The Independent Kixmiller Pigeon MrkII

May 26, 2021

Big Blue's attempt at recapturing the PC market

The IBM PS/2 Family of Computers. Image From Wikipedia.

This is a continuation of the articles covering the history of PC's using emulators to illustrate the march of progress. This is part three in the series of articles covering the IBM PS/2 family of computers.

IBM PS/2 Model 80

Introduced in April 1987, the PS/2 Model 80 was the highest end of the original PS/2 models. The Model 80 shipped with a Intel 386 processor. Intended as a high-end workstation, this system could run various operating systems, like OS/2, IBM AIX, and Xenix. AIX was IBM's variant of the UNIX operating system which ran on there proprietary workstations and servers. The version for the Model 80 was ported to the Intel platform and could run on this PC as well. Xenix was Microsoft's variant of the UNIX operating system. Around this time, Microsoft exited the UNIX market, and Xenix was transferred to SCO (Santa Cruz Operation). Of the different variations of UNIX, Xenix was the most widely used for the Intel platform.

PCEM emulates this machine very well. Like the other models, I needed to download the disk image of the reference/starter disk in order to configure the machine and set up the hard-disk controller. Like the other starter disk, I needed to modify the contents of the disk image to replace ad ADF driver file in order for the hard-disk controller to be recognized correctly. Once I configured the disk controller, I was able to use the hard-disk image and install both IBM PC-DOS and OS/2 Version 1.3. The OS installations went without issue. In the future, I will attempt to install IBM AIX and Xenix on this emulated configuration as well. I was able to install AIX on Virtual PC in the past. Granted, it was a exercise in frustration and agony. The AIX PS/2 installation is one of those things that one has to prepare themselves both spiritually and mentally before undertaking.

OS/2 Version 1.3

More OS/2

Intended as the successor of DOS, OS/2 started shipping around the same time that IBM was launching their new series of computers. The OS was a joint project between IBM and Microsoft. Some could compare this more as a marriage of convenience. IBM understood corporate needs like the back of there hand, and they also had experience writing advance operating systems. At the same time, IBM didn't understand PC users, plain and simple. On the opposite side of the coin, Microsoft didn't know how to write advance OS's at the time, but they understood the needs of PC users much better than IBM did.
OS/2 would have many advance features, which most shipped at the very start. OS/2 could run programs that could use more than 640KB of RAM, something that DOS couldn't do (yet). Applications could multi-thread and multitask as well, something that DOS couldn't do (yet). However, some of the heavily promoted features of OS/2 wasn't ready yet, like its GUI, the Presentation Manager, which shipped in 1988 with Version 1.1. Installable filesystems, which was a feature that allowed OS/2 to use various filesystems, wasn't ready until 1989 with OS/2 Version 1.2, which included HPFS. HPFS, or High Performance File System, was intended as being the FAT (File Allocation Table) killer. It had the ability to use long file names up to 255 characters, and made more efficient use of disk space, along with other features.

OS/2 version 1.x was originally designed to run on the 286 processor. IBM placed this demand in order to keep its promise to customers that purchased 286 systems from them promising the advance OS in the future. Microsoft pushed for the 386 version because the 386 offered far more advance features, most notably the virtual 8086 mode, which would allow OS/2 to run multiple DOS applications at once. It was also easier for software developers to program the 386 processor because of its flat memory model, which could keep the code together without splitting the program into separate memory segment. The 286 segmented it's memory, which made it harder for programmers as they had to split there code into chucks.

It was decided that OS/2 Version 1.x would support the 286 while Version 2.x and above would support the features of the 386. The 286 processor didn't support a virtual 8086 mode like the 386 did. As a result, the processor had to restart when switching to DOS mode. This reset could be done without resetting the computer or disrupting OS/2. The main issues though was that this restricted DOS to a single session that ran fullscreen, and a badly misbehaving DOS program could lock up the computer, forcing the user to restart the computer. DOS compatibility was planned from the start as it would have been a necessity for OS/2 to succeed in the marketplace.

Gordon Letwin, who was the primary architect of OS/2 on Microsoft side, and who also designed and wrote code for the HPFS file-System, had written a awesome book describing the OS design of OS/2 version 1.x called "Inside OS/2", which describes how OS/2 is intended to operate and why the OS design in the way that it is. He described many of the challenges of getting OS/2 to conform to the restrictions imposed by the 286 processor, and what the developers did to bypass those restrictions. Highly recommended reading for programmers and software designers.

At the same time, Microsoft was placing there hopes in OS/2 as Windows hasn't taken off. Windows 1 was a flop, while Windows 2 did better, but it didn't establish GUI computing. The main issues was that most computers in the 1980's weren't cut out for that sort of computing. While the Intel 8088/8086, 286, and most 386 processor could run such environments, they often didn't do a good job though. The graphical demands of the environment placed a huge demand on the processors.
OS/2 didn't take off either for primarily the same reasons. It was a very demanding OS at a time when IBM compatibles were still very expensive. OS/2 couldn't run on a XT-class system for starters. AT and PS/2 class systems were still a very expensive proposition at the time, so the size of the potential audience would have been limited. OS/2 was the OS equivalent of Crysis in the 80's. Also, there were few applications at launch, which limited its appeal. Most of the applications that were eventually developed were home-grown corporate applications made by various Fortune 500 companies like GM, Ford, GE, American Airlines, Bank of America, Chase, etc. Task like handling payroll, data processing, and handling hardware like ATM machines.

These companies had there own LAN's that were the size of small intranets. As a result, OS/2 computers in these environments were often networked together and used the Microsoft LAN Manager for these machines to communicate with each other and utilize the resources on the network. Microsoft even created there own version of the HPFS filesystem, HPFS386, which was intended for the network setting. The main benefit of HPFS386 was performance, as the filesystem was written in 32-bit Intel assembly language. This allowed for faster file access on a networked server.

OS/2 wouldn't get its big break until the middle 90's. By that point, IBM and Microsoft dropped out of the joint project, with OS/2 being exclusively under IBM's direction. Microsoft's OS/2 codebase, after a very massive rewrite and transformation at the hands of Dave Cutler and and friends who defected from DEC, became Windows NT. At this point, I also highly recommended reading "Show Stopper!: The Breakneck Race to Create Windows NT and the Next Generation at Microsoft". This is also a very good book about OS development and how Windows NT became Microsoft's future, with some of the IBM vs. Microsoft drama captured in the pages as well. By the 90's, Microsoft found success with Windows 3.1, and the two Goliath's would square off in a battle that IBM would eventually lose.

The Failure Of The PS/2

Calling the PS/2 a total failure is a bit sensationalist. The computer found plenty of success among corporate customers heavily invested in IBM's services. The computers would play a major role within the operations of various Fortune 500 companies for a couple of years. However, this family of computers failed in one major aspect; they failed to reestablish IBM's control within the computer industry, which is the whole reason why this platform existed in the first place.

These machines tend to be very expensive. While many were purchased by Fortune 500 companies, the models purchased were usually tended for niche uses or used in the server setting. These companies would still have been much more inclined to purchase far more affordable clone PC's from Compaq or AT&T for their office workers. One didn't need to spend +$5,000 while a $1,500-$2,000 clone PC could still run WordPerfect and Lotus 123 reasonably well.

IBM making the MCA bus and related technologies proprietary surely didn't help matters either. In order to use the MCA bus, a company had to pay a license fee to Big Blue for the privilege. The MCA license in itself wasn't that much, but IBM demanded a royalty payment from each system sold. Companies that licensed MCA were required to pay IBM royalties not only for MCA systems sold, but XT and AT compatibles sold previously as well over the years. When IBM created the PC, the motherboard was built from open standards, so Big Blue was not in the position to demand payment from clone-makers. Few bit the bullet. Compaq actually brought a MCA license from IBM, but they never sold any computers with MCA slots in them.

Not only did most companies reject MCA, but several companies, with Compaq taking the initiative, decided to create there own 32-bit bus to compete with MCA called EISA. EISA, or Extended Industry Standard Architecture, was a 32-bit bus that offered features very similar to MCA, but yet, was compatible with 16 and 8-but ISA adapters as well. In retrospect, this standard never caught on either as the buses and motherboards were expensive to make, restricting EISA, like MCA, primarily to the server room. Typical PC users wouldn't see a high-speed bus until Intel made PCI and released it as open standard many years later.

The counter-move by Compaq reflected that IBM wasn't in control of the PC standard anymore, just active participants in it.