Developers show the way with a new compression method
Computers have been a natural part of our everyday life for a long time. We have also become accustomed to the rapid pace of computer development. Since the 1970s, performance has doubled roughly every two year; a phenomenon which has come to be known as Moore's law. The flip side of rapid performance development is increased energy consumption. Our computer centres are responsible for an ever larger part of our electricity consumption - and the proportion is increasing.
Future computer systems must be green
Development has reached the point where we are now faced with problems if we want to build even more powerful computers. The more capacity we want a computer to have, the greater the amount of power it requires, which means not just an increase in electricity consumption but also a shorter battery life. The more power that is required, the more heat is produced, which means even greater electricity consumption due to cooling of the processors. In short, it is high time we found new methods of building computers.
The cloud – a way to save energy
For the business community, cloud computing offers cheaper and more efficient solutions, avoids the need for extensive in-house IT departments and reduces the need for investments in hardware and software. For private individuals, cloud computing is more about accessibility and security; accessing pictures and document wherever you are and not losing everything when the hard disc crashes. Whichever way you look at it, cloud computing saves energy. But this is not enough.
Complex systems require complex solutions
To seriously improve computer systems and reduce electricity consumption requires a wide range of different skills. Knowledge of algorithm theory is required to make computer calculations more efficient. Program analysis is required to translate the algorithms into a language the computer understands. Computer architecture is required to organise the transistors to do the job. Knowledge of circuitry is required to construct computational structures efficiently and economically. Ultimately, completely new materials with unique properties are required. At Chalmers we are therefore bringing together researchers from many different fields to meet the challenge together.
We are building more efficient computer memories…
The major players in the data world are today building gigantic server halls to store all the data we need. One single server hall at one of the major players today uses enough energy to heat tens of thousands of homes. This is unsustainable. A research group at Chalmers, under the direction of Per Stenström, is therefore developing a new compression technology to build more efficient computer memories. Their concept makes it possible to shrink and at the same time speed up the memory.
...and increasing battery life in smart phones
Some believe that we have to build completely new solutions to achieve energy-efficient server halls and longer battery life for smart mobile phones and the various types of e-reader and iPad. Our researchers are convinced that the same concept can solve both problems. Preliminary results show that their compression method can reduce memory requirements by a factor three. This would significantly reduce energy consumption both in servers and in telephones.
But this is not enough
It is becoming more and more difficult to program computers efficiently. The need for even faster, still more efficient computers will not disappear. There is however a physical limit to Moore’s law. This is something on which everyone agrees. The size of a transistor today is 32 nanometres. When its size approaches that of an atom, nasty things start to happen that affect reliability. There is not just one path to sustainable computer development. To meet the challenge, researchers are working with parallel computers and computational accelerators. New materials will obviously provide new opportunities. This is an exciting time for the developers of the computers of the future.
Per Stenström, professor in computer architecture, Department of computer Science and Engineering