The Fairchild 9440 MICROFLAME, also known as the F9440 and μFLAME, was a 16-bit microprocessor introduced by Fairchild Semiconductor in 1977. The 9440 implemented the Data General Nova 2's instruction set in a single-chip 40-pin DIP. The name "MICROFLAME" was part of a wider branding exercise called "FIRE", which was a development software system.
An updated version, the 9445, was announced in 1978 but did not reach market until late 1981. By this time the 16-bit designs were being surpassed by 32-bit designs and hybrids like the Motorola 68000, and Fairchild began turning their attention to their own 32-bit Fairchild Clipper design. The underlying core of the 9445 was also used to implement the 9450, which used new microcode to implement the MIL-STD-1750A instruction set.
The 9440 and 9445 were subject to constant lawsuits from Data General (DG) that dragged on both companies. DG finally settled all ongoing litigation in September 1986 by paying Fairchild over $52 million. Fairchild was purchased and sold several times during this period, ending up at National Semiconductor in 1987. National ended production of the 9445, leaving several users in the lurch. A final version, the IDC9445, was introduced to fill this need.
History
Nova
The Data General Nova was introduced in 1969, implemented using individual integrated circuits (ICs) mounted on a 15x15-inch printed circuit board. In order to lower design complexity, and thus board size and cost, the arithmetic logic unit (ALU) was only 4-bits wide, implemented using a single 74181 IC. This meant it required four machine cycles to complete a 16-bit instruction, but it also allowed the system to be much less expensive than competing minicomputers from Digital Equipment Corporation (DEC) or Hewlett-Packard. The NOVA was very successful, propelling DG into the second-place position behind DEC in the minicomputer market during the 1970s.
In 1970, DG introduced the SuperNOVA, which featured a full 16-bit wide ALU using four 74181's in bit-slice fashion, and thus ran about four times as fast as the original NOVA. This was further boosted by other changes, including faster core memory and optional semiconductor memory, making the SuperNOVA the fastest mini for some time. This also meant there were two different central processing unit designs implementing the same underlying instruction set architecture (ISA). As the development of both designs continued, the two were modified so the faster version could be dropped into existing machines originally running the lower-speed hardware. This led to the NOVA 2, 3 and 4 series.
microNOVA
Improving semiconductor fabrication processes, especially through the 1970s, put increasing pressure on the individual-IC CPU designs like the NOVA. Initially, only 4-bit and then 8-bit CPUs could easily be manufactured on a single chip, but by the mid-1970s, 16-bit designs were appearing. In 1973, National Semiconductor introduced the IMP-16, which implemented a NOVA-like system in a set of only five IC's. The next year their PACE reduced that to a single IC. Several similar designs appeared during this period, including the Texas Instruments TMS 9900, which implemented their TI-990 minicomputer, and the Intersil 6100, a single-chip version of the PDP-8.
DG needed to respond to these systems, and began development of the microNOVA mN601, a single-chip design that implemented the NOVA 3 instruction set. It also added hardware single-precision multiply and divide, formerly an optional add-on for NOVA systems. It was announced in early 1976, and sold to anyone who wanted one - they could be purchased as a single chip, a CPU card with support chips, or a complete packaged NOVA machine. The individual chips sold for $225, or $95 in lots of 100. Later, however, DG stopped direct sales of the 601 in favor of selling complete systems.
The mN601 was implemented using NMOS technology, and thus required four separate input voltages from the power supply. It also lacked an internal clock, and required the system to provide a two-phase clock signal to drive it. It ran at fairly low speeds, normally using a 240 ns clock (~4.2 MHz), completing instructions in 2.4 to 10 µs. This, along with limitations on the bus performance, meant the system performed about half as fast as an original Nova.
Throughout the evolution of the Nova line, the systems had been built primarily using discrete ICs from Fairchild Semiconductor, a leader in IC design during the 1960s. This relationship ended with the microNOVA, which was fabbed at a new DG-owned factory in Sunnyvale, California.