Belgian researchers develop desktop supercomputer.


How much computing power can you cram into a single desktop PC? In our research on image reconstruction we often have to perform large-scale scientific computations, which can easily take weeks on a normal PC. To tackle this problem, we have developed a special PC that is capable of performing our computations just as fast as a cluster consisting of hundreds of PCs. Using this superPC, which consists mainly of gaming hardware and costs less than 4000 euro, we can now perform our three-dimensional reconstructions within a few hours: over 100 times as fast.

The research group ASTRA, part of the Vision Lab of the University of Antwerp, focuses on the development of new computational methods for tomography. Tomography is a technique used in medical scanners to create three-dimensional images of the internal organs of patients, based on a large number of X-ray photos that are acquired over a range of angles. ASTRA develops new reconstruction techniques that lead to better reconstruction quality than classical methods.

Although our reconstruction techniques are very powerful, they have an important drawback: they are quite slow. As the 3D images that we normally deal with can be rather large (typically 1024x1024x1024 volume elements, or more), advanced reconstruction methods can sometimes take weeks of computation time on a normal PC.

Fortunately, these computations can be carried out in parallel, for example using a cluster consisting of hundreds of PC's. Employing a large cluster has some drawbacks as well: it is quite expensive, is not always available, takes a lot of space and requires considerable maintenance. To be able to perform our computations without using a cluster, we develop software for reconstructing 3D images with the aid of 3D graphics cards, that are supposed to be used for playing 3D games. In fact, graphics cards are highly suitable for tomography computations. Each graphics processor (GPU) contains 128 small subprocessors that can all work in parallel. By appropriate programming of the GPUs, many calculations can be performed simultaneously.

Performing our computations on the GPU already results in a speedup of over 40 compared to a single CPU core: great, but still not enough. For our most demanding computations tasks, we developed the FASTRA: a desktop superPC, which contains four dual-GPU graphics cards. Having eight graphics processors work in parallel allows this system to perform as fast as 350 modern CPU cores for our tomography tasks, reducing the reconstruction times from several weeks (on a normal PC) to hours. FASTRA is made completely from consumer hardware and contains four NVIDIA 9800GX2 graphics cards, each containing two GPUs. And guess what .. this system costs less than 4000 euro!

A green desktop supercomputer for a very small price ... just perfect for our applications!