On October 17, 1985, Intel released its 386 CPU, also known as the 80386, and later, as the 386DX. The 386 changed the direction of the computer industry like no chip before it, and maybe since. The phrase “disruptive product” didn’t come into our lexicon until the mid 1990s, but when Harvard professor Clayton Christensen published his theory of disruptive innovation in 1995, the 386 powered one of his key examples.
IBM, the 386, and its repercussions
IBM was conflicted over the 386. Don Estridge, the engineer who oversaw the PC/AT, wanted to build a successor to the AT around the 386. He wanted to buy out Intel’s initial production run so that clonemakers wouldn’t be able to get adequate supplies of the chip. He’d done something similar in the 286 era.
The thing was, in the mid 1980s, IBM still had a very lucrative business selling minicomputers. These were medium-sized computers designed for applications where a mainframe would be overkill but a PC wasn’t powerful enough. In 1987, IBM sold $14 billion worth of minicomputers. IBM’s PC business was worth more like $4 billion a year in that era.
So there was a faction of IBM who didn’t want the 386 because it performed like a minicomputer. They didn’t want businesses buying a 386-based PC and loading Unix on it instead of buying a traditional minicomputer that cost twice as much.
The documentary Silicon Cowboys states that IBM wanted an exclusive on the 386 but doesn’t cite a source. Maybe the filmmaker misinterpreted what Don Estridge wanted to do, or maybe there was a third faction inside IBM with a slightly different opinion.
The IBM faction that didn’t want the 386 won out. And that caused two things to happen.
Intel shedding second sources
When IBM selected the Intel 8088 CPU to power its IBM PC 5150, IBM insisted on Intel licensing the design to at least one other chip manufacturer. This was to ensure that IBM would have a sufficient supply of chips to meet demand. It was also not a particularly unusual request. Both Apple and Atari had multiple sources for the chips they used in their computers, for instance.
Intel did IBM one better. Not only did they license the design to AMD, but they licensed it to several other companies as well. And after IBM selected the 80286 CPU to power its PC/AT, Intel amended the agreement with AMD and several others so they could manufacture not only 8088 and 8086 CPUs, but 80286 CPUs as well.
But with IBM uninterested in the 386, Intel had no reason to amend its agreement and extend the 386 to second sources. IBM wasn’t asking, and no one else had the clout to demand it.
AMD did clone the 386, but Intel fought it in court and kept it off the market until 1991, after the release of the 486.
Compaq taking the lead
Compaq famously released a 386-based PC in 1986, forcing IBM to do the same within 12 months. But that put IBM in catch-up mode, and of course, other 386 clones appeared in the meantime. The 386 was the first step in IBM losing control of the PC ecosystem it had built. The PS/2 line was the second. Compaq took the lead for a while, and by the mid 90s, we didn’t use the phrase “IBM compatible” so much anymore. Instead, “PC” took its place, or “Wintel,” an amalgamation of the words “Windows” and “Intel.”
And as Compaq faded, Intel and Microsoft increasingly took control over the direction of the ecosystem. IBM ultimately left the PC business in 2006.
Why the 386 was special
The 8088 was a bit of an oddball mashup 8-bit/16-bit CPU, designed to make porting 8-bit 8080 software easy, resulting in some design decisions that made lots of sense at the time but got in the way of future performance as time wore on. The 286 was more of a true 16-bit CPU and could be made to behave like it, or it could behave like a fast 8088 for compatibility reasons.
The 286 could run Unix, though it wasn’t the ideal Unix platform. And as things stood in 1985, it looked like the next generation of computer platforms would be powered by something else, likely made by Motorola. High-end Unix workstations and servers used Motorola CPUs. And toward the lower end of the price scale, so did the Apple Macintosh, Atari ST, and Commodore Amiga.
The 386 was a fully 32-bit chip, like Motorola’s brand new 68020 CPU. Its advantage was offering backward compatibility with the earlier 8088 or 286, so you could boot it into DOS and run your favorite legacy DOS apps on it, and much faster than on earlier machines. Or it could operate like a next-generation 32-bit CPU. Like the 68020, it could run a modern operating system without compromises, whether that was Unix, or something else like OS/2 or Windows NT. The 386 meant companies like Microsoft, Digital Research, and Quaterdeck could think about what was next. And what was next was exciting.
The 386 was in the same league as the Data General minicomputer that was the subject of the Pulitzer Prize-winning book The Soul of a New Machine by Tracy Kidder. Not even three and a half years after Kidder won the Pulitzer Prize, Intel was bringing comparable power to the desktop. What a time to be alive.
The Innovator’s Dilemma and the 386
Clayton Christensen’s influential book, The Innovator’s Dilemma, doesn’t mention the 386 by name. But one of its key examples is how minicomputers disrupted the market for mainframes, and then desktop computers disrupted the market for minicomputers, and how the minicomputer makers struggled to enter the market for desktop computers. It’s not meant to be a book about technical details, but about a higher-level concept.
Christensen really wanted his readers to understand that the things that propelled Digital in the 1980s were the very things that derailed it in the 1990s. There was a 1980s biography of Digital founder Ken Olsen that called him “the ultimate entrepreneur” on the cover. Today we remember him for saying he couldn’t imagine people wanting a personal computer.
But if you’re interested in the technical details, the 386 was a key step in the journey microcomputers took to disrupt the market for larger computers. The reason it didn’t become a dead end like, say, Data General and DEC minicomputers, was because of backward-compatible successors like the 486, Pentium, and x86-64 which, yes, originated at AMD, not Intel. PCs undercut the minicomputers in price, and as their power grew, their prices kept coming down. Today what we call a server fills the role minicomputers used to fill. But it’s fair to think of a server as a PC built with industrial-grade parts.
Why the PC didn’t get disrupted like minicomputers and mainframes did
Mobile devices like smartphones and tablets cut into the PC market. But they couldn’t completely replace them. Smartphones and tablets arguably are also examples of disruptive technology. But they ended up coexisting with PCs, if only because the cloud-based services they rely on are running on servers in distant datacenters. Sometimes these servers run on the same CPU architecture as a mobile device–it’s called ARM, if you’re curious–but servers also often use a modern Intel or AMD CPU descended from the 386.
And while some kinds of work lend themselves very well to mobile, not all do. For some types of work, you really need a keyboard and a bigger screen. Maybe multiple screens. I wrote an early draft of this blog post on my phone, for example. But I expanded on it and edited it and finalized it on a PC. I could do it on my phone in a pinch, but it’s faster with a keyboard and a bigger screen.
386 vs 286
Compared to the Intel 286, the 80386 had a longer instruction pipeline, extended the architecture from 16 bits to 32 bits, and added an on-chip memory management unit. The MMU made it much easier to implement operating systems that used virtual memory. The 386 featured three operating modes: real mode, virtual 8086 mode, and protected mode. The protected mode, which debuted in the 286, was extended to allow the 386 to address up to 4 gigabytes of memory. The new virtual 8086 mode made it possible to run one or more real-mode programs in a protected environment, although some programs were not compatible.
The 32-bit i386 ran most code intended for the earlier 16-bit processors such as 8086 and 80286 without issue. It had 275,000 transistors and ran at speeds ranging from 12 MHz to 25 MHz, with a 33 MHz version launching later. The least expensive version of the 386 cost $299 at launch, but magazines cautioned the 386 needed more complex and expensive supporting chips than the 286, increasing the total cost of the system. Intel later released the less expensive 386SX, which helped drive 386 adoption as a whole.
All of this meant a theoretical PC built on the 386 architecture could move upmarket. It had a CPU that could compete with Motorola, so all it would need was a new graphics standard and a standard for sound. Of course, both of those things followed soon after, in the form of VGA and Ad Lib.
The 386 lived a very long life. It was obsolete as a PC CPU by the mid 90s, eclipsed by the 486. But it thrived in embedded and industrial applications for another decade and a half. Intel didn’t discontinue it until September 28, 2007.
David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.