Fast, faster, fastest
The computing power of computers grew exponentially in the past few decades thanks to progressively faster processors. This development is now running into physical limits. While miniaturization continues apace, processor speed is becoming a problem.
Put simply: the faster a processor operates, the hotter it gets. In some equipment, a chip measuring a square centimeter today already gives off more heat than a hotplate.
Our computer scientists are embarking on new paths to develop hybrid high performance computing systems. A simple example is a combination of two processors on one chip. One of the processors has a great deal of storage capacity and the other is especially fast. Our hybrid systems are slated to have tens of thousands of different processors all working on the same task simultaneously.
Co-processors are also being developed to carry out even more specialized tasks, for example, in the graphics field. If necessary, these are to be integral to the system. With so many different processors, programmability is a major hurdle, and solving that problem is also one of our major goals. Other research priorities focus on developing compatible and reusable software components as well as reconfigurable hardware components.
New computer architectures and infrastructures must be designed that marry up all components for optimal functioning. This new dimension in supercomputing will take time as some hardware components still must be refined and adapted to support future high performance requirements. For example, processors need to become less sensitive to heat and more able to better withstand vibrations, humidity and electromagnetic waves.
We are also working on improving the energy consumption of processors and increasing their tolerance to hard- and software caused faults. These robust yet flexible new systems will service the operations of all our Research Areas and will be of major importance in achieving our goals.