MEGWARE HPC CLUSTERS

Technology Challenges of Simulation and Visualization Solutions

The more detailed and larger the volume of data, the greater potential of higher-resolution imagery and the more informative the simulation and visualization can be. With the growing data sets used in various disciplines, excellent image fidelity can be drawn from the massive data volumes. But, the limitations of today’s graphics processors make visualizing such large datasets a challenge. Additionally, users want their simulations and visualization to be interactive. They can then study what-if scenarios by modifying parameters during computation and see the visual result in real time to gain greater insights. Such interactivity can significantly slow down a visualization system, especially with very large datasets. Until recently, dedicated interactive solutions have failed to strike an adequate balance among performance, visual fidelity, and cost. Rasterization techniques on graphics-specific hardware can accelerate rendering of smaller 3D models but do so at the expense of higher fidelity. Ray tracing techniques offer the highest fidelity for simulations and visualizations, but the performance of ray tracing algorithms on graphics hardware has been an impediment. The inherent raster-focused architecture of graphics processing units (GPUs), along with their limited memory and input/output (I/O), fail to deliver adequate performance for ray tracing algorithms.