The first commercial microprocessor
By Scott Hamilton
Senior Expert in Emerging Technologies
This year we celebrate the 50th anniversary of the computing era. In 1971 Intel corporation released the world’s first commercial microprocessor, the Intel 4004. It sold for $60 and was the first of many central processing units to be released by Intel.
The Intel 4004 was the longest running processor in the history of Intel, being mass produced for a period of 10 years before being discontinued. The design was implemented with the Metal-Oxide-Silicon (MOS) silicon gate technology and took only two years to develop. The microprocessor’s developer, Federico Faggin, led the project from its inception to its completion in 1971. In 1969, Marcian Hoff designed the architecture, which is the proposed set of instructions; how the CPU communicates with displays, memory and input devices; the general layout of the inputs and the internal format of the data and instructions. Masatoshi Shima also worked on the architecture and later the logic design. You can think of the architecture like the floor plan of a house, and the logic design is like the engineering diagrams that show the electrical wiring, plumbing and other finer details.
There is one final step in the design of a microprocessor and that is the layout. The layout is similar to the final steps in designing a house; you can think of it as the materials list and construction blueprints. In microprocessors this is the hardest of the design steps. The layout engineer, Faggin, had the task of figuring out how to fit all the logic designed by the other engineers onto the integrated circuit. This was a feat that no one in history had ever accomplished. Faggin managed to design and build this custom chip with intel, containing 2300 random logic transistors. This was the largest number of transistors ever placed on a single piece of silicon.
You might wonder what the big deal was with fitting the logic transistors onto a single chip. Prior to this accomplishment, CPUs were built on several separate circuit boards that could fill a rack about the size of an average refrigerator. It was not only the size that was impacted by this intricate design; it also allowed the processor to run nearly five times faster and cost nearly fifty percent less to produce. He used technology he had originally developed as an engineer at Fairchild Semiconductors in 1968, where he was responsible for building the first commercial integrated circuit, the Fairchild 3708, which was an 8-bit analog multiplexer, a primary component in analog computers.
Faggin moved from Fairchild semiconductors to Intel in 1970 and was allowed to take his buried silicon-gate bootstrap load with him. It was this obscure and buried project that made the design of the microprocessor possible. Faggin admits that there were several engineers at the time capable of designing the logic circuitry of a small CPU, but he felt his bootstrap method was the only plausible method to provide the high speed and high-density circuitry required for a single-chip CPU. Ironically the chip was not designed for an Intel product, but rather as an engineering project for a Japanese manufacturer that were hoping to build a family of electronic calculators.
The Intel 4004 was a 4-bit processor which meant that it could address up to 16 memory locations and manipulate the data between them. It was first produced in April 1971 and introduced to the market in November 1971. It was quickly followed by the first 8-bit processor, the Intel 8008, in April 1972. The 8008 could address up to 256 memory addresses; as you can see, each bit exponentially increases the amount of memory accessible by a processor. The Intel 4004 was outright amazing in its day, with speeds of 740kHz or 740,000 cycles per second and address 640 bytes of RAM and a transistor count of 2300 and size of 10 micrometers. However, compared to the behemoth processors of today, they were tiny. The latest Intel Ice Lake processors contain 290 million transistors at 10 nanometers in size (1000 times smaller), operate on a 64-bit address space, meaning 18,446,744,073,709,551,616 memory addresses, and operate at a frequency of 3.7GHz (100 thousand times faster). To me it is an amazing accomplishment in merely 50 years.
Until next week stay safe and learn something new.
Scott Hamilton is a Senior Expert in Emerging Technologies at ATOS and can be reached with questions and comments via email to sh*******@te**********.org or through his website at https://www.techshepherd.org.
