First came the gHz war, then the number of cores, and now the performance per watt: why the current one is the toughest of all

The race to place the best microprocessors on the market has not always been as it is now. Intel, AMD, and also the brands that at some point have been important in this market, such as IBM or Motorola, among others, have dedicated their resources to the progressive refinement of the microarchitecture of your chips.

However, the leading role in the search for maximum possible performance has been monopolized for decades by the increase in the clock frequency, which is one of the parameters that directly conditions the productivity of a processor.

In just fifty years, microprocessors have developed to the point of being able to operate at a clock frequency clearly higher than 5 GHz.

In just fifty years, those that have passed since Intel launched the first commercial processor in 1971, the 4004, which had only 4 bits and worked at a clock frequency of 740 kHz, microprocessors have developed to be able to operate at a significantly higher clock rate at 5GHz.

In fact, by resorting to cryogenic refrigeration with liquid nitrogen this figure can be increased much more, although it is a practice that we cannot yet consider popular, and with which overclocking enthusiasts flirt above all.

From the gHz war to the number of cores war

Since the early 1970s the improvement of architecture, the increase in clock frequency and the refinement of photolithography have gone hand in hand with the purpose of increase capabilities and the performance of the processors, but there came a time when the silicon technology showed signs of fatigue. I couldn’t go much further.

Working at even higher clock rates could trigger physical degradation of the processor itself due to electromigration, among other limiting phenomena, but just over a decade and a half ago manufacturers came up with a solution: they would integrate multiple cores in the CPU package.

Not all applications could take advantage of multiple cores simultaneously, but those that could take advantage of them could scale your performance without the need to increase the clock frequency of the processor.

Furthermore, when the first multicore chips arrived, technology multithreading it was ready, so the possibilities of multi-threaded processing were very promising. This innovation allows each physical core to implement two logical cores, so that each of them is capable of simultaneously processing two execution threads (threads).

Over the last decade, operating systems, applications, and, of course, processors, have greatly developed their multithreading capabilities.

Intel has Hyper-Threading technology, which is their implementation of this innovation. And AMD has SMT (Simultaneous Multi-Threading), which is yours. In either case, essentially both two-thread processing techniques on a single physical core They work in a very similar way..

Over the last decade, operating systems, applications, and, of course, processors, have greatly developed their multithreading capabilities, so today this strategy has a major impact in the benefits offered by our computers. And also in our experience as users.

semiconductors
semiconductors

However, processors are not the only component of our computers that has developed a lot during the last decades. Main memory, secondary storage solutions and graphics processors have also evolved at a breakneck pace.

And, in the meantime, mobile devices have made their way into our day-to-day life and have caused users to value not only the overall performance of our computers, but also the way they manage energy. We care about speed, but also autonomy, which has placed us where we are now: in the war of brands to offer us best performance per watt.

In the crusade of performance per watt, lithography and architecture rule

The road that has brought us here has been long. Even tedious at times. But above all, it has been full of challenges. Microprocessor manufacturers have been solving numerous challenges to prevent the solutions they propose from becoming stagnant.

They had to put ever-increasing performance in our hands at all costs, and the improvement of the integration technology and the refinement of the microarchitecture have been two fundamental pillars on this tour. And they will continue to be so because they are the two essential ingredients in the quest for the highest performance per watt.

What are light cones and why they are a very valuable tool to better understand the amazing properties of space-time

However, in recent years we have seen that it is not easy for them. Developing the photolithographic techniques that allow semiconductor manufacturers to fit more and more transistors into the same space requires fine-tuning extraordinarily advanced technological solutions. And it is not easy to find them.

In fact, users are witnessing how much it is costing a company with the enormous resources that Intel has. stay ahead in this area to protect its competitiveness.

IBM and TSMC are already working experimentally on the first 2nm chips (or 20 angstroms)

Regarding the development of photolithographic technology the future is hopeful. The roadmap of TSMC, Intel, Samsung, and also that of other semiconductor manufacturers, places us at the dawn of the age of angstroms. IBM and TSMC are already working experimentally on the first 2nm (or 20 angstrom) chips, and Intel says it will have its 18 angstrom (1.8 nanometer) node ready by 2025.

There is no doubt that the next few years are going to be very exciting, although there is a possibility that the chipmakers will not be able to deliver completely. all the promises What are they doing to us now?

In addition, the development of photolithography is only one of the requirements that must be addressed to put chips with higher performance per watt in the hands of users. It is also essential to refine the microarchitecture so that it is capable of carrying out more work in each cycle of the clock signal, and at the same time consume less power.

Improving lithography helps on this path, there is no doubt about that, but it is also essential to devise new approaches that make it possible to implement more efficient microarchitectures. And it is not easy. Microprocessor manufacturers have it harder than ever, but this is one of those paths that, without a doubt, is worth traveling.

Disclaimer: If you need to update/edit/remove this news or article then please contact our support team Learn more