The DEC PDP-8 StoryPart III: the concluding years
by
Douglas W. Jones
|
The DEC PDP-8 StoryPart III: the concluding years
by
Douglas W. Jones
|
Originally published in Historically Brewed, 9 (1997). This copy is an HTML version of the text file I submitted for publication.
In the first two parts of this series, the focus was on the technological and marketing innovations that pushed the Digital Equipment Corporation PDP-8 into a dominant position in the minicomputer market of the early 1970's. In this final installment, the focus is on how DEC continued to successfully market the PDP-8 through the years of 16 and 32 bit minicomputers and well into the era of microprocessors.
The minicomputer market of the early 1970's was large and growing. While DEC rightly claimed, in a January 1971 ad, that the PDP-8 was the yardstick by which all other minicomputers were measured, the market was already dominated by 16 bit machines.
A survey of the minicomputer market published in May 1971 showed 49 minicomputers on the market, made by almost as many companies. Of these, 36 had 16-bit words, 4 had 12-bit words, and only 8 had 8-bit words (Motorola's offering was one of the latter).
A followup survey, published in July 1974, showed only two 12 bit machines, the PDP-8/E and the Nuclear Data 812. All of the others were 16 or 32 bit machines, with the exception of a handful of low-end 8 bit minis.
Looking at the sales figures, on the other hand, the PDP-8 was by far the dominant machine! In the 1974 survey, the PDP-8 was the only machine listed with an installed base of over 12,000 machines. The closest competitor listed was the Hewlett Packard 2100 family, with over 5,000 machines. (It should be noted that Data General's sales figures were not released for this survey.)
While the growth of hardware technology in the 1960's was largely driven by the mainframe industry, many of the new technological developments of the 1970's were driven by the growing market for small computers.
While many of the fastest mainframes were still being built using proprietary hardware technology, the minicomputer industry embraced TTL logic and became the driving customer behind many of the developments in MSI and LSI fabrication.
The announcement of the first microprocessor in January 1972 was widely expected. Nobody knew that it would be the Intel MCS 4, the 4004, but many expected that one of the chip fabricators would soon come out with something of that sort. In fact, the HP-35 calculator, announced in February 1972, clearly demonstrated that HP was also able to produce produce LSI chips of similar complexity.
The peripheral marketplace of the early 1970's was clearly dominated by devices designed for the mainframe market, but certain minicomputer applications demanded smaller peripherals and more convenient storage media. IBM, having first used the floppy disk for microcode loading in their mainframes, was quick to exploit this with the 3740 diskette based data entry workstation, announced in March 1973, and after Shugart announced the SA 900 IBM compatible diskette drive, four months later, the industry quickly recognized that the 8 inch diskette was a winner.
Minicomputers also demanded inexpensive terminals. Nobody really liked the ubiquitous ASR 33 Teletype, and when Lear Siegler Inc. introduced the LSI ADM-1 "glass teletype" terminal priced at $1500 -- the same price as a teletype, it was an immediate hit in the minicomputer and timesharing markets.
While many minicomputers continued to be purchased directly by their end users, more and more machines were being purchased by OEM customers -- the Original Equipment Manufacturers who incorporated computers into their products, ranging from industrial control equipment to elevator controllers to word processing systems.
The OEM marketplace was very lucrative! Sales to end users required an extensive investment in customer service, support and training, but OEM customers were expected to take care of these expensive details themselves when they resold systems to end users. Some of the smaller minicomputer makers only sold to OEM customers, thus completely avoiding these costs.
The DEC PDP-8/E sold for $4990 without a teletype when it was announced in July, 1970. This not only undercut the older PDP-8/L by $2000, but also offered immensely more flexibility with the new Omnibus backplane.
It was not, however, enough to offer the automatic market advantage that was suggested by DEC's October 1970 advertising slogan: "If the new PDP-8/E is so great, how come it's so cheap?"
In fact, only two months before the PDP-8/E came on the market, Digital Computer Controls introduced the D-112 at almost exactly the same price. This machine was advertised as being both plug and software compatible with the PDP-8/L, and it was available only to OEM customers.
The most impressive PDP-8 clone came to market later. The Fabritek MP-12, offered in June 1974, was available to OEM customers for $990 each, in lots of 100. This machine was only 2 inches wide, 9.5 inches high, and 15 inches deep.
DEC's first response to the market demand for low cost OEM hardware was the PDP-8/OEM, announced in OCT 1970 at a price of $3250, or $2800 in lots of 100. This was a PDP-8/E, stripped of its flashy looking front panel and its core memory, and outfitted with 256 words of ROM and 256 words of integrated circuit RAM.
In 1972, DEC brought out a new power supply and backplane for their omnibus machines. The original PDP-8/E power supply used inefficient linear regulators, while the new PDP-8/F and PDP-8/M power supply was a compact and efficient switching supply. DEC continued to make the PDP-8/E for those customers who wanted large configurations requiring more than 20 slots in the backplane, while they packaged the new switching power supply in a small package with no room for an internal backplane extension.
Physically, the PDP-8/F and PDP-8/M are identical machines; the 8/F was sold to end users, while the 8/M was sold to OEM customers. Because of this, most 8/M systems had no front panel, although it was an available option. Many PDP-8/M systems were almost certainly configured along the lines of the 8/OEM, using a ROM to hold the application code.
While the PDP-8 seems small by today's standards, almost all of the early minicomputers were offered with around 4K words of memory, and the limit of 32K 12-bit words on the PDP-8 was respectable -- this is 48K 8-bit bytes, exactly the same as the standard maximum for the PDP-11/20, although some ambitious users managed to squeeze in an extra 8K of core into their PDP-11/20 systems.
In fact, in May 1972, DEC brought out the FPP-12, a floating point coprocessor for PDP-8 and PDP-12 systems, and with this and a new FORTRAN IV compiler, DEC took on the IBM 1130, an aging but still common mainframe.
In July 1972, DEC advertising countered the generally held view that the PDP-8 family was nearing obsolescence with the slogan "It's as current as a PDP-8", featuring the OS/8 single-user operating system and the COS/300 commercial data management system.
Some of the patterns of computer use we associate with the microcomputer revolution were well established in the early 1970's. In a 1973 article on Minicomputers published in Datamation, it was noted that Ken Olson, president of DEC, was a long-time owner of a home computer, and that DEC had a policy of offering PDP-8 systems at a bargain prices to DEC employees who wanted their own home computers.
In the same month, DEC began the EduSystem marketing program, in an effort to enlarge their already large share of the educational minicomputer market. At the time, Hewlett Packard was their only significant competitor in this market segment. The configurations offered ranged from the EduSystem 5, a PDP-8/E with 4k, a teletype, and BASIC pre-loaded in core (with paper tape for re-loading, if needed), to the EduSystem 50, a large PDP-8 based 16-user time-sharing system with disk and DECtape drives.
In February 1975, DEC followed up on the EduSystems with the CLASSIC, the CLASSroom Interactive Computer, a PDP-8 based workstation built into a desk with an RX01 dual 8-inch diskette drive, a 32K PDP-8, and a VT52 video terminal on top, with an advertised cost of $200 a month. This machine was touted as "a new era in educational computing", and it marked a change in DEC marketing, recognizing that computers were now inexpensive enough to put one on each desk.
The PDP-8/A, announced in May 1974, was an essential prerequisite to this change. This was announced as DEC's entry into the microprocessor arena, although the announcement was somewhat misleading. DEC did have a cooperative agreement with INTEL at the time, and they were in the process of developing an 8080 based 8-bit system, but the PDP-8/A was not based on a VLSI microprocessor.
The Intel 8080 was announced in July 1974, priced at $360 in unit quantity. In the same month, the DEC Components Group offered the PDP-8/A CPU plus 1K of RAM at a price of $572 in quantities of 100 for December delivery. In November, the Components Group advertising used the slogan "Unbundling the world's most popular minicomputer system" and advertised this configuration, now renamed the Kit 8/A, as "minicomputer power at microcomputer prices."
The PDP-8/A had a 1.5 microsecond memory cycle time, with a 12-bit add from memory taking 3 microseconds. In comparison, 8080 instructions required a minimum of 2 microseconds and an 8-bit add from memory took 3.5 microseconds, with a 16 bit add being considerably slower. The faster clock rates supported by later versions of this chip would quickly erase this margin, but initially, the PDP-8/A was quite competitive.
The PDP-8/A CPU was less expensive than the PDP-8/E CPU because, by using MSI chips and printed circuit boards half-again as large as the PDP-8/E boards, it was possible to pack a full PDP-8 CPU onto a single circuit card. The PDP-8/A wasn't designed for speed, though, so the PDP-8/E CPU, with its 1.2 to 1.4 microsecond memory cycle time, remained in production and was used in high-end PDP-8/A systems, where its slight speed advantage was useful.
The new larger PDP-8/A board format, combined with use of MSI technology, allowed the introduction of two-board 8K and 16K core memories instead of the 3 board 4K and 8K memory subsystems of the PDP-8/E, and it allowed the construction of a number of multipurpose I/O boards supporting common peripheral mixes on a single board.
By the time the MITS Altair hit the market, with a kit price of $439 in March 1975, DEC's primary thrust in marketing the PDP-8 was towards the OEM market. Although it was still available to end users, DEC's end-user advertising was clearly focused on their 16-bit PDP-11 and their 36-bit DECsystem 20.
In August 1976, DEC published an 8 page slick advertisement entitled the "Illustrated Guide to OEM Success". At this point, DEC was a major supplier of video terminals and DECwriters, and the PDP-8 was still a big seller, with annual sales of 10,000 systems a year and over 30,000 machines installed.
The PDP-8/A 100, a machine with a short backplane, sold for $1835, while the 8/A 620, with a 20 slot backplane and a PDP-8/E CPU, was selling for $3,950. The OS-8 operating system was offered for $400, while the RTS/8 real-time operating system, very important in the OEM market, was offered for $500.
DEC had worked cooperatively with Intel during the development of the 8080, and when it became clear that a PDP-8 on a chip was a real possibility, DEC cooperated with Intersil and later Harris in developing the 6100 and 6120 chips.
These first came to market in July 1977 with the DECstation or VT78. This was advertised as "a big computer system that's small enough for anyone", using a series of photos showing a small girl unpacking one. (These must have been faked, there's no way she could lift some of those parts!)
The basic VT78 was packaged in the same package as the VT52 terminal, but it had an Intersil 6100 CPU, 16K of MOS RAM, 2 serial ports and a diskette interface packed into the box. The machine was sold in a bundled configuration that included an RX01 dual 8-inch diskette drive mounted in a teacart that could support the terminal. The advertised price was $7995, discounted to $5436 in lots of 50 or more.
The VT78 was a full scale member of the PDP-8 family, able to run OS/78 (a somewhat arbitrary renaming of OS/8) and RTS/8, and it was advertised as supporting FORTRAN IV and BASIC. At this time, the standard OS/8 distribution also included the PAL8, SABR and RALF assemblers, PIP, the peripheral interchange program, the TECO and EDIT text editors, and an assortment of other utilities.
At the time of the VT78 introduction, DEC's PDP-8 was well established as a small business machine. A September 1977 Datamation survey showed that the DEC DataSystem 310, a PDP-8/A with 16K 12-bit words of core and supporting the DIBOL programming language was a significant player in this growing market segment. It is clear that the VT78 was well positioned to take advantage of this.
Even in 1978, when microprocessor-based controllers were the obvious choice for most OEM applications, DEC was still listing 12-bit systems in their advertising for that market, although only in passing. Internally, DEC began using large numbers of 6100 and 6120 chips as peripheral controllers on devices for the larger PDP-11 and VAX families of computers, and presumably, others also used these chips, but documentation of this is difficult to find.
DEC's Word Processing Computer Systems Group, with the software product WPS-8, gave the PDP-8 family its last big market. The advertisements for this group began in February 1979, and they rarely mentioned the hardware or software being sold; instead, the emphasis was on ``word processing solutions''.
The word processing market of the early 1980's was large, and it was in turmoil, with many large businesses still using Selectric typewriters and manual methods, while others moved to mainframe or minicomputer based timesharing systems and yet others were trying microcomputer based systems. In this environment, a relatively mature product like WPS-8, supported on a microcomputer based workstation like the VT78 was a very reasonable choice.
The market was large enough that DEC followed up on the VT78 in 1980 with a new machine, the DECmate; This was based on the 6120 chip, packaged in the same case as the ubiquitous VT100 terminal. As with the VT78, this was not an open system. It came with 32K words of RAM, an RX02 dual 8-inch diskette drive, and a printer port. The only options offered, other than variations in packaging, were a second dual diskette drive, a communications board, and an optional disk controller for DEC's RL02 rack-mounted disk drives.
OS/78 was still being sold with the DECmate, and with a large disk such as the RL02, this was clearly a big enough system to handle the basic data processing needs of many small businesses. Nonetheless, much of the market was made up of WPS users, and many DECmates were sold with a keyboard with special WPS keycaps instead of the normal ASCII markings used on the VT100.
In 1982, DEC came out with the DECmate II; although a closed architecture machine, this offered more options than the original DECmate, and it used RX50 5.25-inch diskette drives instead of the older 8-inch RX02 drives. This machine had support for an optional RX02 drive, allowing customers to convert older PDP-8 based documents to the new smaller diskette format, and 5, 10 and 20 megabyte MFM hard drives were available, typically using Seagate hardware.
At the time the DECmate II was introduced, there was a large installed base of small business systems based Z80 processors running CPM, and there was an emerging marked for systems compatible with the 8086 and MS/DOS. With the DECmate II, DEC offered support to both communities, using optional coprocessor boards.
The APU option allowed a Z80 and 64K of RAM to be added to a DECmate II, allowing full support for CPM. The XPU board option allowed for an 8086 and 512K of RAM. DEC's MS/DOS environment used the PDP-8 compatible 6120 processor as a parallel I/O processor, allowing for very fast input-output, but unfortunately for DEC's market share, this made it incompatible with the IBM PC for applications that bypassed the BIOS and dealt directly with hardware.
Two years later, in 1984, DEC introduced the last of its PDP-8 compatible machines, the DECmate III. This machine remained in production until 1990, with almost all of the production going into the word processing market.
It has been 5 years since DEC stopped producing PDP-8 compatible computer systems, but huge numbers are still in service, mostly in industrial control systems sold through OEMs. Today, PDP-8 systems control nuclear reactors, elevators, milling machines, and other automated equipment built from the late 1960's through the 1970's.
There are two or three companies selling spare parts and service for PDP-8 systems; the prices seem quite high for such apparently obsolete hardware, but if you imagine the problem from the point of view of, for example, a nuclear reactor operator, they make sense. It is far easier to maintain a old control system almost indefinitely, paying wildly inflated prices for spare parts, than it is to go through the complex and even more expensive certification process required by the nuclear regulatory agencies for any significant change to such a system.
Alternately, consider the perspective of an automated machine tool owner. Should the PDP-8 CPU running the control system break, would you rather pay hundreds of dollars for a reconditioned but obsolete CPU, or tens of thousands of dollars for a new machine tool that uses a modern control system?
There are probably a number of WPS systems still in use, but these are almost certainly being thrown out at a high rate. PC and Mac based word processing systems are very good, and the cost of converting to such a system is so low that there is little economic incentive to keep old systems that are incompatible with everything else on the market.
To date, old PDP-8 systems have not developed any special collectors market. When systems become available, they are frequently available to anyone willing to haul them away. Old DEC mounting racks, particularly those with casters, have a fair market value of between $10 and $40, but the electronics they contain rarely brings in more than a few dollars at surplus sales.
There are perhaps 100 active PDP-8 collectors, judging by the names that show up irregularly in the USENET newsgroup alt.sys.pdp8 and the pdp8-lovers@ai.mit.edu mailing list. Most of the machines held by this group of collectors are DECmates and various OMNIBUS machines such as the PDP-8/E and PDP-8/A, but there are a fair number of older PDP-8 systems in good working order.
The PDP-8/S is perhaps the least common PDP-8 CPU currently in the hands of collectors. With a production run of only 1024, there were never many, and because of its low speed, most were discarded quickly when faster models became available at lower prices. Some of the older PDP-8 peripherals are quite hard to find. TU56 DECtape drives are particularly rare, while RX01 and RX02 diskette drives are fairly commonplace.
For those who cannot find PDP-8 hardware, a number of good emulators are available. Most are available on the world-wide-web; a good starting point is http://www.cs.uiowa.edu/~jones/pdp8/. Another good starting point is the FTP archive maintained at sunsite.unc.edu; the directory path is a long one, /pub/academic/computer-science/history/pdp-8.
The dates cited here, particularly in the 1970's are all references to the month of publication of the press release or advertisement in Datamation magazine. This magazine is a good general source for press releases and computer advertising through the introduction of the PDP-8/A, but I have yet to find a good source to document the details of the DECmate era.