The End of Moore's Law and its Impact on Design Complexity
JianiZhang 270006W0BW Visits (895)
It is important to first state that Moore’s Law is in fact not a natural law but an observation. Moore’s Law characterizes the growth of transistors in integrated circuits as doubling every 2 years and has long been used as a guide for product development in the semiconductor industry. From a consumer’s perspective, this phenomenon has manifested itself into faster processing speeds in CPUs, larger capacities in memory and more pixels in pictures. However exponential growth cannot be sustained indefinitely. Modern day studies state that by the end of 2013, transistor counts in semiconductors will slow from doubling every 2 years to doubling every three years.
So how does the end of Moore’s Law affect the semiconductor industry? In short, it means that product innovation will be much harder. Gone are the days where new products can be easily introduced just by shrinking the size of the transistor. I remember when I started as an IC design engineer, my first project was operating at a transistor size of 120nm. By the time I had left, the designs were operating at 45nm and yet many of the peripheral circuits had stayed more or less the same. The ability to reuse peripheral designs meant less time spent on development and a faster time to market. However, now in the sub-22nm designs, things are not that simple. Atomic level limitations and material constraints mandate substantial changes to the peripheral circuits, which must compensate for the instability of these miniature circuits.
To further complicate the situation, demanding consumers and fierce competition are forcing semiconductor companies to look beyond their typical competitive advantage. Customers are no longer content with a fast processing CPU. They demand CPUs with GPUs and integrated memory. Modern day semiconductors must be the jack-of-all-trades and the master of all. An example of these multitasking semiconductors is seen in Intel’s paper on its new Haswell chip at this year’s ISSCC (International Solid-State Circuits Conference). This chip not only has its usual faster processor speed but also includes improved graphic performance and battery life due to fully integrated GPU and voltage regulator.
Haswell architecture: Intel showed off improved processing power, a sleeker form factor, and an integrated voltage regulator.
So, along with having to redesign much of their original chip, companies must wrestle with the headache of integrating other types of chips onto their own. And of course, tackling all these complexities means more time and money spent on development.
In order to address these issues, there are two thoughts that come to mind. The first deals largely with organized and well maintained hardware design. More components on a chip means more interfaces, more interconnectivity and more ways to go wrong. Therefore, it becomes even more important to have an integrated development process with clear requirements management and design traceability. The second way to address design complexity is to use more software to compensate for hardware deficiencies. Look for a blog post soon that discusses increased software usage to address hardware complexity.
Get an overview of Rational tools for
Read more about the Intel Chip and the