HP(Hewlett-Packard) 2100 | 1964 ~ 1992

 The HP 2100 is a series of 16-bit minicomputers produced by Hewlett-Packard (HP) from the mid-1960s to the early 1990s. This series sold tens of thousands of units over 25 years, and by the 1970s, HP had become the fourth largest supplier of minicomputers.

The design began at Data Systems Inc. (DSI), originally known as the DSI-1000. HP acquired the company in 1964 and integrated it into the Dymec division. The first model, the 2116A, was released in 1966, utilizing integrated circuits and magnetic core memory. Over the next four years, models A through C were released with various memory and expansion types, along with cost-effective models 2115 and 2114. All these models were replaced by the HP 2100 series in 1971, which was later replaced by the 21MX series in 1974 when semiconductor memory was introduced.

These models were also packaged as the HP 2000 series, allowing the 2100 series machines to run the BASIC programming language in a multi-user time-sharing mode with optional components. HP Time Sharing BASIC gained popularity in the 1970s, and many early BASIC programs were written on this platform, notably the famous Star Trek, which became popular during the early home computer era. The People's Computer Company published their programs in HP 2000 format.

The introduction of the HP 3000 in 1974 provided advanced competition for the 2100 series, and by 1977, the entire line was renamed HP 1000, establishing itself as a real-time computer. In 1979, a significantly revamped version using CMOS large-scale integration was introduced as the 1000 L series, which was the first model to break backward compatibility with the previous 2100 series expansion cards. The final upgrade was the A series, which featured a new processor with over 1 MIPS performance, with the last A990 released in 1990.

HP established DynaC in 1956 to work on development projects that the company typically did not undertake. Their original logo was simply an upside-down HP logo, forming a shape close to "dy," which inspired the name. After discovering in 1958 that Westinghouse owned the trademark for that name, they changed it to Dymec. The company was integrated internally in 1959 and became the Dymec division, renamed the Palo Alto division in November 1967.

Dymec originally created various products for the HP family but gradually transformed into a systems integrator, building test equipment and similar systems used by HP. In 1964, Kay Maglby and Paul Stoft began experimenting with using PDP-5 and PDP-8 computers as controllers for complex test systems, but they felt these machines needed to be modified to meet their needs. At that time, Digital Equipment Corporation (DEC) was still a small company and a potential acquisition target. David Packard deemed negotiations with Ken Olsen difficult, and these plans fell through.

Looking for other designs, Packard was led to Data Systems Inc. (DSI), which was part of Union Carbide. When Packard asked how Union Carbide came to own a computer company, HP Labs manager Barney Oliver replied, "We didn't ask for an answer to that question." Bill Hewlett initially did not consider developing a "minicomputer," but when Packard restructured it as an "instrument controller," the deal was approved.

DSI was acquired in 1964, initially established within Dymec, with four of the original five DSI employees and several staff from HP's instrumentation division. The computer group later moved to a separate office in Cupertino, California, becoming the Cupertino division in a building purchased from Varian Associates.

The new division led by Maglby completed the design of the 2116A, which was demonstrated at the Joint Computer Conference held in San Francisco from November 7 to 10, 1966. It was one of the earliest 16-bit minis to hit the market, noted for its highly scalable design and real-time support as an "unusual new instrument computer." The system featured an oversized cabinet capable of accommodating up to 16 expansion cards, with additional expansion possible through external expansion cages for up to 48 cards.

The system was launched with 20 different instrumentation cards, including "counters, nuclear scalers, electronic thermometers, digital voltmeters, AC/ohm converters, data amplifiers, and input scanners." Additional sets added input/output devices such as tape drives, printers, punch cards, and paper tape. Real-time service was provided by assigning fixed interrupt vectors to each card slot, calling the appropriate device driver.

As the machine entered the market, it became clear that it was selling more rapidly into the business data processing market. This led to the introduction of the 2115A in 1967, a low-cost product with many expansion features removed for commercial users. In 1968, an even simpler version, the 2114A, was released, which had only 8 slots, allowing the power supply to be integrated into the main unit. The 2115 and 2114 also lacked the extensive DMA control capabilities of the 2116, removed some mathematical operations, and operated at slightly slower speeds.

These were the original models using core memory and hardwired CPUs:

2116A: 10 MHz clock, 1.6 microsecond (μs) cycle time. Typically supplied with 4k words, expandable internally to 8k or externally to 16k. The chassis includes 16 I/O slots and is also expandable. Weighs 230 pounds (104 kg). Introduced in November 1966, marking HP's first use of integrated circuits.

2116B: Supported a new 32k memory expansion option. Weighs the same as above. Introduced in September 1968.

2116C: Used smaller cores, allowing the entire 32k to be mounted in the main unit. Introduced in October 1970.

2115A: A short-lived cost-reduced version with DMA and some mathematical functions removed, featuring only 8 I/O slots and operating at 8 MHz clock with a 2.0 μs cycle time. Weighs 65 pounds (29.5 kg) but requires a bulky external power supply, bringing the total weight to 160 pounds (72.6 kg). Introduced in November 1967.

2114A: A simplified version of the 2115 with a new front panel and internal power supply. The front panel features non-mechanical touch switches. Weighs 95 pounds (43.1 kg). Introduced in October 1968.

2114B: A 2114A with a new front panel featuring a single DMA channel and illuminated push buttons. Introduced in November 1969.

2114C: A 2114B with a maximum of 16k memory. Introduced in October 1970.

In November 1968, the company launched the 2000A time-sharing system, which later became known as HP 2000/Access. It was based on the 2116B (an 8k core memory expanded version of the 2116A) and ran HP Time Sharing BASIC, using a separate 2114 as a terminal server. T-S BASIC allowed for the creation of multiple user accounts, enabling up to 16 users to log in simultaneously.

Subsequent models of the 2000, from B to F, were released using new CPU versions. Some models used a low-cost version of the same CPU as the terminal server. For example, the 2000F used the 2100S as the main CPU and basic storage controller, while the 2100A operated as the terminal server. Models B, C, and F were dual-processor systems. The final version, the 2000F, utilized the 2100S and 2100A CPUs, with the 2100A connecting up to 32 serial terminals via a serial multiplexer interface.

Despite relatively high costs—$105,000 for the 2000F in 1974, equivalent to about $649,000 in 2023—it was the first minicomputer to offer time-sharing BASIC, gaining significant popularity in the early 1970s.

The HP 2100 computers were continuously improved throughout the 1970s, with new models maintaining compatibility with the existing 2116 in software and expandability.

In the early 1970s, Fred Allard was tasked with designing a new magnetic core memory system at AmenTech's memory core division. Using a new 18-mil core, reduced from 22 mils, he was able to mount 8kW memory on a single expansion card using a single sense/suppress line. This was utilized from the 2116C model onward.

In 1971, the updated 2100A replaced the existing 211x lineup. The overall system was similar to previous models and still based on core memory. Physically, it resembled the 2114 most closely, with an integrated power supply and limited internal expansion. However, the CPU was restructured with user-programmable microcode, adding hardware multiplication and division, and a simple memory protection system that triggered high-priority interrupts when activated. A two-channel DMA controller provided higher throughput. An optional floating-point device was also available. The front panel buttons were illuminated by small incandescent bulbs due to usage. The use of a switch-mode power supply allowed the chassis to be much smaller than previous models.

By 1972, HP had shipped 4,500 minis, becoming one of the largest companies in the market. That year, they merged the Mountain View division, which manufactured magnetic tape drives, with the Cupertino division to create the Data Systems Division (DSD). At this point, the HP 3000 project faced serious issues, and in February 1973, Packard sent Paul Ely to take over the division. Sales of the 2100 series remained strong, with the 6,000th unit shipped in August 1973, the 8,000th in February 1974, and the 10,000th in February 1975. They were declared obsolete in 1978.

The 2100A typically had 4kW but was expandable to 32kW, with 14 I/O slots expandable to 45. Introduced in 1971.

The 2100S included floating-point options, a time reference generator for interfacing with time-dependent hardware, and telegraph interface cards, based on the 2100A. Introduced in 1973.

In 1972, the division decided to move to 4kbit SRAM memory chips instead of core. This was twice as fast as the low-cost core at the time and allowed for the construction of a 32kword machine in a form similar to the 2114. At the time of this decision, 4kbit SRAM was not actually available, so initially, experimental machines using the core of the 2100 series were used, later replaced by 1kbit SRAM. Initially, it seemed that 4k parts would not be available until the rest of the machines were upgraded, so the company investigated various manufacturers, some of whom were actively developing 4k parts. Notably, Intel, Mostek, Motorola, and Texas Instruments suggested that these parts would be available in volume by late 1974.

Using this memory in the same expansion chassis as previous systems would allow memory capacity to expand into the megaword range. However, due to the CPU's 15-bit addressing, the size was limited to 32kword. To address the demand for larger systems, HP developed the "Dynamic Mapping System" or DMS. DMS expanded the address format from 15 bits to 20 bits, allowing for a maximum of 1,048,576 words, representing a dramatic expansion of the original system.

These changes led to the introduction of the first 21MX series machines in 1974. The "21-M" processor and "21-X" memory meant that the memory control system was separated from the CPU, providing flexibility in production if selected 4k parts were changed. This was effectively utilized, allowing the machines to mix and match Motorola or Texas Instruments parts, and higher-density systems could support normal 21-X/2 controllers by replacing them with 21-X/1 based on 16-pin Mostek parts.

The entire 2100 series was replaced by the first 21MX series machines in 1974. When using the highest density parts, the new machines could support up to 1.2MB. When the CPU was further upgraded in 1976, the new models became known as the "21MX E series," while the original models were retroactively named the "21MX M series." The main difference between M and E was that E used clever timing to double the speed. As part of the 21MX series, a wider variety of machine styles were released, including small systems with 4 slots and larger systems with 9 or 14 slots.

The 21MX series featured memory management devices and semiconductor memory, expandable to a maximum of 1,048,576 words (1 megaword). The bit display of the front panel buttons used small red light-emitting diodes (LEDs) instead of the incandescent bulbs used in previous versions.

M series – 2105A, 2108A, 2112A (blue line on the front panel)

E series – 2109A, 2113A (yellow line on the front panel; E stands for expansion)

F series – 2111F, 2117F (red line on the front panel; F indicates a separate 2U chassis for the floating-point processor)

The 21MX ran the HP RTE (real-time) operating system. They initially started as refrigerator-sized 19-inch rack-mounted systems with lights and switches on the front panel. The last models used a single-chip processor placed under the desk, utilizing a console terminal instead of a front panel.

The new L and A series models featured HP-IB interface capabilities, but, like all HP systems at the time, the blinking LED lights on the front panel were removed. Despite customer demands for real-time capabilities and HP R&D efforts, the RTE-A OS using installable real-time cards was not as suitable for real-time tasks as the RTE of the 21MX. This was a significant reason why this computer did not easily fade away. Many companies used real-time tasks to perform measurements, control processes, and operate pumps, heaters, valves, and adjust motor speeds.

By the end of 1977, the entire line was renamed "HP 1000," becoming the "HP 1000 M series" and "HP 1000 E series." The following year, the "HP 1000 F series" was introduced, which added a floating-point unit to the E series. By 1978, the success of this line made HP the fourth largest manufacturer in the minicomputer field, following DEC, IBM, and Data General.

In 1980, the "HP 1000 L series" was introduced, featuring a new processor based on HP's large-scale integrated silicon-on-sapphire process. Additionally, the expansion cards included their own processors, allowing them to access main memory and perform input/output operations, relieving the CPU of some burdens. This was the first break in compatibility with previous models but significantly improved overall performance. These models did not have memory management devices.

HP1000L silicon-on-sapphire (SOS) CPU and I/O processor

The L series was replaced in 1982 by the "HP 1000 A series," which included a new "Lightning" CPU design that reached 1 MIPS and a faster "Magic" CPU that reached 3 MIPS. Various models were produced, including desk-side towers and rack-mounted systems of various sizes. In 1986, the low-cost A400 model was introduced, and in 1990, the last high-end A990 was released. Each model could address up to 32MB of RAM.

1981:

A600: Based on the Am2900 bit-slice processor, 1 MIPS, 53kFLOPS, code name: LIGHTNING

A600+: Based on the Am2900 bit-slice processor, supports code and data separation, and optional ECC (error correction) memory.

1982:

A700: Based on the AMD AM2903 bit-slice processor, supports optional hardware floating-point processor, 1 MIPS, 204kFLOPS, microprogramming, and optional ECC memory. Code name: PHOENIX

1984:

A900: Provides a pipelined data path, 3 MIPS, 500kFLOPS, supports ECC memory. Code name: MAGIC

1986:

A400: The first single-board CPU with four serial lines; CPU manufactured using VLSI technology in a CMOS-40 process, with 512KB RAM on the board. Code name: Yellowstone

1992:

A990: CPU implemented with two 208-pin CMOS application-specific integrated circuits (ASICs), supports 298 instructions, and can address up to 512MB of memory.