The IBM System/3 family was a very successful line of small business-oriented computers in the 1970's. The first member, the Model 10, shipped in mid-1969 and was a card-based batch-oriented system intended to replace unit-record (punched card) commercial applications: billing, inventory control, accounts receivable, sales analysis, payroll, etc. The System/3 applications were written in RPG, a non-procedural language originally based on the data-flow of IBM card-based accounting machines. Of particular interest, the Model 10 introduced a new format of punched card along with I/O devices for it. This card was much smaller than the standard IBM card but could contain more data: 96 columns vs. 80 columns.
Figures 1 and 2 show our console from the card-oriented IBM System/3 Model 10.
The System/3 Model 6,shipped in late 1970, was the second model introduced. As opposed to the card-based batch-oriented Model 10, the System/3 Model 6 was designed for interactive usage, both for business applications, but also for engineering and scientific applications. It had no attached card equipment, but rather had a good keyboard and an (optional) CRT display. The business applications used the same general operating system and RPG language, but there was a new interactive BASIC-language operating system for engineering and scientific applications.
Figure 3, from an IBM ad, shows a person using the System/3 Model 6 with its BASIC software package. Figures 4 etc. show the museum's Model 6 (we do not have the optional CRT display).
In 1968 I was working at IBM San Jose on the design of a follow-on to the IBM 1130, a small engineering and scientific computer. Late in the year, our project was redirected by upper management to use another IBM processor that was being developed in Rochester: the IBM System/3 processor. This was probably the worst architecture ever imagined for engineering and scientific computing:
I was redirected as well--to the newly forming Boca Raton development laboratory. I arrived in Boca in early 1969 and became the manager of one of two departments developing the IBM System/3 Model 6 BASIC program. I managed the operating system underlying the BASIC interpreter. My department implemented a 64KB virtual memory (in software, there was no address translation hardware), the math library, the I/O subsystem, a powerful desk calculator function (ala the later UNIX "bc" function), and an integrated, disk-based help system. In spite of my complaints about the System/3 instruction set, the BASIC system actually performed well.
In addition to the BASIC project being my first IBM management job, I also met my wife while working on this project; she was an IBM programmer in Boca working on operating system development. (In fact, as a nice touch, my System/3 Model 6 arrived literally on the day of our 42th anniversary.)
After my work on the System/3 Model 6 was finished, I worked on the design of another small computer system. That project was cancelled and, in the Fall of 1971, I moved to Rochester where my new job was ss the second-line manager of the system software for another new small computer system: the IBM System/32. Amazingly, the museum also has our own System/32, check it out.
The IBM System/3 Model 6 hardware together with its BASIC software package was, I claim, the first "IBM Personal Computer"; that is, a computer on a desk with a keyboard and a CRT display device, a printer, a removable disk drive, and software for a user to program in interactive BASIC. Effectively, this system could do what the Apple II did, only much better (internal fixed disk, 64KB virtual memory, etc.), and 9 years earlier.
The only flaws with our solution was that it was a large desk with an large lump on one end, it weighed 1,300 lbs, it used 220V power, and it rented for about $1,000 a month. Accordingly, sales were low and the BASIC software received little attention in the market or within IBM. (In spite of the BASIC system having no visibility, I and another person from my department [Brad Beitel] went on to be very successful at IBM.)
My Model 6 was procured in August, 2013. It came from the original owner who had used it to run business applications for her small company. It was used for this purpose until about 1985 when it was decommissioned by IBM and stored in the owner's basement, where it gradually became buried under typical basement storage items (show in Figure 23).
Getting it physically to the museum (in Austin) from the basement in Omaha was a complicated logistically operation, but it finally made it. (Mark Rothbauer of Centaur was the key person here; he went to Omaha and freed it from the basement.) The system's condition is dusty but good considering the it is over 40 years old and hasn't been used for over 25 years. The usage meter indicates 2977 hours powered on. The Model 6 also came with 17 disk packs including the operating system, as well as a lot of documentation.
I (with help from others here at Centaur) will try to restore it to running condition. The likely things that need repairing are replacing electrolytic capacitors and drive belts, but my biggest worry is the disk drive. Stay tuned for progress.
The Model 6 is very rare these days. A interesting web site that covers System/3 family tracks existing devices. It knows of only two other Model 6's, neither of which are operational.
Figures 3, 4, and 5 show the general layout. Figure 6 shows the keyboard, which is very nice, especially for that time. Included are a numerical keypad and 16 function keys. The keyboard complex also included the basic controls for the operator along with "halt code" lights which the program can set. Figure 7 show detail of the 16 lights corresponding to the function keys. These keys are used by the application program and thus have user-writable labels on the lights to remind the operator of what the keys do. To the right of the function key labels there are another eight output lights that can be set by the program. These also have user-writable labels and are used to tell the operator what function to perform next.
Figure 8 shows the back of the swing-out logic gate with its wiring. There are five logic boards installed. (The missing one is for a communication option). Figure 9 show the other side of the gate and Figure 10 shows it with the board covers removed. Figure 11 shows a closeup of a particularly interesting board. On the left of this board are eight pluggable cards, each containing many modules, primarily IBM MST logic modules (more on this technology below). On the right of this board is the magnetic-core memory module. Our system is fortunate in that it has the largest possible Model 6 memory: 16KB. The smallest memory for a Model 6 was 8KB. (The BASIC software package I worked on ran in only 8KB).
(The Model 6 was probably the last IBM computer to use core memory. Subsequent System/3 models used solid state memory as did the S/370 IBM mainframe family at this time.)
The System/3 instruction set was very simple. It had 28 instructions, two index registers, a condition flags register, and some control registers. There were no accumulator or general purpose registers; operations were performed memory-to-memory. For example, the ALC (Add Logical Characters) instruction (4-6 bytes long, based on addressing options) adds one memory field to another field, one-byte at a time for a specified length from 1 to 256 bytes. Instruction timing is simple: 1.52μs for each memory byte touched. For example, a 4-byte ALC instruction moving 16 bytes in memory takes 30.4μs.
Figure 12 shows the CE (Customer Engineer) control panel. This was our primary debug tool when I was developing code for this system. We would single step the code examining processor state using the lights. You could also easily modify memory from this console. Figures 13 and 14 show the back view of the system including the cables the connect the two halves together. For shipping and installation, the system could be separated into two components: the "table" part, and the "box" part. (Our system had to be separated into two pieces in order to get it though doors.)
Figure 15 shows the back of the 5444 disk drive unit at the left of the table (from the front). There is a compartment for a second 5444 under this, but our system only has one disk drive. There were two models of this drive; the larger had a capacity of 5MB split equally between a fixed platter and a removable platter.
Figure 16 shows the disk drawer opened with the removable disk installed. Below this is a fixed disk platter that shares the same access arm as the removable disk. Figure 17 shows the disk drawer with the removable disk removed. The fixed disk is barely visible through a hole on the right (where the index transducers are). Figure 18 shows a view looking at the retracted heads and the cleaning brushes on the right.
Figure 19 shows some of the documentation received and five disk cartridges containing system software. Figure 20 shows 12 more cartridges that we have. Figure 21 show a closeup of the removable disk cartridge with the access holes for the heads and cleaning brishes.
Finally, Figure 22 shows the a cassette tape used by the IBM Customer Engineer to run diagnostics and deliver software fixes. This was used since the only removable medium is the disk cartridge, too large for the CE to haul around. So, looking at Figure 12, there is an audio jack ("J1") used to plug in a ordinary tape player.