, and these cost-reduced LSI machines, well, they used a cost-reduced version of that bus. That should tell us something important. The anti-obsolescence powers of the PDP 11 didn't mean that you would buy an 11's chassis in 1970 and use it all the way up into the 90s, just replacing cards as needed.

Rather, it meant you could always buy a top-of-the-line PDP 11 from deck. The Electronic Corp would always be able to create a PDP 11 that would keep up with competing machines. that chassis you bought in 1970 and the new 11 you bought in 1990 would look totally different, they'd be built totally different, and they'd be totally incompatible on a hardware level. However, you could run the same exact software. So how much did this new bottom-of-the-line PDP-11 cost? Well, according to computer world, introductory price was $2,400 for a model with 4K of RAM. That inflates up to just over 15,000 in 2025 money. Comparisons here are, again,

kind of hard to make, because the 11 was pretty unique in this period. The IBM 5100, another microcode machine, cost wild, wild amounts of money, and was a portable. So that's only similar in very broad terms, and only really if we're talking about underlying hardware technology. Comparison here is difficult, so let me do something unorthodox. You might not expect this. The Altair 8800 is released in 1974. That's around the same time as the first LSI 11 machines. It's the first computer that a hobbyist can realistically buy pre-assembled. At least That's its claim to fame. It actually turns out that some nerds were doing group buys of PDP 1103s and 76,

which makes me suspect some other weird deals may have been going on, but that's an avenue that I do not have the brain power to investigate right now. Anyway, a stock and assembled Altair 8800 cost $621 in 1974 money. but that's not really a usable machine. The stock model had no serial or parallel interface and only came with 256 bytes of RAM. That's bytes with a B, no kilobytes, no megabytes, just bytes. That's enough for a toy, but not enough to even run basic.

To fully trick out an 8800 with 4K of RAM and with usable, I.O., you had to get a bunch of expansion cards. The Altair was built around a bus, similar to another computer we may know of. Now, to fully trick out the machine, it would cost you about $1,500, or $10,000 after inflation. Now I know there's a mountain of difference between those two costs. but I bring this up as a comparison because at the time there were some hobbyists actually comparing the two machines. The cost difference here ends up being $1K in 1970s.

Again, significant, but hobbyists were already shelling out a lot of money for computer hardware. This is bleeding edge technology. For the extra money, you end up with a machine that's totally complete and has a library of software ready to go. Which brings us, I think, to the most interesting question of all. Where exactly does the LSI 11 show up? Early computer hobbyists were a fascinating breed.

They were the classic early adopters. Hobbies were willing to put up with all kinds of garbage to get a computer to work. The Altair is the perfect example of this ethos. A machine with 256 bytes of memory is barely useful. But groups like the Homebrew Computer Club were absolutely a gog with the device. Members figured out all kinds of wild things to do with stock altars. And when that got old, they figured out how to expand the machine. That kind of down-and-scrappy attitude is, to me, a very hardware mentality.

to get an early home computer up and running, and to make it useful, you had to be comfortable messing around with hardware. But hardware is only part of the digital equation. To me, it's always the how of software. How do you make a program run? Why, with hardware, of course. And in the case of machines like the Altair and its contemporaries, there were a lot of caveats to that how. This is where the LSI 11 becomes truly fascinating to me. In 1976, Bight ran an article about the new PDP-11-03, the first LSI-11 machine. Medium is just as important as message.

Byte's tagline was the Small Systems Journal. It was founded in 75. Think about that for a minute. And then think about the PDP 11. In 1970, the PDP 11 was meant to be a small machine. As the microprocessor hit the scene, the small side of computing shrunk significantly. When byte is founded, small systems include things like the Altair 8800 or the soon-to-be Sol 20 or Kim 1. These are machines the size of toys that, with a little extra work, can become usable computers.

As the small side shrunk, the PDP 11 was no longer a small computer. It was a mini-computer, after all. That's something a business or an office buys. But the 11, well, it's a chameleon. It can adapt and overcome anything. With the release of the 1103 and, importantly, the LSI 11 inside, the PDP 11 can suddenly be a microcomputer. It can compete on a different stage.

And so it ends up getting covered in byte next to machines like the Selby and the Apple One. That to me is wild that you could pick up a magazine announcing the Apple One and see benchmarks a few pages away about a PDP11 model. Okay, so getting back on track, the Bight article from 76. The teaser for the article opens like this, quote, how would you like a PDP 11 slash 40 in your basement computer room? The price would possibly be too high for the typical amateur. But Digital Equipment Corporation also makes the LSI 11, a microcomputer which implements the PDP 1140 instruction set

and inherits a wealth of existing PDP11 software. End quote. The article itself isn't anything new to us. It describes the PDP 11's architecture, the wonders of the LSI 11, and all the gritty details about the machine. It comes complete with glossy photos of the computer. The interesting part is at the end of the article. To quote, For those interested in purchasing the board version of this computer,

the Southern California Computer Society is organizing a group purchase for amateurs. This purchase will involve an original equipment manufacturer quantity of 50 or more machines on a basis of cost plus 2% minimum contribution to SCCS, end quote. So we have this user group in SoCal that was organizing a group buy of PDP 1103s. Why would you do this? Well, simple. to save money and to make a PDP 11 accessible. Deck offered bulk rates and even had discount programs for OEMs,

but you had to have a minimum order. This is the kind of thing that, I don't know, I've talked to my friends about doing this before. I think a lot of us have. Oh, if we sign an OEM contract and get a big purchase order, we can save a lot of money, and then this server hardware is actually the same. cost as a desktop. It's always something you talk about, but never do. But this scheme

actually worked. Therein lies, I think, the difference between early hobbyists and today's hobbyists. You get all these stories where the scheming pays off. This is also one of those fun times where one article in a magazine is followed up in another zine. Later that year, Dr. Dobbs' Journal of Calistinics and Orthodontia ran a letter from one of the hobbyists that was part of the PDP11 group by. As, on aside, Dr. Dobbs is a great old zine. I have a pile of them in my magazine shelf.

The name is a joke about the pain of early home computers. The first issue has the tagline, Running Light Without Overbite. I assure you, it's very funny in print. Anyway, That letter in Dr. Dobbs explains how the group buying worked out. The Southern California Computer Society got a deal with a subsidiary of SDC, the System Development Corporation. According to the letter, SDC was buying a pile of OEM hardware from deck and either had extras lying around or was persuaded to order a few more boards.

SCCS snuck in on SDC's bulk rates. That let a number of hobbyists get their hands on PDP11 hardware. How much did that all cost? Well, according to this one letter, a system cost $1,000. That's a really good discount from the whole scheme. That also puts us in the realm of a fully tricked out Altair. In exchange, the hobbyists got a fully complete, computer. That takes all the pain out of home computing, to quote the letter itself.

Quote, for about 1K, you get a processor, 8 kilobytes of memory, and a serial I.O. card, and a backplane fully assembled to industrial standards that works when you plug it in. It took me about 15 minutes to go from box to teletype, end quote. That's about as easy to set up as an Altair. In fact, it might be a little bit easier. You also get a much more powerful and much more useful machine. You instantly have access to Dex Software Library. In 1976, user groups were forming around home computers, but the PDP 11 already had long-standing groups. You could hook up with Deccas, which had been around since the literal 1950s. You had access to software, hardware, and practical know-how from decades of experience.

In the biz, we call this a turnkey solution. End users normally don't have access to turnkey solutions. It's when you pay a company, you get a product, and you just have to turn the key and it works. No jumping through hoops, no pulling teeth, you get access to support, you get access to anything to make it easier for you. I also find this little chapter hilarious because it shows how scratch. early hobbyists were, just in a different way than usual.

The PDP 11 nerds wouldn't have to mess around with hardware. They would be ready to jump into the realm of software in, perhaps, 15 minutes. From here, we can start to venture out into truly bizarre waters. Now, DEC would make a number of models using the LSI 11, and the LSI 11 chipset itself would be refined. Over the next few years, revisions were released, and the chip set would even shrink down to a single chip. This is also the era when we start to see the influence of the PDP 11 at play. The 68,000 and the 8086 both come out of the end of the 70s.

Those chips would bring the design of the PDP 11 to microcomputers for decades to come. And therein lies a question. Deck was able to make a microprocessor version of the PDP 11. Where were all the 11-based microcomputers? I mean, this has a simple answer. By the early 80s, when these new 11-inspired chips are ascendant, the PDP 11 is a pretty old design. The 68,000 and 8086 are also mired in legacy, especially in Intel's case. The key difference, however, is that these new chips aren't exact copies of the PDP 11.

They're inspired by the earlier computer. They have a lot of room to make sweeping changes and to be more modern. One huge differentiating factor is address bus size. Both of these newer chips could address more memory than the PDP 11. But that's not an interesting answer. Instead, let's look at how the LSI 11 was used outside of deck. There were actual home computers that used the chipset. It's just that those machines were few and far between.

One of the better attested machines was the Heathkit H-11. You may have heard of Heathkit before. They were an electronics company that sold kits. I mean, the name kind of speaks for itself. This was originally things like oscilloscopes and power supplies and radios. In the 1970s, they started selling computer kits. The H-8 and an 80-80 machine is probably their best-known computer. That machine came out in 1977.

The very next year, 1978, the H-11 was released. This was a kit computer built around an LSI-11, and it retailed for $1,200. The ads for the machine actually mirror what we saw in the SCCS group by, quote, The world's most powerful microcomputer comes home. Computer hobbyists have always wanted the power and speed of professional machines, but they've had to settle for less. Professional machines were too expensive, not anymore. Now there's the Heathkit H-11, end quote. Ads even call out the fact that with an H-11 you have access to Deccas' software catalog. You're buying your way into an entire software ecosystem, not into painful tinkering with unproven hardware. Oddly enough, this feels a lot like a turnkey solution, so maybe I should rescind what I said earlier about end users not getting turnkey solutions.

Of course, this was still a kit computer. For your cache, you would receive a big box of parts. You would have everything you need to build an H-11 in the form of circuit boards, chips, wires, and even folded sheet steel case components. However, there was one concession. The LSI 11 came as a pre-assembled board, and therein lies the key difference between the LSI 11 and, say, the 8086. Deck in this period sold LSI11s as complete boards, not as microprocessors.

The board hooked into a cubus backplane. Compare that to the 8086, which Intel sold as a single chip, and it would just plug into a socket. Why does that difference matter? Well, Dex's approach is something of a double-edged sword. Many OEMs like HeathKit bought up LSI 11 boards. that was often a way to add a computer control to, say, an industrial CNC machine or some other big device that you want to computerize. Once you have an LSI 11 board, you have to connect it up to a queue bus and throw in some

memory. Now you have a PDP 11, which is good, because now you have a PDP 11. Who doesn't want a PDP 11? The downside is now you have a PDP 11. You don't really have flexibility with your system design. And as I mentioned, the 11 was starting to show its age. One of its biggest issues was that small address bus. It just couldn't handle as much memory as contemporary microprocessors.

These LSI 11 boards also missed the mark on price. I don't have exact OEM pricing, but we can assume it was in the high hundreds of dollars. In 1978, you could get an 8086 for a staggering $80.80. For that price, an OEM gets more power, more memory, and more design flexibility. An 8086 can be at the heart of basically any computer you want. It ends up becoming synonymous with the IBM PC, but there's actually quite a bit of design diversity prior to the PC. Xerox, for instance, used a number of 8086es, in a portable graphics workstation.

That's a kind of flexibility that just wasn't quite as possible with an LSI 11. As a result, you get machines like the H-11, which is literally just a PDP-11 with some solder joints you have to do on your own. We're still, by the 80s, the LSI 11 isn't giving as much oomph as it used to. In 1984, Bight ran an article that was full of benchmarks for different, Unix platforms. This included a pile of different flavors of Unix and also different computers that were capable of running Unix.

For us today, I want to look at three crucial machines. The PDP 1123, the IBMXT, and the Sun 2 slash 120. The PDP 1123 is an LSI 11 machine. The IBMXT was just a PC with a hard drive, and the Sun 2 was a Motorola 16. 68,000 workstation. This represents the old guard of Unix and two newcomers. All these computers used somewhat similar processors. At least, we can say that the 80-88 in the XT and the 68,000 in the Sun machine were both very heavily influenced by the PDP 11. Better still, these newcomers were formidable. The comparison to the IBM machine is closer to a wash. The PDP 1123

it wins some and the XT wins others, but margins are close. The Sun 2, however, beats the pants off the deck computer. In one of the benchmarks, the Sun 2 is actually 20 seconds faster than the PDP 11. It's no contest. The 68,000 is just more powerful than an LSI 11, and that's before you even take memory space into consideration. But that doesn't actually spell doom for the 11. Remember, it's a family of computers. It just happens at the lower end of the family. Well, it was a little low.

The 1170, a larger member of the family, is still hanging out in these benchmarks. It actually outperforms the micros. There's also a machine that we haven't even touched on yet. You see, Deck wasn't just planning on the PDP 11 dying out. Instead, they expanded the machine. By the end of the 70s, DEC had already introduced the 11's successor, the Vax.

This updated the PDP 11's design and expanded to 32-bit. But Vax was on the upper end of things, and will be a tail for another time. Down on the micro side, things were increasingly competitive. Despite its fantastic design, the 11 just couldn't keep up in that part of the market. Over the coming years, the LSI 11, would continue to see improvements.

Single-chip versions would hit the scene, but the chip would never be as successful as earlier PDP-11s. It would show up in a number of special-purpose machines and some cost-reduced models, but it served as something of a sideshow. It's this interplay between the PDP-11 and later chips that feels really neat to me. The 68,0886 showed that the PDP-11's design

could make a killing as a microcompetitive. Dex's first-party attempts were successful, but were never aimed directly at hobbyists in the same way that other chips would be. Despite that, a good number of home users did end up rocking PDP-11s. All right, we've made it through the midpoint of the PDP-11 saga. So where does this leave us? We've seen the genesis of the PDP 11 and how it adapted to one of the largest changes since sliced bread.

We've also seen how it inspired some very important processors. But there's a strange afterlife to these smaller end 11s that we haven't gotten into. Next episode, we're going even further into the future. We'll be looking at the odd places that the PDP 11 snuck into. That will include microcomputers, video game consoles, and even the Soviet Union. Until then, thank you so much for listening to Advent of Computing. You can find everything over at advent ofcomputing.com. And as always, have a great rest of your day.