Home / Seminar & Event /Past Seminars / (324) HPC Trends in Computational Chemistry
Home / Seminar & Event /Past Seminars / (324) HPC Trends in Computational Chemistry
(324) HPC Trends in Computational Chemistry
Seminar: (324) HPC Trends in Computational Chemistry
Speaker: Dr Rika Kobayashi, Australian National University
Time: 2018-09-26 16:30 to 2018-09-26 17:30
Venue:
Organizer:

Health Science Platform


Venue: No. 3 Meeting Room, Conference Building


Abstract

    The pace of technological change is such that supercomputers rapidly become more powerful, advancing capability in many directions, allowing researchers to tackle larger or more complicated problems that could not hitherto be done. However, the traditional Moore’s law view of computers doubling in power every two years, based on the number of transistors in an integrated circuit, no longer applies. Limitations on computational chemistry now not only include technology but commensurate software development, environmental factors and even economics, requiring a shift in paradigm to traditional approaches to what calculations can be done and how they can be done. The early days of computational chemistry just required a computing unit, algorithm and program; the advent of massively parallel computers required a rethink of  algorithms and software. Nowadays, modern architectures, especially the emergence of accelerator technology such as GPUs, have affected software development and given rise to new ways of looking at chemical problems through machine learning. The increased computational capability has also had consequences affecting the handling of all the data that is being generated. Finally, economic considerations are seeing the demise of many traditional compute clusters in favour of the “cloud”. I will briefly go through these developments especially in the context of how they impact computational chemistry.

 


(324) HPC Trends in Computational Chemistry
Seminar: (324) HPC Trends in Computational Chemistry
Speaker: Dr Rika Kobayashi, Australian National University
Time: 2018-09-26 16:30 to 2018-09-26 17:30
Venue:
Organizer:

Health Science Platform


Venue: No. 3 Meeting Room, Conference Building


Abstract

    The pace of technological change is such that supercomputers rapidly become more powerful, advancing capability in many directions, allowing researchers to tackle larger or more complicated problems that could not hitherto be done. However, the traditional Moore’s law view of computers doubling in power every two years, based on the number of transistors in an integrated circuit, no longer applies. Limitations on computational chemistry now not only include technology but commensurate software development, environmental factors and even economics, requiring a shift in paradigm to traditional approaches to what calculations can be done and how they can be done. The early days of computational chemistry just required a computing unit, algorithm and program; the advent of massively parallel computers required a rethink of  algorithms and software. Nowadays, modern architectures, especially the emergence of accelerator technology such as GPUs, have affected software development and given rise to new ways of looking at chemical problems through machine learning. The increased computational capability has also had consequences affecting the handling of all the data that is being generated. Finally, economic considerations are seeing the demise of many traditional compute clusters in favour of the “cloud”. I will briefly go through these developments especially in the context of how they impact computational chemistry.