Recently, I saw a post on the internets that stated companies like Intel should focus on creating new technologies and not try to optimize the already existing ones. Specifically, they should “try to find something amazing that could replace SOI”.
This point of view is obviously simplistic, but it made me think about the wrong perceptions we have when newbies to the technologies out there.
First of all, we need to be able to talk in terms, and the ITRS does a pretty good job helping us on that. So, there’s three categories at the moment: More Moore, More than Moore and Beyond Moore. More Moore means scaling of the current CMOS technology by incorporating different materials or doing process adjustments like high-k dielectrics and SOI, or even different FET structures (FinFETs) etc. More than Moore means the incorporation of a new technology into the standard CMOS, like MEMs. And finally, beyond Moore involves a leap further from conventional CMOS and not just integration into it. Examples in this case are only limited by your imagination and can be carbon electronics, spintronics, memristors, you name it.
Therefore, SOI is not really a different technology at all. Despite it being effective in providing memory devices for the radiation hardening community, the real goal of SOI was to solve some issues that involved the scaling of the CMOS process itself and not invent something completely new. And this is the whole point of Beyond CMOS: To overcome the limits of the latter. Whatever new and exciting you find in the More Moore spectrum, the imminent reach of the physical limits means that you will soon have to abandon it.
But on to the real issue here: What research should the industry do? Well, I don’t know who gave them these crazy ideas that companies operate for the good of humanity, but the real answer is “whatever will bring them the highest income with the least expenses”. So, apart from the obvious minimization of the risks and a reluctancy to undertake a project that might not be successful, it also means that they operate on a “milk the cow” basis. And you do that by staying close to the technology that has the better chances of involving and being adopted in the future.
Intel, for example, has been reducing the size of its transistors using process adjustments every two years according to :
In this white paper, it also mentions III-V semiconductors as probably being a project that they undertook. But they also go a little beyond that by mentioning spintronics and quantum computing, as something whose feasibility they are estimating, or at least that’s what I make out of it. Visiting their research website you can really tell where this company is oriented to: standard consumer electronics, servers etc. Fast speed, low power consumption. However, IBM’s website has a different tone. It is full of different interest areas. Spintronics, carbon nanotubes, optical chipsets etc.
Beyond Moore scaling of electronics involves intense research that spans over different science areas (materials, chemistry etc.). To reach the point where a company is interested in a specific new idea, my view is that not only it needs to be something that has a high probability of actually working (and working better than CMOS in the aspects of interest like speed and power consumption), but it also needs to be scalable so that the cow can be milked accordingly.