The grand challenge for computational biology is to provide fundamental insight into how molecules function inside cells. Computational models provide the capability for making systematic, controlled perturbations to molecular structure at the atomic scale. However, biological systems often function over length and time scales that are intractable using current technology, even with the exponential """"""""Moore's Law"""""""" rise in speeds of Central Processing Units (CPUs). Recent developments by our team have demonstrated that the power of Graphical Processor Units (GPUs) can be harnessed to drastically improve our abilities to study complex systems over physiologically relevant time scales. With greater than hundred-fold improvements over CPUs readily accessible, we propose to establish a shared, large-scale GPU-based cluster for use by collection of interdisciplinary faculty from the departments of Chemistry, Bioengineering, Mechanical Engineering, Bioinformatics, and Computer Science. Complementing our team's scientific track record, our team has also pioneered the development of GPU-based software, including molecular dynamics calculations. This software has made possible more accurate molecular-level simulations of protein folding, chaperonin function, and membrane fusion. Moreover, the recent release of a GPU-engine for Matlab will make the cluster useful for a broad range of biological problems that benefit from large-matrix linear algebra computations, from RNA structure, to motor mechanochemistry and design, to cellular mechanics. In addition, as new classes of biological problems become feasible, our team is poised to utilize these studies to develop new physical models of biological phenomena and to develop algorithms that extend the capabilities of the cluster. Rather than the usual incremental increases in resources, this GPU-based cluster will represent a landmark improvement that revolutionizes the set of biological problems accessible to computational biology.