It was 1965 when Intel's future co-founder, Gordon Moore, prophesied that the number of integrated circuit components would double every year until it reached 65,000 in 1975.
In 1968 Gordon Moore founded Intel and in 1975 his prophecy came true with surprising precision. It was then that the prediction became one law, which predicts that the number of transistors on a chip would double every 2 years.
An economic law
Moore noticed that, thanks to the invention of the integrated circuits on small squares of Silicon, a few years earlier, the price of each piece of circuit would have been inversely proportional to the number of pieces present in the circuit.
This can be trivially translated by saying that the more transistors would have been added, the less the price of the single transistor would have been.
Miniaturization engineering was still only in its infancy, so Moore noticed that there was ample room for progress and from there, in fact, his law has accompanied technological progress for almost half century, being able to predict almost perfectly the trend of the computing power of the new processors.
The construction of ever more powerful and ever smaller chips quickly became the main objective of technological research, so fundamental that the progress of the world still depends on it today. It is estimated that innovations in the IT sector depend on one third of the growth of the entire American economy from 1974 to today.
In fact, every new technological gadget is nothing more than a goal of this type of research: from megacomputers with a few thousand transistors, to smartphones with millions and millions of those "megacomputers" inside, everything has been allowed by miniaturization.
A self-fulfilling prophecy
1965 was still the beginning of that technological boom that led humanity to experience the most flourishing period ever. It is surprising to see how Moore managed to extrapolate from a graph with a few points such a hard-to-die law.
The choice to follow this law was purely economic and all in the hands of the large silicon companies, which, with almost Swiss punctuality, placed their new top models on the market, which outclassed exponentially the computing power of the above following the forecasts.
More than half a century ago, when personal computers were only a dream, Moore had foreseen a future of self-driving cars, internet-connected terminals and portable communication equipment.
Research has thrown itself headlong into the creation of that land promised by Moore, so much so that, every year since 2001, almost all the discoveries that have marked a milestone in technological progress have been possible thanks to this type of progress.
This was true until we found ourselves faced with the inevitable impossibility of miniaturizing indefinitely.
The end of an era
It took Intel 5 years to develop the technology a 10 nanometers which followed on from the 14 nanometer one. We are no longer able to follow Moore's rate of progress said Charles Leiserson, MIT computer scientist and pioneer of parallel calculation.
For many years the death of Moore's law has been declared and the data certainly does not give comfort to those who claim to follow the old guides, while the CEO of Nvidia declares the era of progress following Moore's law definitively extinct.
A slow death
Already in 1999 we were sailing on sight and the future appeared dark: Moore himself saw the end of his prophecy near and some Intel researchers were concerned about side effects that a miniaturization under 100 nanometers would have brought. Many feared that they would not be able to develop this innovation by 2005 because the side effects were insurmountable and there was no known solution in those years.
Scientists, it is known, are quite stubborn and do not accept that something is not possible if there is no physical law that prevents it.
Many technologies were born that allowed to optimize the use of electrons and create increasingly sophisticated and precise construction processes that, in fact, led us an order of magnitude below what was believed to be the minimum limit just 20 years ago.
This did not happen without sacrifices: the research requires ever greater funds so much so that it is estimated that since 1971 the funds invested in it have increased by a factor of 18.
The cost of building factories that can keep up with new technological advances is becoming prohibitive. It is estimated that by 2022 more than 16 billion will be needed to maintain cutting-edge production.
These costs are not easily tackled, in fact the market has drastically shrunk: from 25 competitors in 2002 it has fallen to 10 in 2010 and only 3 for the current generation. One of these companies is Intel, which does not give up on accepting the imminent end of the law that takes its name from its co-founder.
It's not all over
Jim keller, chief engineer of the Silicon division of Intel since 2018, believes that it will still be possible to keep Moore's law alive.
Keller even pointed out that there are probably hundreds of solutions to double the amount of chips on the transistors, thanks to the solutions provided by new technologies, such as the3D architecture is new designs transistors.
Intel, according to Keller, has planned to continue its silicon technology for at least the next 10 years, bringing the transistors on a chip to the impressive number of two thousand billion, with a calculation ability 30 times higher to the current one.
However the point is that the time when you had access to smaller, faster and faster chips economic every two years it's really over.
Where the future leads
The new frontiers of technology currently under study at MIT, foresee the increase of the computing capacity of the next CPUs by using increasingly optimized architectures and software.
The very high calculation capacity of current machines allows programmers not to focus too much on program optimization, preferring computer languages at a higher level, such as Python, but with a higher use of resources, instead of basic languages, such as C , which would require much more work but far fewer resources spent by the PC.
The exploitation of multi-cores in the current generation of programs is very poor, so much so that it has been seen that the mere rewriting of some computationally difficult calculations from Python to C speeded up the process by a factor of 47.
In terms of time, running on an 18-processor multicore is a problem that in Python requires 7 hours of calculation, the same problem is solved in oneself 0.47 seconds if written in C.
This, of course, does not save Moore's law as it is, but allows for significant improvements in terms of performance from existing hardware.
Other companies, such as Microsoft or Google, are focusing their energies on developing Artificial intelligence is Deep Learning knots using highly specialized chips, such as those produced for GPU, who use a very large number of parallel processors (much larger than what is used by CPUs) to increase the computing capacity of devices, but their architecture flaw in versatility.
The future is not yet close. As Moore's law exhales its last breaths, it is doing the impossible to keep it alive because it is not known yet which technology will really replace the current one: years of basic research are still needed to create a substrate on which to start again for a new technological boom, for this reason we try to encourage new logical devices and new ways of doing computation.
Quantum computation, carbon nanotube transistors or technology that will exploit the spin of the particles (spintronics) or the atom (atomtronics) or the light (photonics), are all equally possible paths, which research is trying to follow, but none of these guarantees us a new development that follows Moore's law. We just have to wait, the future is around the corner.
