Will the speed of silicon-based computers continue to increase? Showed the increase of computer speed over the past 30 years on a graph. He indicated that there was an increase in computer speeds that followed Moore's Law, namely, the speeds doubled every two years. He indicated that the speeds will level off around 2010. Yes, we expect to get higher speeds through miniaturization but the limit will be hit at the atomic level due to limitations of X-Ray lithography. What are the opportunities post 2010? We can place many processors in one box. We can move to non silicon computer chips, for example, quantum, chemical or bio inspired (all have parallel architectures). Unless new technology appears we will need to rethink the architecture of computers. There are a variety of architectures at both at microscale and macroscale. The core notion is that the computer or computers that we are using must appear as if we are using ONE machine. What is a consequence of architecture diversity? We may wish to specify an architecture that suits a particular class of applications. What is the ramification of dealing with a large number of processors? Do we know how to use them? No. We need to investigate different types of parallelism. We need new paradigms to exploit more than 100 processors that work together. What are the ramifications to MDO? The old driver was to reduce number of visits to full analysis and to distribute design work to many people. The new drivers are to make many analysis simultaneously (with analysis being almost instantaneous) and to distribute design work not only to many people but distribute it all over the world. What is the bottom line? We need to compute as fast as engineers can think
展开▼
