It goes without saying that the geoscientist's appetite for more detailed information about the Earth can never be completely satisfied. As geoscientists we want to understand the Earth at higher and higher resolution and on a larger and larger scale. Over the past decades, geoscientists have developed methods for investigating the Earth using complicated theoretical models applicable to huge data sets. The many game-changing applications deployed in recent years include: 3D prestack depth migration for imaging the Earth's reflectivity, 3D wavefield tomography for deriving complex velocity models, sophisticated fluid-flow simulators for understanding oil and gas production, and digital rock simulators for understanding the subtle relationships between the measured fields and the underlying geology. All these applications are linked together by one attribute: They are all extremely compute-intensive.