Don't take it to the limit: Jim Keller promised Moore's law another twenty years of well-being

Released last week interview with Jim Keller, who leads the development of processor architectures at Intel, helped to remove the fears of some market participants about the imminent demise of Moore's law. Scaling semiconductor transistors will be possible for another twenty years, according to this Intel representative.

Jim Keller admitted that he had already heard many times the prophecies about the imminent termination of the so-called Moore's law - a rule of thumb, which was formulated in the last century by one of the founders of Intel, Gordon Moore (Gordon Moore). In one of the initial formulations, the rule stated that the number of transistors placed on a unit area of ​​a semiconductor crystal can double every year and a half. Now, according to Keller, the scaling factor over a two-year period is about 1,6. This is not such a big regression compared to the original Moore's tract law, but it does not guarantee an increase in performance by itself.

Now Keller is trying not to worry about the approaching physical barrier in the development of semiconductor computing and encourages everyone to do the same. According to him, engineers and scientists will find a way to create transistors whose linear dimensions will not exceed a dozen atoms in each of three dimensions. Modern transistors are measured in thousands of atoms, so their dimensions can still be reduced by a hundred times, at least.

Technically, this will not be so easy to do, for noticeable progress in lithography requires the efforts of specialists in many disciplines, from physics to metal science. And yet, an Intel representative believes that for another ten or twenty years, Moore's law will be relevant, and computing performance will grow at a steady pace. Progress allows you to make computers more and more compact, it changes the way you interact with them and the whole human life. If semiconductor transistor technology ever hit a wall, as Keller believes, software developers would have to drastically redesign the algorithms to achieve an increase in performance with the existing hardware. In the meantime, there is an opportunity to develop in an extensive way, although perfectionists will not like it.

Source: 3dnews.ru