Today our new GPU card arrived – our first go at using a TESLA card. We have received an NVIDIA TESLA C2075 which has a rather useful 448 CUDA cores and 6GB memory. A remarkable feature of these TESLA cards is that they are designed for high-performance computation and do not even have a video output! We anticipate trialling this in a number of projects, including the most recently awarded BBSRC grant for analysis of Biolog data. There may well be significant advantages to be gained in using GPU technology for such high throughput data.This will also be an opportunity for me to learn a completely new computing paradigm, and I might take the time to learn Python as well and make use of the PyCUDA libraries rather than worry too much about C++ (I’m not sure I’ll have the time any more for too much lower-level programming).
Particular thanks go to Matthias Gerstgrasser for starting us off down the GPU route, Theo Kypraios with whom we have recently had funding to trial GPUs in statistical inference methods, Dorota Herman who has been working with us for the last three months on using GPU technology, and John Veasey for helping set up the hardware and software.