The Rise and Fall of the IBM PC
Posted: Wed Dec 08, 2004 11:01 am
With IBM's sale of its PC division, I've been thinking about the decisions IBM made in that arena, and how those decisions have shaped the PC universe. I've been there from day one -- my dad worked for IBM and my first PC in 1982 was a first-gen IBM (64K RAM! Cassette port! RF adapter! 160K floppy drive!).
The original PC project was *not* a high-profile effort within IBM. The budget was low, as was the interest from the higher-ups. PCs were still viewed as toys, but IBM figured they ought to have an offering in the market, rather than handing sales to Apple, Radio Shack and Commodore. The team was free to do pretty much what they wanted, and what they did was *very* un-IBM -- they basically grabbed a bunch of off-the-shelf parts and published the specs.
IBM's dominance in the "big iron" mainframe and midrange markets was based on control of the standards. It was not trivial to make an "IBM-compatible" mainframe, though it was attempted, because the hardware and software was highly proprietary.
However, the 5150 (the first PC) was very different.
There was no budget to build a processor in-house, so the PC team went to Intel, a big player in the then-dominant (at least for business) CP/M market. They were just about to release the 8086, one of the first 16-bit processors. it was a no-brainer to base the 5150 on a 16-bit CPU, as there was no point in expending effort on a soon-to-be-obsolete 8-bit platform. But, all the parts out there were 8-bit. Developing all-new components to mate with the new 16-bit CPU would have been prohibitively expensive and price the 5150 out of the market.
The solution was the 8088. This was a 16-bit chip internally, but unlike the fully 16-bit 8086, the external interface of the 8088 was 8-bit, allowing the team to use all the readily available 8-bit components out there. This was a genius move, since it undercut the price of the "true" 16-bit systems while delivering most of their advantages. The 5150 was released with an 8088 chip running at 4.77MHz. (This was somewhat conservative -- the 8086 commonly was used at 6 and even 8 MHz. But IBM didn't release any faster systems until the 286-based AT in 1984.)
The other genius move was the open architecture, which would never have happened if the PC project was seen as at all important by the IBM brass. But since this was a below-the-radar project, a little something got released called the IBM PC Technical Manual. I had one a long time ago, it was a dark blue hardbound three-ring binder in a slipcase. Inside was all the information a competent engineer needed to build hardware compatible with the 5150, with the exception of the proprietary BIOS. This became known as the Industry Standard Architecture, or ISA.
The effect of this was to usher in a flood of third-party expansion products for the 5150, which debuted at the same time as the system itself. Memory expansion cards, all kinds of interface cards, joystick adapters, specialized graphics boards, anything you could imagine (and most importantly, things IBM didn't imagine). This huge base of third-party support is what sealed the deal for the dominance of the IBM PC platform -- exactly the opposite strategy from what had worked for IBM in big iron.
The same strategy was used for the OS. Rather than cooking up a closed system in-house, the 5150 team went shopping. They famously first stopped off at Digital Research, which made CP/M. The expected move would be to release the 5150 with a 16-bit version of CP/M (CP/M-86), since that was the closest thing to an accepted standard in the business world. However, DR balked at IBM's non-disclosure agreement. Instead of waiting around, IBM put out feelers with other companies. They had already contracted with a little company called Microsoft to provide some language compilers for the new 16-bit architecture.
Bill Gates got wind of IBM's OS woes. He knew of a company called Seattle Computer Products that had hacked up a port of CP/M to the 8086 chip over a weekend so they could play with it (referred to internally as QDOS for Quick and Dirty Operating System). So he bought QDOS from them for what SCP must have thought was an amazing amount of money (a couple hundred thousand, I think), did some quick polishing and made it less like a CP/M clone, and presented it to a delighted IBM.
The 5150 debuted with a choice of operating systems -- you could buy CP/M-86 for $250 or IBM PC-DOS for $40. Bill Gates had made the move of several lifetimes. DR was reduced to selling enhancements for DOS, and eventually created a DOS clone called DR-DOS before throwing in the towel and selling out to Novell.
Of course, after the 5150 (and the 5160 "XT", which introduced official IBM hard disks to the line, and the 5170 "AT" 286, which was the first true 16-bit IBM PC) was a huge, market-dominating hit, IBM reverted to form. They watched as the BIOS was quickly reverse engineered and a wave of clones hit the market, offering lower prices, different form factors and higher CPU speeds. (None of these original clone companies still exist, now that Compaq has been consumed by HP.) They saw Microsoft become enormous by selling what it called "MS-DOS" to all these clone makers.
IBM saw this as a bad thing -- their big iron experience told them that they had to own the standards. As it was, anyone could make compatible hardware and IBM didn't get a dime directly.
So, in 1987, IBM attempted to take over. It introduced the two-fisted attack of the PS/2 and OS/2. The PS/2 was based on a new hardware standard called the Micro Channel Architecture, which replaced the ISA on high-end (286 and 386) PS/2s. It was an undeniable upgrade from the ISA, with a catch -- if you wanted to make cards or systems based on it, you had to pay IBM a royalty. Similarly, OS/2 was IBM's baby (though it was largely developed under contract by Microsoft).
This was a resounding flop. The aftermarket largely ignored MCA and carried on with ISA-based 386 systems. The MCA-based PS/2s started strong but soon were selling only to die-hard IBM shops. IBM was forced to rush-introduce several ISA-based models to keep sales alive. The emergence of Windows (and eventually Windows 386 and 3.0) crushed OS/2 by providing many of the same benefits without the high cost and incompatibility. OS/2 2.0 was much improved, but by then it was much too late. By the 90s, IBM had completely lost their leadership position in the PC market in both hardware and software. The PS/2 contributed exactly two lasting hardware standards -- VGA and the "PS/2-style" mouse and keyboard connectors. MCA was forgotten. IBM never introduced another market-standard hardware or software innovation.
The IBM PC changed the rules in many ways -- it killed the dominant MIS (what they called IT in the old days) big-iron dumb-terminal paradigm and fast-forwarded the home-computer market. But it also killed the model of closed standards. Apple still holds a minority of the market with its closed Mac line, but even they are now running an OS based on Unix. And, most interestingly, IBM is rumored to be thinking about spending some of the dough it got for its PC business on an investment in Apple.
Stay tuned...
Ryan
The original PC project was *not* a high-profile effort within IBM. The budget was low, as was the interest from the higher-ups. PCs were still viewed as toys, but IBM figured they ought to have an offering in the market, rather than handing sales to Apple, Radio Shack and Commodore. The team was free to do pretty much what they wanted, and what they did was *very* un-IBM -- they basically grabbed a bunch of off-the-shelf parts and published the specs.
IBM's dominance in the "big iron" mainframe and midrange markets was based on control of the standards. It was not trivial to make an "IBM-compatible" mainframe, though it was attempted, because the hardware and software was highly proprietary.
However, the 5150 (the first PC) was very different.
There was no budget to build a processor in-house, so the PC team went to Intel, a big player in the then-dominant (at least for business) CP/M market. They were just about to release the 8086, one of the first 16-bit processors. it was a no-brainer to base the 5150 on a 16-bit CPU, as there was no point in expending effort on a soon-to-be-obsolete 8-bit platform. But, all the parts out there were 8-bit. Developing all-new components to mate with the new 16-bit CPU would have been prohibitively expensive and price the 5150 out of the market.
The solution was the 8088. This was a 16-bit chip internally, but unlike the fully 16-bit 8086, the external interface of the 8088 was 8-bit, allowing the team to use all the readily available 8-bit components out there. This was a genius move, since it undercut the price of the "true" 16-bit systems while delivering most of their advantages. The 5150 was released with an 8088 chip running at 4.77MHz. (This was somewhat conservative -- the 8086 commonly was used at 6 and even 8 MHz. But IBM didn't release any faster systems until the 286-based AT in 1984.)
The other genius move was the open architecture, which would never have happened if the PC project was seen as at all important by the IBM brass. But since this was a below-the-radar project, a little something got released called the IBM PC Technical Manual. I had one a long time ago, it was a dark blue hardbound three-ring binder in a slipcase. Inside was all the information a competent engineer needed to build hardware compatible with the 5150, with the exception of the proprietary BIOS. This became known as the Industry Standard Architecture, or ISA.
The effect of this was to usher in a flood of third-party expansion products for the 5150, which debuted at the same time as the system itself. Memory expansion cards, all kinds of interface cards, joystick adapters, specialized graphics boards, anything you could imagine (and most importantly, things IBM didn't imagine). This huge base of third-party support is what sealed the deal for the dominance of the IBM PC platform -- exactly the opposite strategy from what had worked for IBM in big iron.
The same strategy was used for the OS. Rather than cooking up a closed system in-house, the 5150 team went shopping. They famously first stopped off at Digital Research, which made CP/M. The expected move would be to release the 5150 with a 16-bit version of CP/M (CP/M-86), since that was the closest thing to an accepted standard in the business world. However, DR balked at IBM's non-disclosure agreement. Instead of waiting around, IBM put out feelers with other companies. They had already contracted with a little company called Microsoft to provide some language compilers for the new 16-bit architecture.
Bill Gates got wind of IBM's OS woes. He knew of a company called Seattle Computer Products that had hacked up a port of CP/M to the 8086 chip over a weekend so they could play with it (referred to internally as QDOS for Quick and Dirty Operating System). So he bought QDOS from them for what SCP must have thought was an amazing amount of money (a couple hundred thousand, I think), did some quick polishing and made it less like a CP/M clone, and presented it to a delighted IBM.
The 5150 debuted with a choice of operating systems -- you could buy CP/M-86 for $250 or IBM PC-DOS for $40. Bill Gates had made the move of several lifetimes. DR was reduced to selling enhancements for DOS, and eventually created a DOS clone called DR-DOS before throwing in the towel and selling out to Novell.
Of course, after the 5150 (and the 5160 "XT", which introduced official IBM hard disks to the line, and the 5170 "AT" 286, which was the first true 16-bit IBM PC) was a huge, market-dominating hit, IBM reverted to form. They watched as the BIOS was quickly reverse engineered and a wave of clones hit the market, offering lower prices, different form factors and higher CPU speeds. (None of these original clone companies still exist, now that Compaq has been consumed by HP.) They saw Microsoft become enormous by selling what it called "MS-DOS" to all these clone makers.
IBM saw this as a bad thing -- their big iron experience told them that they had to own the standards. As it was, anyone could make compatible hardware and IBM didn't get a dime directly.
So, in 1987, IBM attempted to take over. It introduced the two-fisted attack of the PS/2 and OS/2. The PS/2 was based on a new hardware standard called the Micro Channel Architecture, which replaced the ISA on high-end (286 and 386) PS/2s. It was an undeniable upgrade from the ISA, with a catch -- if you wanted to make cards or systems based on it, you had to pay IBM a royalty. Similarly, OS/2 was IBM's baby (though it was largely developed under contract by Microsoft).
This was a resounding flop. The aftermarket largely ignored MCA and carried on with ISA-based 386 systems. The MCA-based PS/2s started strong but soon were selling only to die-hard IBM shops. IBM was forced to rush-introduce several ISA-based models to keep sales alive. The emergence of Windows (and eventually Windows 386 and 3.0) crushed OS/2 by providing many of the same benefits without the high cost and incompatibility. OS/2 2.0 was much improved, but by then it was much too late. By the 90s, IBM had completely lost their leadership position in the PC market in both hardware and software. The PS/2 contributed exactly two lasting hardware standards -- VGA and the "PS/2-style" mouse and keyboard connectors. MCA was forgotten. IBM never introduced another market-standard hardware or software innovation.
The IBM PC changed the rules in many ways -- it killed the dominant MIS (what they called IT in the old days) big-iron dumb-terminal paradigm and fast-forwarded the home-computer market. But it also killed the model of closed standards. Apple still holds a minority of the market with its closed Mac line, but even they are now running an OS based on Unix. And, most interestingly, IBM is rumored to be thinking about spending some of the dough it got for its PC business on an investment in Apple.
Stay tuned...
Ryan