On 19 April 1965, Gordon Moore, an electrical engineer with a P.Hd in Applied Physics and Mathematics working for Fairchild Semiconductor, wrote in a May 2015 issue of Electronics magazine that the number of transistors on a chip would double roughly every year, with the price of the processors remaining the same as time passed and scales of economy allowed electronics to become cheaper and cheaper. That wasn’t the writing that ended up being Moore’s law, but it was later amended by Moore to reflect his theory that the number of transistors would double on a chip roughly every 18 months to a year. Moore’s law has since held true, and even predates the day of its original creation by some margin.
To get a sense of the scale of just how things have changed since then, in 1965, you have to pick up a pencil with a rubber on the end. Do you see how small that is? That little eraser is the size of one transistor back in 1965. Just one of them. At the time of writing, various companies were able to cram in 60 of those things on to a single die for manufacturing. According to Moore’s calculations, that would mean that by 1975, his hypothetical processor design would include 1920 transistors, a 32-fold increase over ten years. He only ever imagined it being valid for ten years before production and better technology began to slip forward, but it didn’t.
By the time he founded Intel with Robert Noyce in 1968, the computing world was already following his theory closely and it did indeed pan out more or less as he’d expected. By 1971, Intel’s first processor, the 4004, has 2300 transistors. In 1972 the 8008 had 3500 processor and by 1974, the Intel 8080 had 4500 transistors, tallying close to Moore’s law. This continued on for the next 50-odd years and in an interview with Intel, Moore said that it wasn’t really something they thought about initially.
“In the beginning, it was just a way of chronicling the progress,” Moore, now 86 years old. “But gradually, it became something that the various industry participants recognized as something they had to stay on or fall behind technologically.”
Moore’s law hasn’t slowed down in recent years, but the process technology has. Moore’s idea initially for the time frame of around 18 to 24 months was due to the amount of time it took to retool a fabrication plant to allow for a new production process to work. More and more, though, companies like Intel are putting their extra transistor counts into better graphics and dedicated hardware to drop power consumption, because moving to smaller and smaller process nodes has become more and more expensive. At 7 nanometers between each transistor, Intel plans to move to something that isn’t silicon in order to keep the chip from overheating and melting.
Will Moore’s law transcend silicon and move into different processor types? Possibly. For now, though, he can gloat about being right for five decades on end.