Advent of Computing - Episode 126 - IBM Compatible (No, Not Those)

Episode Date: February 19, 2024

This episode wraps up the System/360 trilogy by taking things back to where they started for me. We will be looking at System/360 clones, how they could exist, why they existed, and why IBM didn't crush them. We close with a discussion of how these earlier clones impact our understanding of the IBM PC story. The truth is, by 1981 IBM was no stranger to clones. This is the culmination of a wild story, so prepare!   Selected Sources:   https://archive.org/details/iclbusinesstechn0000camp/mode/1up - ICL: A Business and Technical History   https://archive.org/details/impactreportamdaunse/page/1/mode/1up - Impact Report by INPUT   https://www.stayforever.de/ibm-pc-a-conversation-with-dr-david-bradley/

The history of IBM is, to be fully honest, kind of wild. I've hit a point in the research madness where some of IBM's shine is starting to rub off on me. Once you're deep down in this hole, you understand that IBM isn't just a computer company. It's a lot closer to an institution. Don't believe me? Well, this might help. Did you know that IBM published a book of devotional songs? It's called The Songs of IBM, or sometimes just the IBM Songbook, depending on the edition. These were actually sung at corporate get-togethers and events. The book was first published, I think, around 1930, with revised

editions printed as late as 1940. At least, those are the bounds I can find from scans online. The book is about 30 pages of music. Here's just the lyrics to the chorus of the most catchy tune in the collection, Ever Onward, IBM. Ever onward, ever onward. That's the spirit that has brought us fame. We're big, but bigger we will be. We can't fail, for all can see that to serve humanity has been our aim. Our products now are known in every zone. Our reputation sparkles like a gem.

We've fought our way through, and new fields we'll conquer too. Forever onward, IBM! Or should I say, forever onward, IBM! IBM is a giant force in the history of computing. They existed before the first digital computers. The consequences of decisions made inside IBM still shape the digital landscape today. What I'm trying to get to is that it's weird to see blemishes on IBM's record. It's weird to see times that Big Blue didn't get the last laugh. That's one of the reasons that the story of the PC is so compelling. It's a time that IBM lost. It's a story about smaller companies treading in the domain of giants. But it isn't the first time that this would happen.

Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 126, IBM Compatible. And, no, I don't mean that in the usual sense. No, today, we're completing my System 360 trilogy by going back to where this entire descent started. Allow me to elaborate. Back at the start of the year, I decided to work up a fun, light-hearted episode about the RCA Spectra 70 series, a family of computers that were inexplicably compatible with the IBM 360. As my research began, I ran into something that I didn't expect.

More 360 clones. IBM isn't exactly known for their leniency with clone manufacturers. That's actually a reason that the IBM PC was looked at as so unique for IBM. That computer ends up being cloned because its design made it possible for a third party to do so without violating IBM's intellectual property. The usual story is that the PC was the first time this happened to IBM, that it was very unique. But the Spectra 70, and a number of other computers, seem to say otherwise. The main questions over this trilogy have been how these earlier IBM clones existed, how they survived, and how they changed the story of the IBM PC. But to get to that,

we've had to take a pretty long route. To talk about 360 clones, we first had to talk about the 360 itself. Then we had to look at USVIBM, an antitrust case that bloomed out of the success of the 360 and, in large part, created an environment for clones. Today, finally, we're coming to the clones themselves. So yeah, it's been over a month, but now I get my fun, lighthearted episode about weird clone computers. This episode, I'll be focusing on two companies in particular, RCA and Amdahl Corporation. There were a few other clone manufacturers, but I think this will give us a good outline of the field. RCA is notable as the first of the cloners, and Amdahl is notable because of its founder. To close out the episode and the trilogy, we're going to look at the million-dollar question.

This is something that's been in the back of my head for quite a while now. How does this change the story of the IBM PC? What does this context do to this very well-known legend, to really one of the biggest stories in computing history? We have a lot to cover, but this will be a little more fun than last episode, so let's get started. I'm not a very patient man, so I'm going to start this episode with what got me sucked into this rabbit hole to begin with.

That's RCA, the Radio Corporation of America. RCA was actually a very early entrant into the computer market. They built their first computer, BizMac, in 1956. That puts them squarely in the same realm as better-known players like IBM and Univac. During this early era, the 50s and 60s, RCA was very much a player. They were one of the seven dwarves that fought over the market share that IBM didn't capture. This doesn't mean that RCA had a straightforward relationship with computing. There are some twists and turns, even right off the beginning. computing. There are some twists and turns, even right at the beginning. First of all, RCA didn't stay in the computer business indefinitely. Their initial computer division was sold and dissolved in 1971. It was restarted in 1972 with a new focus on home computers.

That second computing division is what I covered in my Cosmac episode late last year. What you should be seeing is that RCA's corporate history is a little strange. And it gets stranger. So check this out. One of my sources for this segment is a book called ICL, A Business and Technical History. The relation to RCA is a little tenuous, but you see, there isn't a book on the history of RCA computers. Back in the 1960s, RCA had business dealings with this British company called the International Computers and Tabulators Company, or ICT.

That company was folded into another company called ICL, or International Computers Limited, in 1968. So this corporate history of ICL just happens to tie into the history of RCA computers. Once again, there's some annoying stuff going on here. But complications aside, let me spin a yarn for you. In 1961, ICT had bought and imported a number of RCA machines. These machines were then relabeled and sold as ICT computers. It's a bit of a classic gambit in the tech world. Anyway, this had opened up communications between the two companies. Now, this is where things get interesting. Apparently, IBM wasn't the only company thinking about compatible families of machines. ICT had recently bought up and merged with a number of computer companies.

Buoyed by that influx of power, they were planning to release a line of compatible computers. These were early plans, mind you, but there's still an intent there. At the same time, RCA had been devising very similar plans. So here's the wild twist, as told by the ICL book. In 1964, RCA reached out to ICT to see if the British company would be interested in helping out with this whole computer family thing. As far as I can tell, RCA had no idea that ICT was on the same tip. And neither company knew that IBM was also planning the same thing. The date here is crucial. This conversation happened in early 1964. RCA invited ICT engineers over to the US to hash out the plans, and that's when things got really wacky. To quote,

As luck would have it, the System 360 announcement of 7 April 1964 occurred at the very moment of the ICT visit. There was no industrial espionage, and RCA obtained details of the 360 from the publicly available manuals. While the ICT team toured the United States on other business, RCA immediately investigated the implications of System 360 for the RCA range. When the ICT team returned a week later, RCA had decided to make the new line fully 360 compatible. The new RCA range was subsequently announced as Spectra 70. End quote.

Two big things that I want to highlight here. Besides, you know, the whole chain of events here just being very improbable and kind of, frankly, funny. The whole chain of events here just being very improbable and kind of, frankly, funny. Thing one is that IBM was not the only company thinking about compatibility. I think we have enough samples here to assume a pattern. In the early 60s, maybe call it the late 50s if you want to add some lag time, computer manufacturers were starting to get concerned with compatibility.

The reason for this comes down to sheer complexity. That's what IBM was facing and what led to the 360. RCA seems to have held the same sentiment. At the 1964 Fall Joint Computer Conference, David Sarnoff, board chairman of RCA, had this to say. This comes to us via an article in Electronic Age. There are eight computer word lengths in use. There are hundreds of character codes in being, at a ratio of one code for every two machines marketed. Four magnetic tape sizes are employed with at least 50 different tape tracks and codes. End quote. Sarnoff is talking specifically about the complexity of incompatible software, While IBM's public approach also had a lot of hardware-based arguments in it, but that's inconsequential. The core problem expressed here is that there were simply too many incompatible machines and incompatible systems in the world.

There were too many languages. That had all these ethereal impacts on progress in theory and concrete impacts on profits. I think the argument could be made, and I'm going to make it, that by 1960 the computer world was becoming too complex. The spread of incompatible machines and incompatible languages was leading to something like the information problem. There is just too much to sort through. That could, in theory, lead to a total stall in new developments. That stall could either come from pure market forces, or from research and development running up against an impenetrable wall of prior arts. The market needed to collapse a little bit in order to progress.

The wagons needed to circle around fewer languages and fewer kinds of machines. I think that's why companies were starting to flirt with compatibility. This leads to the interesting implication that, at least in theory, a clone computer could be made for very altruistic reasons. Anyway, that's thing one. Thing two is that clone piece, that RCA decided to follow IBM's lead and follow it very closely. This gets a little tricky, so let me tread carefully. On December 4th of 1964, RCA announces the Spectra 70 series. This is just about nine months after the huge fanfare of IBM's 360 announcement.

RCA's press release says nothing about IBM. It says that Spectra 70 will be a family of compatible computers. But it does have this interesting line tucked away. This is in a list of big, cool features of the new family. To quote, an advanced multilingual capability enabling Spectra 70 computers to speak all the most common programming, data, and communications languages, including those of the most recently announced computer systems. End quote. Given the context, and the fact that a lot of people still programmed in assembly language, I think it's very easy to read this as RCA announcing a 360-compatible computer, or at least a computer that can use the same programming languages and same data and

communicating languages as the 360. Once again, this is couched in terms of programming languages, in terms of software, but come on, it's right there. The most recent big announcement, the most recently announced computer system, was the System 360. But does that mean that the Spectra 70 was just a ripoff of IBM's big new machine? The simple answer is no. If that were the case, then RCA would have been destroyed by Big Blue legally and lawfully. The fact is, RCA built a new family of computers using new technology that, under certain circumstances, could run IBM software. The first thing to point out here is that some Spectra 70s, at least the fastest and best, used integrated circuits. Just some quick brass tacks here. The first

Spectra to use integrated circuits was the 70-45. That was announced in 1964 and shipped in 1966. That was announced in 1964 and shipped in 1966. There was also the Model 35, which was integrated and shipped in 1966, but that was announced in 1965. This would make the Model 45 the first commercial computer to ship and be announced with integrated circuits inside. That's a huge leap forward from previous technology. This is also a feat that IBM couldn't really match. I've seen some arguments that the System 360 Model 91 was actually the first computer to use ICs, but that's just not true. That machine languished in a lab for years, only shipping in the latter part of 1967. The 360 series did use these STL components, which were

something like very primitive low-density integrated circuits. And here I mean very primitive. We're talking about just very small components on little glass wafers. In 1966, RCA was shipping computers with honest-to-goodness lithography. RCA was able to beat IBM here because, well, RCA was more of an IC shop than IBM. They had been on the forefront of the development of ICs. In fact, the first MOS integrated circuit was developed in a lab at RCA. So they were ahead of the game. They had the know-how and the manufacturing capacity to use ICs more aggressively, whereas IBM just wasn't quite there yet. That alone made the Spectra 70 a whole different kind of computer than the System 360.

In a lot of the marketing material, RCA calls the Spectra 70 the first, quote, marketing material, RCA calls the Spectre 70 the first, quote, third-generation computer. And while that is kind of a gimmick, it is largely true. RCA was the first to do ICs on this kind of scale in a commercial machine. The compatibility, the clone-like aspect, only really comes in terms of software. You weren't going to be plugging an IBM tape drive into an RCA computer. At least, I haven't seen any press about that practice. Rather, you'd be running select 360 software. This was possible because the Spectra 70 had the same rough architecture as a System 360. Now, I do say rough here. The RCA machine offered most of the instructions that IBM's newest computer had. But there's two big caveats.

The first is that the Spectre 70 ran its own RCA-made operating system. This matters because, normally, a program relies on certain operating system features to run. A program can call out to an operating system to, for instance, allocate memory or load a file from the disk drive. That, right off the bat, made for a significant difference. It meant you couldn't actually just drop a binary from a 360 on a Spectra and go. The second caveat is that the Spectra didn't implement the 360's privileged instructions. Once again, a bit technical, but bear with me. On the 360 architecture, there are so-called unprivileged and privileged instructions.

Unprivileged instructions are your basic bread-and-butter things, like adding numbers and moving around data. It's the stuff you use most often. Privileged instructions are for more dangerous things, like input-output, managing memory protection, or halting the computer. Those operations can only be used by the operating system, by OS 360. So, really, these two caveats tie together. RCA computers can't run OS 360 because they don't have IBM's special privileged instructions. So, we must ask, how was a Spectra 70 really useful? Well, you could run IBM application software on a Spectra. The actual details here are a little shaky. Officially, the Spectra 70 supported languages. That was the selling point. You could program a Spectra in

Fortran or COBOL, or in Spectra 70 assembly language that would have looked very suspiciously similar to IBM's own assembly language. Application-wise, though, I think you'd have to recompile programs. This would be due to the operating system differences. You'd at least have to relink programs, since RCA's operating system's interfaces would have been different than the ones used by IBM's operating system. But the documentation doesn't explicitly say how to run IBM software on your Spectra, and I think that's partly just to keep RCA a little bit safe. Software wasn't the only type of compatibility on the system. Since the Spectra was a 360 clone, that meant it had to speak the same language,

and I don't mean programming language. Rather, I'm talking encoding. Spectras used 8-bit bytes, a standard taken from the 360. Spectra's also used the venerable EBCDIC character encoding, once again borrowed from IBM. You could even use IBM tapes in RCA tape drives. What you end up with is a machine that can't exactly replace a 360, but can be used instead of IBM hardware. It can do basically everything a 360 does, but a little bit different. You can use the same software, mostly, and the same data, mostly. This, I think, would be the best value proposition for a company that has a lot of in-house software. That would already be written in Fortran or COBOL or assembly language, you'd already

have the source code, you could just recompile it and save all those software development costs. What we're dealing with is kind of a quasi-clone, and that status is made clearly apparent in all of RCA's press and documentation. The Spectra wasn't called a clone. It was a machine built to industry standards. It had, quote, industry-wide compatibility. I haven't seen a shred of RCA-labeled documentation that says the letters IBM.

It also seems like a lot of RCA's core software was written in-house. Spectra's ran their own operating system, and RCA even had their own Fortran compiler developed by RCA engineers. I think that's really how the Radio Corp avoided a lawsuit. They had an almost-clone, but they never sold it as a clone. Everything had a wink and a nod to IBM, but the quiet part was kept out of print. That said, the Spectra could be ran as a clone. I read an interesting thread while working on this episode where an old RCA technician was saying that some client was even able to get OS 360 to run on a Spectra. that some client was even able to get OS 360 to run on a Spectra. Now, I don't know the veracity of that story, but the bottom line is that folk knew what was up, it just wasn't shouted from the rooftops. With that, it's time to dive into something more loud. In 1970, Gene Amdahl, designer of the System 360, went rogue. This had

been a long time coming. Big Blue, in this period as much as any other, was a business. Amdahl wasn't really a business person. He was a computer kind of guy. In 1965, he became an IBM fellow, which put him into a more research-oriented position. As a fellow, Amdahl was able to choose and really follow any line of research he wanted. It was the culmination of a spectacular career at IBM, and then he started down a dangerous path. A major source I'm using in this section is the book The Legend of Amdahl by Jeffrey Rodengen. And as a bit of an aside, this was recommended to me by Charles Edge of the Computer History Podcast, so big ups to Charles. Anyway, according to The Legend of Amdahl, the crux of this issue came down to the performance versus cost ratio. To paraphrase

Rottingen, IBM in this period built on performance. A more powerful computer, read a faster 360, cost more money. That's plain and simple. It's a very reasonable approach for business, but it's also very counter-revolutionary. This means that IBM didn't have the flexibility to, say, embrace new technology. Amdahl wanted to do just that, use new technology to create faster and cheaper computers. That would undercut the 360. So, while Amdahl saw his research as blazing a new path forward, IBM saw it as a 360 killer. And IBM, at this point, was the 360. But what exactly was Amdahl proposing?

Was there some secret project that spooked IBM? Well, this is where things get tricky. Amdahl, despite being a fellow, had a bit of a strained relationship with corporate. He describes these periods where he was kept isolated from other IBMers. By 1965, he was even out in IBM's California research lab, which, despite being the new and exciting lab around the company, was on the very edge of Big Blue's empire. Near the end of the 1960s, IBM was working on a new computer design. This was, of course, provisional. At the time, IBM was doing this internal proposal system for new projects.

Teams would bring in designs, those designs would be evaluated by a board, and a winner would move forward. Amdahl and a single engineer, I can only imagine through sheer force of will, won out on one of these proposal competitions. They had worked up a newer and much faster machine, something that could seriously be a 360 killer. But that's where we reach the second issue, pricing. And this is a bit of a weird one. What we have to remember is that when we're pricing out the profitability of a computer, that includes both the actual cost per unit and the cost of development. And new technology, while eventually being much cheaper per unit, is expensive on the uptake. This is also a weird one because it all comes back to the impending US v IBM trial. Every IBM machine had to turn a profit.

They couldn't sell computers at a loss because that could easily be seen as predatory. because that could easily be seen as predatory. And IBM had been selling some computers as lost leaders in order to explicitly hurt competition. IBM had a history of using lost leaders as a way to be anti-competitive, to push around their monopoly powers, so every machine had to be profitable. That would be a problem for Amdahl. To quote, then we decided we would go and find out from pricing what would happen if we put that machine out in the field. As a note here, that machine being the first proposed new machine by Amdahl and this one engineer. To continue, well, they found it'd be a loser, like the Stretch had been, and in the Stretch case,

they had gone back in history and looked at the technology that Stretch had generated and been billed for and how it'd been used in other machines, and they were able to go back and show that Stretch finally made a minute but positive earnings, which was important from the standpoint of antitrust position. End quote. This is one of those very subtle things. Back in the 50s, Amdahl had designed Stretch, IBM's fastest computer at that point. That machine had been a loss leader, but over time, Big Blue actually turned a profit. This was partly because of that unit cost versus development cost, and partly because the technology and stretch allowed IBM to make more money down the road

in other machines. It just took a while to become profitable. This new machine may have turned out the same way. It may well have turned a profit after a few years of sales and years of refinement, but in the short term, it would be a loss. While IBM could take a hit, that would look very suspicious to the Department of Justice, especially with US v. IBM right around the corner. Remember that IBM had years of warning that the antitrust suit was coming. Amdahl would rework the computer, he actually did so twice, and neither time went well. Calculations showed that the first revision would still be a loss leader, and the second revision would only be marginally profitable. Things just weren't working out for him at IBM,

he was too restrained by the company. So in 1970, he decided to leave. This is the danger of having such talented employees. When IBM wasn't giving Amdahl the space to do what he wanted, he'd just up and quit. He would take a handful of other IBMers with him and found the Amdahl Corporation. This type of defection, while not a daily occurrence, did happen often. The usual outline of the story, or at least the traditional form, is that disgruntled employees leave to form a competitor or join up with an existing competitor and start a competing product. That's how we got companies like Fairchild Semiconductors and Intel, just to pull a few names out of my hat. Amdahl Corp would very much follow this tradition. The initial business plan was to make a better System 370. Now, this will turn into a bit of a

mess in a minute, but let's stick with the very simplest story for now. The 370 was the direct successor of the 360, and backwards compatible with the older computer. As far as I can tell, Amdahl didn't have a direct involvement with the 370 project, but the machine was meant to be compatible with his 360 design. 360s were also still very active. IBM officially discontinued the line all the way forward in 1978, so, you know, it's one of those potato-potato kind of things. If I slip into calling Amdahl's machine a 360 clone, well, you'll have to forgive me. Most literature also falls into that same trap. So, what do we mean by a better machine here?

The key metric is the so-called price-performance ratio. That's the cost of a computer compared to its power. Amdahl knew that he could beat IBM at that game. He could make a 360 that was cheaper and faster than a genuine IBM. In fact, an Amdahl computer may have been even more genuine than an IBM, since he did design the 360 after all. The trick was new technology, the integrated circuit, or to be more specific, VLSI, very large scale integration. This was the latest wave of miniaturization technology. As we've discussed earlier, integrated circuits were still pretty new. Chips existed, but were relatively low density. A single integrated circuit may have

contained hundreds to thousands of transistors. VLSI chips were just starting to appear in 1970, and they could contain tens or hundreds of thousands of transistors. This meant that complex circuits could be made faster, smaller, and cheaper than ever before. If built right, then a VLSI machine could be had for a fraction of the price of an earlier machine, and it could run multiple times faster. But that wouldn't be enough for Amdahl to compete with IBM. Recall that Big Blue was a one-stop shop. You buy IBM, you get hardware, software, and services. That's why no one was ever fired for buying IBM.

For Amdahl Corp to compete, they needed to be three companies in one. That's exactly how IBM kept new competition at bay. But this was 1970, and IBM had been forced to loosen their grip. In 1968, IBM had unbundled hardware, software, and services. Previous to that point, software and services always came free with IBM hardware. Software and services always came free with IBM hardware. But now, as Amdahl was setting up his new offices, new markets were forming. IBM software cost money now, for the first time ever. That had two major implications. Firstly, it meant that IBM customers had to pay for software.

They had to have a separate line item on their budgets and receipts. Since software was no longer free, that meant it was an open market. Folk could shop around, and that meant in a few ways. Secondly, non-IBM customers could now buy IBM software. Of course, without context, that might sound a little weird. Why would someone without a 360 buy, say, a Fortran compiler for the 360? But we know that clones already existed. The unbundling meant that a Spectre 70 lover could now buy real IBM software without raising any eyes. IBM software without raising any eyes. And just as a side note, I don't exactly know how RCA expected their customers to get IBM software before the unbundling. If someone knows, please hit me up because I'm sure there's some weird stories around that. Anyway,

returning to the main point, Amdahl's business plan was to rely on first-party IBM software. But it went deeper than just the unbundling. To pull and reuse a quote from an interview with Amdahl, quote, Business plan was to make a compatible machine. Just one year earlier, well, in July of 1969, IBM announced internally that it was putting its old software in the public domain and was going to price the new software separately. And that immediately told me that if we made a compatible

machine, IBM would, if they were not going to violate antitrust, would have to provide the software to us because they were not allowed, under antitrust law, to have tie-in sales on separately priced products." So Amdahl Corp customers would have two sources of software. Buying new software from IBM directly, or freely available older software. This meant that, on day one, an Amdahl computer would be useful. It would be ready to jump up and go, without Amdahl Corp having to write a single line of code. That lowered the barrier to entry and made competition much more possible. This doesn't, however, mean that competing with IBM would be easy.

One of the first orders of business was funding, which the legend of Amdahl turns into something of an epic. Gene himself ended up doing a bit of globetrotting to secure venture capital. Over the first few years of the 1970s, Amdahl scraped together funds from America, Japan, Germany, and the UK. One early supporter was actually another computer company. Fujitsu, a Japanese computer and semiconductor manufacturer, would invest millions in Amdahl. Over the next few years, a computer took shape, the 470-6. As for the meaning of the name, well, it seems to have been something of a joke. Amdahl Corp. publicized that the computer was a fourth-generation machine, but still

compatible with IBM's hardware. Hence, 4 for fourth-gen, and 70 to reference the 370. Development would continue until either 1973 or 1972, when Amdahl Corp was actually forced to throw out their entire design. The fuzziness around this date is due to conflicting sources. The key event here is IBM's announcement that, going forward, System 370s would support virtual memory. That feature in itself is a bit of a big topic. It basically just means that newer 370s could play some more memory tricks, and that helped with multitasking and timesharing. Let's leave it at that and avoid another tangent. The announcement was made in August of 1972. Now, this may not have been totally public. It was first printed in a publication called IBM News. I'm not entirely sure if this was a publicly available newsletter or just an

internal thing at IBM. I assume this was somewhat public since the scans I have are on newsprint and they don't appear to say anything about keeping secrets. There are also references to these announcements in newspapers that were very much publicly circulated in 1972. But, you know, it is IBM. They might have been playing things close to the chest and being a little bit coy. Most of the newspaper articles I've read don't give out a whole lot of detail. I'm a little cagey here because the legend of Amdahl says that the good old corp found out about IBM's switch to virtual memory in 1973. Which, maybe they did. But I don't know, I'm kind of of two minds on this one. I could see Amdahl trying to stay at arm's length from IBM in order to avoid litigation, but the IBM news

announcement just feels like something meant to be publicized. I'm not sure if the particular article was passed around, but as I said, there are news articles about IBM adding virtual memory to the 370s. Maybe Amdahl viewed those as rumors more than the truth, so I just don't know. It's a weird aside, I guess. The point is that the 470-6 was dead in the water as soon as that announcement was made. That design didn't support virtual memory, which meant that it would suck at timesharing. IBM's commitment to virtual memory meant that Big Blue in the future would have a major edge. It also meant that Amdahl computers wouldn't be compatible with the best IBM software. Amdahl computers would also just kind of be obsolete

before they ever hit the market. So that initial design was just thrown out. To support virtual memory, the 470-6 had to basically have all of its memory pathways reworked. It had to have every way it dealt with memory changed. That change was so fundamental that it made more sense to just make a new computer. That design was started in, well, 72 or 73, and it was called the 470v-6, since, the 470V-6, since, you know, virtual, you gotta love these names, right? The 470V-6, now actually competitive with IBM, would be released in 1975. The development cycle was arduous, to say the least. Amdahl had to ensure 100% IBM compatibility, faster operation, and lower costs. That's a very no-compromises kind of position. But they did it.

That's kind of the amazing part of the whole story. Tests were ran right up until release day, and they showed that the 470V would, in fact, beat the pants off any IBM hardware. would, in fact, beat the pants off any IBM hardware. Once released, reports came flooding in about how this new computer was faster than anything Big Blue could sell you. You can actually go back through Computer World's archives and see these cascading articles from universities and research labs with their own independent tests. This new machine was between 1.2 and 2 times faster than a new 370. It could run any IBM software, new or old. It could use IBM peripherals. Now, this is a bit of a weird thing, but check this out. The 470V was plug-compatible with IBM hardware. That meant you could replace a 370's processor with a box

made by Amdahl, and you could plug in all your disks and tapes and terminals to the thing. Amdahl wasn't the only plug-compatible manufacturer out there, so in theory, you could buy a 470V and hook it into a Telex tape drive, a Memorex hard drive, and a data point terminal. You could make a 370 setup and the only IBM logos might be inside some of your software. That's starting to look a lot like the PC era to me, but I'm getting ahead of myself. Once we get to the technical aspects of the 470V, well, it's, uh, it was basically an IBM mainframe. Or, rather, it's so close that technical specs are a little boring. The main differences come down to more aggressive use of integrated circuits,

and the fact that 470s were air-cooled instead of water-cooled. 370s of the era actually used pretty complex and power-hungry water-cooling systems, but cooling's one of those differences that, while important, quickly devolves into a pretty heated debate. The bottom line is, if you bought an Amdahl, you were just buying a better System 370. That's a long-winded way of saying, I'm not doing a little hardware corner. What I do want to close with is a market analysis. I know, exciting stuff.

This is, I think, very important to look at for two reasons. Antitrust and the simple fact that market analysis was crucial to Amdahl Corp. And who are we to argue with Gene himself? Once again, the source here is the legend of Amdahl. It discusses two studies carried out in 1973. These were used to guide the final stages of development. The first study covered IBM customers in general, and the second focused on 360 users. The first study found, paraphrasing from the text, that for IBM customers to switch,

the new machine needed to be either 20% cheaper or 20% faster than current hardware. No surprise there, right? In order to consider the switch, there had to be some type of incentive. This would have only confirmed what Amdahl was working towards. He thought there was a market for better IBM hardware, machines that broke the whole performance equals price paradigm. Turns out that they didn't even need to break the paradigm that hard. 20% fell well within what Amdahl was churning out. hard. 20% fell well within what Amdahl was churning out. The second study is what I find more interesting. Now, I don't have the full text of this report. I just have a cited paragraph in The Legend of Amdahl. The citation is, expectedly, stashed away somewhere that I

can't find or reach, but that's enough to glean something interesting and to really bring something into focus. Amdahl was cloning the 370, but he wasn't necessarily going after 370 users. This is a fine distinction, so check it out. The 370 was still a very new computer in 1973. was still a very new computer in 1973. The first models were just three years old. There wasn't a huge reason or desire for those customers to make the switch. The 360, however, was still a very popular platform. In fact, I'd wager good money that there were more 360 installs at that point in time than 370 installs. Those customers had been on their platform for a while. The march of progress meant that most were looking to upgrade. The study Amdahl commissioned showed that 67% of 360 users wanted to upgrade in the next few years. And here's where we loop all the way back to episode 1 of the series.

When users have the chance to upgrade, that opens up the chance for them to leave. In designing the 370, IBM was taking a risk, which they did hedge. The new family was backwards compatible, which made it easier for customers to upgrade from their aging 360s to a brand new 370. They could keep their software and just move to better hardware. But the 470V was just a 370. It could play the same backwards compatibility trick. So why bother with IBM at that point?

Amdahl was trying to sneak into that crack, to grab up those 67% of 360 users looking to upgrade. The bottom line was that the 470V, on release, was faster and cheaper than a 370, and it could run any IBM software. So there was very little reason to stick with IBM, at least very little reason to stick with IBM. At least, in theory. In practice, things would take a little bit of coaxing. IBM was a triple threat, and so far we've only covered how Amdahl dealt with two of those threats. The hardware was better in all regards. The software was literally from IBM themselves. That leaves services, and another view into IBM's monopoly powers in this period. Prospective customers were, to pull from Rodengin directly, quote,

nervous about alienating IBM by engaging an untested competitor, end quote. IBM customers were, quite literally, and this is crucial to understand, afraid of angering IBM. I think this right there is about all you need to prove that US v. IBM was justified. We have a competitor that comes along. They offer a strictly better product. And existing IBM customers were scared about what IBM would do if they made the switch. So what were those specific fears and how were they quelled? For this, I'm going off a 1976 impact report carried out by Input.

It's a survey of 19 companies that considered switching to Amdahl, 10 of which decided to commit to the new computer. This report is interesting because it not only shows how service was the main concern, but how IBM attempted to protect its accounts. Out of the surveyed companies, one common refrain was that Amdahl was new and unproven. There was a common concern that this new company wouldn't survive, that it wasn't viable, that Amdahl would buckle. And while Amdahl did provide support, that was just for the time being. If the company folded, then any support would evaporate. That's a very real fear, and it's one of the domains that Amdahl couldn't compete in. IBM had, at this point, existed for nearly 80 years, depending on how you

count it. That is some serious longevity. It's one of the reasons that IBM was viewed as a very safe company. They had survived and thrived decade after decade. Amdahl, on the other hand, was totally untested. All the genius and good design in the world can't beat IBM's rock-solid stability. The response from Amdahl was pretty smart. In addition to expected service agreements, Amdahl also shipped out spare parts with new computers. Customers received a 60-day trial period for new hardware. Both of those choices helped to build trust with new customers. It helped to make Amdahl look less risky. The final touch was actually Fujitsu. The Japanese firm, at this point, was very invested in Amdahl Corp. They had invested money, signed joint research agreements, and even sent engineers over to the

states. Better still, Fujitsu was an old company. They had weathered the same decades as IBM. So, to smooth things even further, Fujitsu agreed to back any service agreements. If Amdahl disappeared, then Fujitsu would step in to help customers. Now, I did promise some heavy-handed IBM debauchery. The input report has these two separate sections that cover how IBM reacted to customers evaluating Amdahl hardware. These are primarily composed of direct quotes from customers. These are delightful sections to read. I've pulled out the spiciest quotations to share here. So, to just hit on the highlights, quote, they didn't say it directly, but they warned us about what might happen to our hardware and software support if we left IBM.

They brought out all the big guns to talk to my boss and his boss. They had friends of mine in IBM call me to talk me out of switching, but these guys were friends of mine and refused to do it. They offered to fly me to Poughkeepsie in a private jet, but I turned them down. The local branch manager sat in my office for hours. He almost cried. They crawled all over us. They used red herrings, trying to scare us off Amdahl. They were professional, but insistent.

End quote. The sections continue in a similar manner, and honestly, this is the kind of stuff that I expect from IBM in this period. IBM viewed Amdahl as enough of a threat that they put the clamps down. But they had to be careful because of antitrust law. In some cases, though, this may have been enough. Nearly half of the companies surveyed by Input stayed with IBM. Some of those companies even acknowledged that Amdahl had a better product, but they were worried about leaving Big Blue. They worried that Amdahl was a company that wasn't viable. I can't help but imagine that IBM's personal touch had an effect here. How could it not? The net result is that Amdahl was able to capture some of the market, but not all. I've seen reports that Amdahl had between 8 and 24% share of the overall mainframe market. This is all unconfirmed stuff, but the numbers

seem reasonable. Amdahl at least did well enough to get IBM shook. There are stories of IBM customers keeping Amdahl mugs on their desks to try and get better lease agreements from IBM. There's also evidence that Amdahl forced IBM to be more competitive in the market. There's also evidence that Amdahl forced IBM to be more competitive in the market. The Input Report found that IBM planned to reduce mainframe prices by 20% in 1976. Part of that would be done by making better use of integrated circuits. In that way, Amdahl was able to push IBM around from the bottom of the market. Now that we've run the gauntlet, I think it's time to come back to the classic fable of the PC.

It goes something like this. In 1981, IBM announces the personal computer. It's unlike anything that's ever come out of Big Blue. The PC is an open platform. Its hardware and software interfaces are published and well-documented. Anyone can write software for the PC, and anyone can make expansion cards for the PC. It's also made from off-the-shelf parts. All chips are labeled and easily sourced from the outside.

The only custom parts of the PC are its circuit schematics, its case, a little bit of glue logic, and IBM logos strewn around the computer. The circuits are fully documented. You can actually buy that documentation from IBM directly. So if you have the time and some acid, you can actually etch your own PC motherboard. The case, well, it's kind of just a case. You can make something that looks close enough out of sheet steel and spray paint in your garage. The glue code, the BIOS, is just a few kilobytes of software. It's all well documented and can actually be reverse engineered pretty easily. The IBM logos are strictly non-functional. You can actually use reverse-engineered pretty easily. The IBM logos are strictly non-functional. You can actually use any logo you want and still have a working PC.

Now, IBM had gone with this open design for a number of reasons. The PC was meant to be a relatively cheap machine and a relatively quick project. Using existing parts made that possible. It was also intended to be sold and serviced by retailers, not by IBM themselves. Off-the-shelf parts meant that some random repair shop in somewhere in Iowa could fix a PC just as easily as a technician in an IBM lab in Poughkeepsie. IBM had also made the decision to pull third-party software closer than ever before. The PC would technically ship with IBM DOS, but that was just a rebrand of Microsoft Code. The only IBM software in the PC was, realistically, the BIOS. IBM also encouraged third parties to

develop hardware and expansion cards for the PC. The open architecture design encompassed all of this. It was a radical shift for IBM. When the PC launched, it was an overnight success. Units moved at a very rapid pace. But the open architecture backfired. At least, it backfired on IBM. You didn't actually need Big Blue to make a PC for you.

Companies started making clones of the computer almost immediately. By 1982, you could buy a 100% IBM PC compatible, but there is a bit of a hiccup here. These clones had to have their own BIOS code, and it had to be reverse engineered in a very particular way. If there's even a whiff that these companies stole IBM code, something that very much was copyrighted, well, the hammer would fall on them pretty hard. The first clones got around this using so-called clean room reverse engineering. This is where one team goes over IBM's docs and writes a specification. Then an isolated team writes code to that spec without looking at anything that says IBM on it. Everything is documented, all lines of communication are put down on paper,

so there's an actual paper trail for the eventual lawsuit. Thus, by 1982, there are 100% legal IBM PC compatibles. These machines are cheap and, in many cases, better than the original PC. While IBM would like the fire, it's these clones that really blow up. They can use all the same software that was written for the PC, they can use all the same expansion cards and peripherals, and they cost less than IBM's beige box. This is often spun as a totally new set of circumstances for IBM. Thus, the PC story is this perfect storm. Big Blue took a risk that got wacky. IBM made a very non-IBM computer. They were more open about their design than ever before, and that made it possible for clones to exist.

This was all happening right as microcomputers were really taking off. Now, I recognize this may seem like I'm setting up a straw man to knock over, but I assure you, that is not my intention here. This is my understanding of the PC story prior to this year. If you go back and listen to my earlier episodes, every time I talk about the PC, it's using this framework. So, I guess the straw man here is actually myself. I think the first handful of straw to pull out is the uniqueness hypothesis.

The PC itself was relatively unique for IBM. Now, this does get a little tricky. IBM had been working on a very similar off-the-shelf machine, the System23 Datamaster. You know it, you love it. This used an Intel processor, had an internal expansion bus, and used a lot of generic chips. It was developed around the same time as the PC, being announced just one month prior to the personal computer. In fact, some folk that worked on the design of the PC had just previously worked on the Datamaster.

You can almost think of it as a dry run. So while you can technically say the PC wasn't unique, I think that is a wasted argument. It was a new philosophy for a new decade. Let's just go with that. Previous computers, like the System 360, had been made from totally custom chips and circuits. RAM had been woven in IBM factories and all that jazz. Documentation that was made available was about interfaces, programming, and repair. At least, repair when it had to be by court order. So while there was some visibility into the internals of a 360, you couldn't buy an entire schematic of the machine, with the whole parts list and everything.

an entire schematic of the machine, with the whole parts list and everything. I think that's enough to make the PC something new. The earlier IBM machines also weren't sold through distributors, they were sold through IBM directly, so that's another point in favor of the PC being new and unique. That said, we should all see how the PC clone situation was not unique. We can argue about when IBM hit the first clone crisis, but I think I'd place that at 1975 with the release of Amdahl's 470V-6. That's when we hit the stage where a 360 machine room could exist without a single IBM logo. In 1975, a company could spec out a machine room with an Amdahl 470V. You could get your tape drives from Telex, disk drives from Memorex, terminals from Datapoint, a Potter printer, and maybe a data general setup as a modem or general communication system. You could be running your own in-house operating system. In fact, that input report I mentioned in the last segment talks about a few people that were doing just such

a thing. You could buy a Fortran compiler from RCA, and you might be using a text editor that was written by one of your buddies over at another college nearby. You might have a service contract with Fujitsu or some totally separate shop in your city. You could exist without ever needing to contact IBM. In fact, IBM may not even know you existed. This is a shockingly similar state of affairs to the PC clone situation of the 1980s. You could have a Compaq Desk Pro running DR-DOS with a Hercules graphics adapter and some random generic screen, and a keyboard manufactured by Datacomp. You would have no connection to IBM at all, and IBM would be none the wiser. In both cases, the only IBM exports

are the overall platform design and maybe a certain level of trust. A key difference between these two clone markets comes down to the success of the clones versus the original hardware. As we saw with the input report, IBM was able to keep a majority share of the mainframe market because customers didn't want to upset IBM. I mean, there are a lot of reasons, but they boil down to fear of men in blue suits. In the mainframe world, IBM was really the top dog. They were this huge monolith that had existed before you were born and would exist long after your bones crumbled to dust. Amdahl and other mainframe cloners were going after existing IBM clients, people that knew the power of Big Blue first hand and already had

dealings with the giant. The PC market, on the other hand, was totally different. IBM was trying, for maybe the third or fourth time, to break into the small business and office market. They were trying to produce a microcomputer instead of some breed of beast. They were entering a market where they didn't have majority market share. A market that had eluded them in the past, one where the IBM name wasn't feared. Most customers that would think of buying a PC had no past dealings with IBM. That was exactly the market they were going after. And since PCs were sold through distributors, prospective customers would, in all likelihood, never meet a single person in a blue suit. They would never have these suited ghouls crawling all over them, telling them about how clone manufacturers couldn't be trusted and Compaq

would dry up in the next six months. I think this further strengthens the story of the PC as this perfect storm scenario. This just adds another bit of inclement weather to that forecast. This just adds another bit of inclement weather to that forecast. It also speaks to IBM's just raw monopoly power. They did not have a monopoly in the microcomputer world. They didn't have power over the market for small machines, so IBM ends up getting wrecked. They didn't have the same kind of weight to throw around. Sure, the owner of some random car dealership in Iowa would know the name IBM,

but that name wouldn't carry very much gravitas when they went out looking for a machine to handle payroll for their small office. Without their huge market share, without the monopoly powers, IBM had a hard time. Put another way, when forced to compete in a fair and open market, IBM wasn't as dominant. Yet again, US v. IBM was very much warranted. Which brings me around to another crucial aspect of the PC story, and one that I never even considered until working on this trilogy. US v IBM is dismissed in January of 1982. That's four months after the PC first ships. Maybe you see where this is going. I personally think it's likely that US v IBM had some impact on the development, release, and cloning of the IBM PC.

on the development, release, and cloning of the IBM PC. This would also mean that the System 360 had a direct impact on the PC. The trick here, however, is finding good proof of this impact. I can't sit here today and tell you that IBM's top brass was worried about the implications of antitrust law on the PC. I bet there's actually a memo out there that could prove this one way or another, something that was turned up in discovery during some court case, but I can't even start to think of where to search for that. What I can do is give you some wild speculation and drop a few choice pieces of evidence. The simple fact is that IBM did sue some clone manufacturers. Notably, Eagle Computer and Corona Data Systems were both sued by IBM. Those manufacturers had copied the PC wholesale right down to the BIOS.

Now, IBM did have intellectual property rights over the PC BIOS, so that was a clear violation of copyright law. Both Eagle and Corona settled out of court, paying IBM some restitutions and agreeing to no longer infringe on the sacred BIOS. IBM did not, however, sue every clone manufacturer. IBM never sued Compaq, Phoenix, Columbia, or the host of companies that produced clean-room versions of the BIOS. Those companies had prepared to be sued. They had built their clones to be legally bulletproof. Over this series, we've seen examples of IBM going up against bulletproof legal defenses to intimidate competitors. Take IBM's countersuit threats against Telex from last episode. IBM used the pressure of a threatened lawsuit to get Telex to stop the appeals process,

on a lawsuit that, in a lower court, Telex won. That, I think, is a perfect example of throwing around some weight, even if that weight wasn't actually backed up by the law. My reading here, and this is speculation, is that IBM did not want to upset the DOJ again. They were tiptoeing and only filing suits that they knew they could win. This is very similar to how IBM tiptoed around Amdahl. They did what they could without appearing too anti-competitive. They walked up to a line. In the case of Amdahl, we know that was a result of USVIBM directly. I think it's highly possible that their treatment of cleanroom PC clones

was also related to USV IBM. If it wasn't for that trial, IBM may well have pushed around more weight to try and stop a company like Compaq. Let me pivot to my final question of the episode, and of the trilogy. Did IBM's experience with mainframe clones prepare them for PC clones? I really like this question because the answer here is roughly no. And honestly, that's something I would never have expected. In my head, IBM has this 100-plus year corporate history, over a century of institutional knowledge and memory. They should be able to predict just about any turn of events. And, to be fair, in a lot of stories that's used to paint IBM as this perfect scheming villain, or on the flip side, as this brilliant business that can do just about anything.

or on the flip side, as this brilliant business that can do just about anything. But when it comes to PC clones, they seem to have been genuinely blindsided. My evidence for this comes from an interview with David Bradley on the website Stay Forever. Bradley wrote IBM's PC BIOS himself, so he's about as core to the story as we can get. In the course of this interview, Bradley is asked if he expected clones and if IBM prepared for them. His response, I think, is very telling. To quote, Poor planning on our part. Based on the S100 experience, we should have expected clones very quickly, but that would have required us to assume that the PC would be a world-changing The conversation continues on those lines for a while.

Now, what I think here is so interesting is that Bradley doesn't reach for the very recent example of System 360 and 370 clones. He reaches for S100. That was a platform, really just a bus, that started out on the Altair 8800. It's microcomputer technology. It was quickly cloned and turned into a standard in the micro world. That whole saga happened around 74, so we're looking at the same time frame, it's the same recency as System 360 and 370 clones. This, to me, speaks to the separation inside IBM. The PC was a small computer. It was seen as totally separate from the rest of Big Blue.

Anyone familiar with the deep lore here could have probably spotted this from a mile away. The PC was developed in Boca Raton, Florida. That is, over a thousand miles from IBM's metropole up in New York. There was already a huge separation between Poughkeepsie and Endicott, despite being just a few hours away by car. Imagine how isolated a Florida office would be. The other crucial piece that Bradley brings up is cross-licensing. Think all the way back to the early days of IBM's antitrust troubles. They were forced, by a consent decree, to cross-license patents. If someone asked to license a patent from IBM and they did so in writing, then IBM legally could not say no. They also had to charge a reasonable fee. The PC was actually protected by

a number of patents. This was easy enough to get around, most clone manufacturers did, but they were there. If a manufacturer actually wanted to infringe all the way, all they had to do was send a letter to IBM and pay a fee. IBM had to say yes. Normally, the incentive here, besides not being sued again by the DOJ, was that IBM could license back technologies from these manufacturers. But when it came to the PC, well, that incentive did not exist. Clone manufacturers were making clones. They didn't, as a rule, develop any new technology that was

really worth patenting, and certainly nothing that IBM wanted to license. Bradley explains that, further down the road, this made IBM amend their licensing policy. Fees started to go up at least as much as they could. But I think, if I can speculate just a little more, that gave IBM even less of a reason to think about PC clones. IBM was, in large part, powerless to do anything about it. That's it. I'm free. After this trilogy, I am not planning to even read the letters IBM for quite a long time. So now that I've escaped, where does this leave us? Well, I think we've covered a huge swath of history here. The System 360 was a massively influential machine. Everything from its design philosophy to the project management used during its development would

impact the computing world. Its success and IBM's willingness to skirt right up to the law led to US v. IBM. That mix of wild success, the drive for standardization in the industry in general, and legal roadblocks made 360 clones possible. While clones never dominated the market, they did nip at IBM's heels. There is a significant separation between

IBM's mainframes and their small systems, but despite that, we can see a very similar story in the PC clone market. Really, the whole debacle with 360 clones presaged what would happen on desks in the early 80s. IBM really should have taken some cues from the past, but they didn't. This was probably due to a combination of factors. IBM was still somewhat restrained by US v. IBM, so they had to tread carefully. They didn't care about micros in the same way they cared about mainframes, and crucially, IBM did not have the same power in the micro world that they did in the realm of big machines. This meant that the PC clone market was unchecked. It was really uncheckable. That, more than anything, is what made the PC special. IBM reached a little too

far out of their domain and lost control of their own computer. They should have seen it coming, but lucky for us, they didn't. Or maybe this is the unlucky outcome. IBM's own hubris, to be a little dramatic, led directly to the PC monoculture of today. They should have seen the writing on the wall, but they just didn't. There were risks associated with the PC, and given IBM's history, the fact that the PC was coming out under very similar circumstances to the 360 and 370, I would have put clones on the top of that list. Thanks for listening to Advent of Computing. I'll be back in two weeks time with, in all likelihood, a live episode. As you might know, as this is being released, I'm just getting back from VCF SoCal, where I'm giving a presentation on Cryotrons.

As long as the recording from that presentation is good, I'm going to be putting that on the feed as my first live show. After that, I'll have a normal show, but let me bask in a little freedom for a couple of weeks. If you like the show, there are a few ways you can support it. If you know someone else who'd be interested in the history of computing, then please take a minute to share the podcast with them. You can also rate and review the show on Apple Podcasts. If you want to be a super fan, you can support the show directly through Advent of Computing merch or signing up as a patron on Patreon. Patrons get early access to episodes, polls for the direction of the show, and bonus content. You can find links to everything on my website, adventofcomputing.com.

If you have any comments or suggestions for a future episode, then go ahead and shoot me a tweet. I'm at Advent of Comp on Twitter. And as always, have a great rest of your day.