02 February 2011
C ya' CPU - New computational method uses the wrong end of the computer
Traditional computers are too slow to run the computational methods Tim Hansen seeks to develop - so he decided to build his own. Thanks to a Siemens Foundation grant that he applied for and recieved, Hansen will build a computer where the graphics card handles the calculations.By Jes Andersen
Tim Hansen is a Chemical Sciences graduate student who hopes to develop a method able to predict which molecules are most suitable for use as components in molecular electronics (moletronics). Needless to say, this requires a few calculations.
Too much to keep track of
Molecular components are typically tested whilst situated between two gold electrodes. If one is to calculate the quantum mechanical interactions between electrical fields in the gold and the molecule itself, there are 12,000 electrons to keep track of. Therefore, Hansen is developing a method which includes all the electrons in the molecule, but calculates an average for the gold's electrons. The number of calculations is reduced, but nevertheless quite numerous. In comes the graphics card.
Adding computational muscle
Computation usually occurs within the central processing unit (CPU) of a computer. CPU's might have eight or ten cores, each able to perform one calculation at a time. By comparison, a graphics card, or GPU, is equipped with 448 cores. When calculating in unison they provide a great deal of power, assures Hansen.
"I hope to develop a method able to predict the molecular shape necessary for moving electrons. It would save synthetic chemists an enormous amount of work," says Hansen. There is still much code needed to be written so that the computer can compute using the graphics card as opposed to the CPU, but Hansen is looking forward to the challenge.
"It was a wonderful surprise to be awarded the Siemens grant. I always liked to build my own computers, so it's fun to get 30,000 kroner to do so," summarizes graduate student Tim Hansen.