I've just returned from Mentor Graphics' EDA Tech Forum 2010, titled: Delivering the latest in 10X design improvements.
The opening keynote by Pravin Madhani, GM, Place and Route Division, Mentor, could have been better -- well, Dr. Walden C. Rhines, chairman and CEO, Mentor, had also delivered a similar lecture at this year's VLSID 2010 conference.
However, the other two keynotes -- by Dr. Kota Murali, lead scientist & program manager of nanotech, IBM India, and Manjunath Hebbar, VP & Head - Strategic Services, HCL Technologies Ltd, lived up to their billing.
The photomask industry is between the proverbial hard rock and the hard place. For instance, at 32nm, the mask cost works out to be $2 million today. In his keynote, Madhani said that the manufacturing industry would surely figure out a way to control mask costs.
Even fab costs are pretty high today -- estimated at $50 billion in 2010, that is ~10 percent of the annual market. The global fab industry continues to figure out how to decrease costs. While design costs are projected to grow logarithmically, cost per function will continue to decline long after Moore's Law is obsolete.
So, will we have any use for so many transistors? Down the years, growth in unit volumes has always distinguished the semiconductor industry. The semicon industry has been growing at 13 percent (10-year CAGR), while transistors have grown at 49 percent. These sit very well, as compared to say, computers - 9.3 percent, steel -- 5.3 percent, and automobiles -- 0.1 percent. The 49 percent transistor growth drives the semicon industry.
Madhani said that the note/netbook market seems to have several years of growth ahead. The Apple iPad has also created a new segment. Cell phone adoption has been in high-growth mode in the emerging markets. Smartphones are changing the video dynamics.
So, will applications require 10K more transistors by 2018? And, do we have the necessary design tools? Well, there will likely be a ~10K increase in transistors over the next eight years, going up to 40 billion transistors by 2018. Therefore, the industry will require tools ready now in order to design for 2018.
Four principal areas will require 10X improvements in design methodologies — system level design, verification, embedded software development, and back-end physical design and test. A 10X increase in the number of transistors will also require 1000X increase in verification.
In summary, reduction in costs per functionality will continue on a predictable learning curve long after Moore's law is obsolete. The industry will also witness ~10X increase in transistors over the next eight years, leading up to 40 billion transistors by 2018.