The End of Scaling? Revolutions in Technology and Microarchitecture as we Pass the 90 Nanometer Node

The normative evolution in photolithography, together with Dennard scaling theory has provided continuing performance growth for over two decades. As an industry, this has allowed us to become complacent. For two decades, performance growth was achieved fairly straightforwardly, and the thing being optimized (usually performance) was not a moving target. And the processor itself has remained a high-value item, hence has received a continuing (large) stream of capital investment in continuing its (rather straightforward) growth. We have played by the same rules for over two decades, and have not been pushed too hard to think "outside the box." Our path has been almost purely "evolutionary." Frequently, really new ideas were ignored because the evolutionary paths were already "good enough."

As lithography scales past the 90 nanometer node, it is clear that not only does the lithography scaling slow, but classical (Dennard) device performance scaling slows as well. And whatever performance scaling is possible will be further denigrated by problems with device variability. Further, frequency scaling is coming to an end, as dealing with power and power density impose physical limits and cost limits on systems. In addition, as chips contain larger and larger portions of systems, the value proposition of the core itself is diminished. And finally, virtualization technologies make the specific machinations of a chip, and perhaps the system itself less important to the end user (customer).

Philip Emma joined the IBM Thomas J. Watson Research Center in 1983 where he spent more than a decade working in the areas of computer architecture, microarchitecture, and systems, primarily in the context of what is now called zSeries. Phil holds over 100 patents and has been an IBM Master Inventor for several terms. He has authored and co-authored numerous technical articles, and several book chapters on various topics. He has taught and served as an adjunct at a number of universities. He is a member of the IBM Academy of Technology, and is a Fellow of the Institute of Electrical and Electronics Engineers. Phil holds B.S., M.S., and Ph.D. degrees in electrical engineering from the University of Illinois.