How small are the transistors on a chip?
Время обновления: 2021-12-06 10:56:07
In the most advanced chips, transistors are as small as a virus, that is, about 50-100 nanometers (a nanometer is one millionth of a millimeter). We will see in this article how the size of transistors has evolved from the invention of the integrated circuit (IC) in 1959 to today.
Figure: Color image of a chip taken under a microscope. What you see are the metal interconnect tracks of the transistor, they are underneath and not visible
Moore's Law, a self-fulfilling prophecy
In 1965, Gordon Moore, the founder of Intel, published the famous article that, over the years, has become a prophecy: the number of transistors being integrated into a chip doubles every two years. This article is a true gem in the history of electronics and probably one of the most influential in the subsequent development of this branch of science and technology. The following figure shows Moore in his famous article.
This growth in mathematics has a name: the exponential law. Since then, the increase in the number of transistors in ICs or chips (they are the two usual ways to call them) has followed this trend, which is known in the electronics world as Moore's Law.
This prediction has served as a kind of "self-fulfilling prophecy" in the microelectronics industry, which manufacturers insist on fulfilling year after year. Of course, it's not just about fulfilling predictions, but about seeking (and finding) the benefits of a trend that has almost a "divine mandate" for the industry. When you read Moore's article, it remains striking that the predictions he made in it not only came true, but that they have done so year after year for half a century. It is not easy to find a prediction "chase" as determined as this one in any other industrial sector.
Thus, since their introduction in the late 1950s, ICs experienced unprecedented growth in the following years, initially driven by the U.S. space program and the military industry. More and more transistors were manufactured. If the first ICs integrated a few dozen transistors, a few years later there were thousands of ICs on the market, and today, there are billions of transistorized ICs. the following figure shows this trend.
Figure: Moore's Law for CPUs and memory. The vertical scale is logarithmic
The number of transistors in an IC can be increased dramatically because of the tremendous development that its manufacturing process has undergone. Microelectronics employs a process that is partly reminiscent of the mass production of automobiles, so that a large number of identical complete circuits can be reproduced simultaneously on a single silicon semiconductor wafer. The process consists mainly of printing multiple geometric patterns on the silicon surface, which allows defining each of the devices that make it up, and then selectively depositing various insulating and conducting materials in order to properly interconnect the different components. All this can be made possible with the help of the real bottleneck technology in IC manufacturing: lithography.
What does Moore's Law really mean?
To get a better sense of what Moore's Law means, imagine the reader going back in time. Go to 1971 and prepare to listen to an opera in an auditorium with a capacity of 2,300 people, with exactly the same number of transistors as the first microprocessor Intel built that year, the 4004. If the capacity of that auditorium had evolved in accordance with Moore's Law, but without changing the space it occupied and restored in 1982, then 134,000 people would have gathered in the same venue - the capacity of a large soccer field and the capacity of Intel 286 processor -. Years later, in 2000, the auditorium had the capacity to hold the entire population of Tokyo - 37 million people, the same number of transistors as in the latest version of the Intel Pentium III processor -; if the audition had been held in 2011 it would have encountered 1.3 billion people - the entire population of China or the number of transistors in one of the Intel Core i7 processor versions - and if it had taken part in 2019, the audience would have held 7.4 billion people, that is, the population of the entire planet - and the number of transistors built into the IBM z13 storage controller. Finally, if this audition were to take place in 2021, the capacity would become the equivalent of two Earths, since it would have to gather 15 billion people, that is, the number of transistors of the processor Bionic A 15 (the model equipped with the new iPhone 13). As shown in the figure below.
Figure: The amazing Moore's Law
Strictly speaking, the chip area has also increased slightly over the years, but in much smaller proportions. For example, the Intel 4004 takes up 12mm² (3mm x 4mm), while the Bionic A 15 has a surface area of 107.7mm² (8.58mm x 12.55mm), so that the ratio of transistors is multiplied by 6870 .000, while the area is 9.
Need to make transistors smaller and smaller
Integrating more and more transistors on a chip means reducing its size, which brings advantages, but at the cost of increasing the complexity of the manufacturing process. The smaller the transistors, the smaller the chip itself and the more chips can be mounted on the wafer. At the same time, the cost of processing a wafer remains approximately the same regardless of how many chips can be obtained from each wafer. This means that reducing the size of the transistors will result in cheaper chips. Alternatively, the chip can remain the same size so that it has more components inside. This makes it more powerful, but not more expensive. Most importantly, reducing the size of transistors can improve their performance without increasing their power consumption. That said, chipmakers have a strong incentive to reduce the size of transistors. That's exactly what they've been doing for the past few decades, in which the number of transistors on a chip has increased from hundreds to billions. That's an amazingly small size for a transistor. The following figure shows one such transistor.
Figure: Current transistors are as small as a protein or a virus
To understand what this small size means, imagine that we scale up the size of a chip with integrated transistors until it is similar to the area occupied by the community of Madrid (8,000 square kilometers). At this scale, each transistor would occupy or area. 5 square meters, similar to the screen area of a 40-inch diameter TV.
Mark Lundstrom, a Purdue University professor who began working in the chip industry in the 1970s, wrote an article for Science magazine in 2003 predicting that Moore's Law would reach its physical limit in 2015. Lundstrom says that many times in his career he has thought, "Well, it's over." He remembers attending his first chip manufacturing conference in 1975. In his words, "There was a guy named Gordon Moore giving a talk. He was well known in the technical community, but no one knew who he was, and I remember his talk. Moore said, "Soon we'll be able to put 10,000 transistors on a chip. What can someone do with 10,000 transistors on a chip?"
Today, there are 15 billion transistors on the market. What can your designers and manufacturers do with them?