preview.tinyurl.com/8yvr4k5 -> arstechnica.com/business/news/2011/11/the-40th-birthday-ofmaybethe-first-microprocessor.ars
dual in-line package, was an entire CPU packed onto a single integrated circuit (IC). At a bare minimum, a CPU is an instruction decoder and an arithmetic logic unit (ALU); the decoder reads instructions from memory and directs the ALU to perform appropriate arithmetic. Prior CPUs were made up of multiple small ICs of a few dozen or hundred transistors (and before that, individual transistors or valves) wired up together to form a complete "CPU." The 4004 integrated the different CPU components into one 2,300-transistor chip. Since its founding in 1968, Intel was a memory company, making various kinds of RAM, boasting some of the fastest and highest density memory in the industry. It wasn't in the business of making CPUs or logic chips. Nonetheless, Japanese electronic calculator company Busicom approached Intel in 1969, asking the memory company to build a new set of logic chips for its calculators. Busicom proposed a fixed-purpose design requiring around a dozen chips. Busicom had designed the logic itself, and even verified that it was correct; Ted Hoff, manager of Intel's Application Department, realized that the design could be simplified and improved by using a general-purpose CPU instead of the specialized calculator logic that Busicom proposed. Hoff managed to convince both Intel and Busicom management that his approach was the right one. Work started six months later when Intel hired Federico Faggin in April 1970 to work on the project. As a memory company, it didn't have methodologies for designing or validating logic circuits. Intel's processes were geared towards the production of simple, regular repeating structures, rather than the highly varied logic that a CPU requires. Faggin's job was also made more complex by the use of silicon gate transistors. At the time, aluminum gates were standard, and while silicon eventually won out, its early development was difficult; silicon gates needed different design approaches than aluminum, and those approaches hadn't been invented yet. Nonetheless, Faggin was successful, and by March 1971 had completed the development work of a family of four different chips. Intel paid Busicom for the rights to the design, allowing the firm to sell and market the chip family. Branded as MCS-4, the chips started production in June 1971, before being advertised to the commercial markets 40 years ago today. Clumsy and cutting-edge The 4004 itself was a peculiar mix of cutting-edge technology and conservative cost-cutting. As an integrated CPU it was a landmark, but the design itself was clumsy even for 1970. Intel management insisted that the chip use a 16-pin DIP, even though larger, 40-pin packages were becoming mainstream at the time. This means that the chip's external bus was only four bits wide, and this single 4-bit bus had to transport 12-bit memory addresses, 8- and 16-bit instructions, and the 4-bit integers that the CPU operated on. Reading a single 16-bit instruction thus took four separate read operations. The chip itself had 740 kHz clock, using 8 clock cycles per instruction. It was capable of 92,600 instructions per second--but with the narrow multipurpose bus, achieving this in practice was difficult. As with the 4004, this was built for a third party--this time terminal manufacturer Datapoint--with Datapoint contributing much of the design of the instruction set, but Intel using its 4004 experience to actually design the CPU. In 1974, the company released the 8080, a reworked 8008 that used a 40-pin DIP instead of 8008's 18-pin package. Federico Faggin did much of the design work for the 8008 and 8080. In spite of these pioneering products, Intel's management still regarded Intel as a memory company, albeit a memory company with a sideline in processors. Faggin left intel in 1974, founding his own processor company, Zilog. Zilog's most famous product was the Z80, a faster, more powerful, software-compatible derivative of the 8080, that powered early home computers including the Radio Shack TRS-80 and the Sinclair ZX80, ZX81, and ZX Spectrum--systems that were many people's first introduction into the world of computing. Faggin's decision to leave Intel and go into business for himself caused some bad feeling, with Intel for many years glossing over his contribution. Nonetheless, he left an indelible mark on Intel and the industry as a whole, not least due to his decision to sign his initials, FF, on the 4004 die.
Intel The 8080 instruction set was then extended to 16 bits, with Intel's first 16-bit processor, the 20,000 transistor 8086, released in 1978. This was the processor that first heralded Intel's transition from a memory company that also produced processors into the world's leading processor company. In 1981, IBM picked the Intel 8088--an 8086 with the external bus cut to 8-bit instead of 16-bit--to power its IBM PC, the computer by which all others would come to be measured. But it wasn't until 1983, with memory revenue being destroyed by cheap Asian competitors, that Intel made microprocessors its core product. The processors of today continue to owe much of their design (or at least, the design of their instructions) to the 8086.
Sandy Bridge E CPUs using 22 billion transistors, a million-fold increase on 4004 and 100,000-fold on the 8086, the basic design elements are more than 30 years old. While the 4004 is widely regarded as the first microprocessor, and is certainly the best known, it arguably isn't actually the first. Texas Instruments' TMS 1000 first hit the market in calculators in 1974, but TI claimed it was invented in 1971, before the 4004.
Four-Phase demonstrated systems built using AL1 in 1970, with several machines sold by early 1971. However, at the time AL1 was not used as a true standalone CPU; instead, three AL1s were used, together with three further logic chips and some ROM chips. Intel and Cyrix came to blows in a patent dispute in 1990, with TI's patent being one of the contentious ones. To prove that TI's patent should not have been granted, Four-Phase Systems founder Lee Boysel took a single AL1 and assembled it together with RAM, ROM, and I/O chips--but no other AL1s or logic chips--to prove that it was, in fact, a microprocessor, and hence that it was prior art that invalidated TI's claim. As such, although it wasn't used this way, and wasn't sold standalone, the AL1 can retrospectively claim to have been the first microprocessor. The 4004 is, however, still the first commercial microprocessor, and it's the first microprocessor recognized and used at the time as a microprocessor. Simple and awkward though its design may have been, it started a revolution. Ted Hoff, for convincing Busicom and Intel alike to produce a CPU, Federico Faggin, for designing the CPU, and Intel's management, particularly founders Gordon Moore and Robert Noyce, for buying the rights and backing the project, together changed the world.
|
http://csua.com/feed/