A pump-priming funding bid to explore the potential for GPU-based computing in computational chemistry
Computational chemistry is a fundamental tool that helps us to understand the molecular world. However, even with the power of modern computers many calculations are very time-consuming and there are limits to what systems we are able to study. Supercomputers can be used to perform large calculations on a reasonable timescale, but such machines are rare and very expensive. This project investigated whether graphics processing units, originally developed for computer gaming, can be used to perform large computational chemistry calculations much faster than conventional PCs and more cheaply than using supercomputers in areas such as liquid crystals and catalysis.
This project investigated the feasibility of using GPU-based quantum-chemical calculations for various applications in computational chemistry. The GPU-test node and software funded by the grant enabled some scoping work to determine the potential beneifits and any limitations of this approach. It was found that for molecules containing no metal ions, there is a huge potential for this hardware to speed up the workflow associated with a quantum chemical investigation. This approach was used as part of collaborations between Slattery and Russell (Bristol) and Slattery and Fairlamb (York) that are now in the final stages of preparation for publication. Preliminary results from the project were used to support several grant applications including two large (>£1m) bids for funding from the ERC by Slattery. The project also exposed the current limitations of GPU-based quantum chemistry, as metal-based systems involving 2nd or 3rd row transition metals cannot currently be modelled, due to software limitations. There is significant potential for this area in the future, but significant 3rd party software development is needed and we eagerly await developments here.
|Effective start/end date||1/02/11 → 1/02/12|