Explore hardware selection and trends in hydraulic modelling in this 59-minute webinar from the Australian Water School. Learn from Greg Collecutt, Team TUFLOW's senior GPU code developer with over 25 years of experience in scientific modelling software. Discover how new computer hardware generations favor certain hydraulic solution schemes and offer lower solution costs. Understand the implications of processor speed limitations and falling silicon process scales on solution schemes, emphasizing the need for highly parallel computation and explicit solvers. Examine how increased modelling resolutions reveal deficiencies in existing methods, particularly in turbulence closure formulation. Gain insights into adopting software that aligns with hardware advancements and new methodologies for smaller modelling cell sizes and Monte-Carlo style statistical studies. Recognize the importance of staying grounded in hydraulics and interpreting results accurately as model resolutions increase. Compare CPU and GPU performance, exploring how current GPU technology can significantly reduce solution costs. Dive into topics such as parallel code writing, CPU and GPU parallelisation, real code examples, and GPU capabilities through various syllabus points, concluding with a Q&A session.
Overview
Syllabus
Introduction
Clock speeds plateau
CPU and GPU performance
Writing parallel code
Example
CPU performance
CPU parallelisation
GPU parallelisation
Results
Results summary
Real codes
GPU capabilities
Example model
The role of modelers
Cost comparison
Hardware summary
GPU selection
Summary page
QA
Conclusion
Taught by
Australian Water School
