Qualcomm said today it’s joined Sematech, an organization dedicated to advancing research into semiconductor manufacturing — the first chip company that doesn’t have a manufacturing effort to do so. While it may seem odd that a so-called fabless chip company would help fund R&D for better manufacturing, it’s an indication that the struggle to cram more transistors on smaller chips is becoming a bigger problem for all chip makers.
It’s a struggle that hardware vendors and large data center operators are worried about as well, and something about which I’ve spoken with people inside Microsoft and Yahoo, both of which have massive server farms and business that rely on cheap computing. I wrote about it my first day on the job here at GigaOM; moving down the process node, as it’s called, has only become more of a problem in the two and a half years since.
But Moore’s Law, the idea that the number of transistors on a chip doubles every 18-24 months, drives down the cost of electronics and computing while also keeping performance gains coming. In a recent GigaOM Pro article (sub req’d), I discuss the medium- and long-term solutions for ensuring that semiconductor performance keeps rising even as it becomes difficult to keep Moore’s Law on track with conventional chip manufacturing. They include building massively multicore chips either though efforts such as Intel’s cloud chip or by using graphics processors, as well as 3-D stacking technologies for logic circuits. Other performance boosts fall outside of the logic and memory circuits; they rely on using optical interconnects to communicate on the chip and between chips as well as putting more memory closer to the CPU so the information needed by the CPU doesn’t have far to travel and there’s more data available nearby.
Longer term (think decades), HP is researching memristors, an alternative to the transistor, while IBM and various universities are pushing nanomagnetic materials for an entirely new and more energy-efficient semiconductor. There’s also the ever-present hope of quantum computing panning out. For the full story on pushing processors beyond Moore’s Law, read my full analysis.