A team of researchers used a cluster of Tesla GPUs to make a breakthrough discovery, finally explaining how gas giants are formed.
United by a common expertise in planet-disk interactions, four researchers from three countries recently set out to close this gap between theory and observation. They took on the challenge of modelling the physics between protoplanets and the protoplanetary disk.
Running these simulations is no simple feat. They demand immense computing power, capable of modeling the hydrodynamics of the gaseous nebula surrounding an ever growing and moving celestial body, and also of computing the transfer of radiation within this nebula. Running all their models would have taken up a million CPU hours. That’s why the research team chose a cluster of Tesla K20 GPUs instead.
Using CUDA helped the researchers develop a publicly available code, called FARGO3D, which accounts for all the hydrodynamics and radiation transfer processes. Powered by the NVIDIA Kepler GPU architecture, the Tesla K20 cluster could run this code on each of the authors’ models within hours — gaining a 40x speedup on one GPU versus one CPU core.