Four-Phase Systems AL1

The AL1 was an early 8-bit microprocessor designed by Four-Phase Systems and first run in April 1969. It is the first single-chip central processing unit (CPU) to be produced,^[1] pre-dating the Intel 4004 by two years. Although it could be used as a stand-alone general purpose CPU, Four-Phase did not use it in this fashion at the time. Instead, they used three AL1's in a bit-slice system to produce a 24-bit minicomputer, the System IV/70. The company never advertised the AL1 as a product and did not sell it to other customers, the 4004 was the first design to be sold in standalone form.

AL1

General information
Launched	April 1969[a]
Designed by	Four-Phase Systems
Common manufacturer	Cartesian
Performance
Max. CPU clock rate	1 MHz
Data width	8 bits
Address width	16 bits
Architecture and classification
Application	System IV/70
Technology node	10 μm
Physical specifications
Transistors	~4000
Package	40-pin dual in-line package
Socket	DIP40
Support status
Unsupported

In 1990, Texas Instruments began to enforce patents on the basic concept of a microprocessor, which they had initially filed in 1971. These plans were upset when a patent was granted to another designer, Gilbert Hyatt. The resulting flurry of lawsuits led to the AL1 becoming famous in 1995 when Lee Boysel built a small computer to demonstrate his design incorporated all of these concepts using a chip manufactured two years before TI's design and a year before Hyatt's.^[b]

History

Fairchild work

Lee Boysel started work at Fairchild Semiconductor in 1966 after working at several other company's semiconductor departments. At Fairchild he worked on MOS design, which at that time was a very new concept. Over the next two years he developed several new MOS chips, including a 256-bit static RAM and an 8-bit adder. The adder was the first integrated circuit with over 100 gates.^[2]

Four-Phase forms

Boysel left Fairchild in October 1968 to start his own company, Four-Phase Systems,^[3] which incorporated in February 1969.^[4] Corning Glass, who was at that time investing in the emerging semiconductor market, provided start-up funding of $2 million. Four-Phase was what would today be known as a fabless designer, using another Fairchild spin-off, Cartesian, as a semiconductor foundry.^[5]

The company name refers to the way the individual transistors in the MOS circuits were powered by the clock generator. This design allows the transistors to be made smaller, although it requires some additional design-time complexity to arrange them. Smaller transistors translates to cheaper ICs, which he intended to use to build lower-cost computers that would compete with systems from Data General and the mid-range machines from IBM.^[5]

AL1 and System IV/70

The first engineering samples of the AL1 were available in 1970. Three AL1s were each paired with a read only memory (ROM), a random logic chip that ran the bus, and the external clock generator onto a board to produce the 24-bit CPU for the System IV/70.^[5] The company never sold the AL1 or the complete processor boards, nor did they file patents on the design, but they did publish an article in the April 1970 issue of Computer Design magazine describing the design and how it could be used to build a simple computer.^[6]

The IV/70 was successful, selling for about half the equivalent machines from other major vendors. By the mid-1970s they had sold 350 systems, and by 1980 they had 5000 employees and annual sales over $250 million. When sales of the AL1-based systems dwindled by the late 1970s, the firm introduced a series of Unix-based computers based on the Motorola 68000, which led to Motorola purchasing the company in 1982.^[2]

TI lawsuit

In 1970, Datapoint corporation (then known as CTC), sent a contract to Texas Instruments (TI) and Intel asking them for a IC that implemented their entire Datapoint 2200 smart terminal circuitry in a single chip. Intel responded in late 1971 with the Intel 8008. TI had initially designed a three-chip solution, but when Datapoint passed them a copy of Intel's proposal in late 1970, they redesigned it in a single-chip form as the TMX 1795. Several delays at Intel meant they didn't deliver their 8008 until after TI had sampled the 1795 in mid-1971. Datapoint ultimately decided to use neither design, due to performance issues, and TI abandoned the design in 1972.^[7]

TI nevertheless filed several patents on the design and these were awarded over a long period of time. In 1986, TI began demanding royalties from every company making microprocessors, amounting to 2 to 3% of the cost of every computer.^[8] This launched a series of lawsuits in what became known as "TI vs. Everyone".^[9] Further confusing the issue was the 1990 award of a December 1970 patent on a "Single Chip Integrated Circuit Computer Architecture" to Gilbert Hyatt, which pre-dated TI and apparently rendered their patents moot.^[10]

This new wrinkle had the side-effect of causing all of the law firms involved to begin looking for additional prior art. This led to Boysel being called as an expert witness by no less than twenty-five of the law firms involved, as his article in April 1970 pre-dated all of these claims. TI admitted that they were aware of the 1970 article, but, as in the case of Hyatt, they claimed that it wasn't a "real computer". This turned out to be a bad idea, as it made Boysel mad and determined to prove them wrong.^[6]

In order to drive home the point that the AL1 was indeed a system that incorporated all of the ideas TI was claiming, Boysel built a tiny computer system consisting of an AL1, with its 1969 manufacturing date still clearly stamped on it, mounted on a plexiglass card. This was connected to RAM, ROM, an I/O unit housed in cartridges from the Nintendo Entertainment System.^[11] The ROM held a simple program from Datapoint for customer lookups called WSTR, which had been given to TI and used to validate the TMX 1795. The only difference was that the AL1 ran the program ten times faster than the TI or Intel chips.^[12]

Boysel noted that when he described the demo just prior to the trial:

Faces went white; the place turned into chaos. They just realized that they’d lost, and it was over.^[6]

Design

The AL1 was fabricated using the then-current 10 micron process, the same used by the Intel 4004 and other designs of the era. However, the four-phase logic allowed them to place over 1,000 gates (~4000 transistors) on a 130 by 120 mil (100 mm²) chip. In comparative terms, the chip was as complex as the Intel 8008 but the same size as the 4004, and also ran at a much higher speed of 1 MHz, compared to 740 kHz for the 4004 or 500 kHz for the 8008.^[5]

An additional improvement was in the packaging. The 4004 was packaged in a 16-pin dual in-line package (DIP) and the 8008 in a 18-pin DIP. This is too few pins to allow data and memory addresses to be expressed at the same time, meaning that accessing memory requires several cycles. The AL1 was packaged in a 40-pin DIP, which allowed it to access memory in a single cycle. This would be a common size for later CPU designs. The wider busses and higher operating speed meant the AL1 was almost 10 times the speed of any other design of the era.^[5]

The AL1 was an 8-bit design with eight processor registers, one of which was used as the program counter, feeding an 8-bit arithmetic logic unit (ALU).^[5]

Notes

1. The CHM states the first samples were in March 1969, not April.
2. Another often quoted contender for first is the MP944 inside the F-14 CADC. This began design in 1968, but working samples were not delivered until June 1970, after the AL1 was already in use and had been disclosed in April.