Is Moore’s law still applicable today? And is there any limitations to just how small a transistor can actually get?

165 views

[ad_1]

Is Moore’s law still applicable today? And is there any limitations to just how small a transistor can actually get?

In: Physics
[ad_2]

For now, yes. We can still expect the transistors in a chip to roughly double every two years. But there is a limit and we are getting somewhat close to it, so it’s not really clear how long it will continue to apply.

Basically, if you imagine transistors as a switch that open and closes, once you get small enough the electricity will just sorta teleport across it even if you try to turn it off. That keeps it from being very useful. Using other kinds of materials than silicon may help, and even after that we’ll probably still be able to find other ways to make computers more efficient, so it’s not like they’ll be the end of computer advancements. It just means we might have to get more creative and might not make big leaps at quite the same rate.

There’s two sides to the equation.

One is the size of transistors another poster mentioned. This is reaching atomic limits at this point where they’re literally counting the number of mile it in each transistor.

The other is the number of transistors per processor. Software improvements in designing processors continue to increase the number of transistors per processor, but today it’s more about improvements in the *number* of “cores” more than the *speed* of individual “cores” for the reasons stated above.

It is common consensus among the Computer Engineering research community that Moore’s Law, doubling the number of transistors in a chip every 18 months, is dead as of the mid-2010s. There are multiple reasons for this, but one of the biggest is that smallest transistor sizes today, 10s-100s of atoms wide, are extremely difficult to produce without errors (intolerable variations in size and electrical properties). As a result, chip producers have fewer good chips to sell and the chips aren’t cost-effective to produce. They (Intel, Samsung, Global Foundries, TSMC) can make chips with smaller transistors, but advances in chip manufacturing to make it financially worth it have slowed well below the rate predicted by Moore’s Law.

An inaccurate illustration: if your transistors are 25-atoms wide, a manufacturing defect of merely 5 atoms changes the electric property by 20%.

It is unclear whether single-atom or sub-atomic transistors are possible. In my opinion, the areas of focus in computer hardware development are shifting away from reducing transistor sizes (“technology scaling”), and so consumer-grade, mass-produced <1nm-tech chips will not be produced in our lifetime. I hope I’m proved wrong.

Source: Am PhD student in the field.

Moore’s law always was only an observation. People started to call it a “law”, like it followed some kind of rules, but that is not it.

A long time ago transistor count increased because new functions were added. Nowadays, the transistor count increase is mostly cache.