Speaking at the Nvidia GPU technology conference last month, Tom-Michael Thamm and Marc Nienhaus of Mental Images (MI) described how its experience of GPU-based computing for DreamWorks and Sony has been leveraged in a seismic data pilot for an unnamed ‘US oil and gas major.’ MI’s scalable platform for 3D graphics includes the Mental Ray ray-tracing flagship and RealityServer, a web services based GPU compute server.
But the core component for the seismic display environment is MI’s distributed compute environment, ‘DiCE.’ DiCE, whose customers include Autodesk and Dassault Systèmes, hides GPU/multi core complexity from the developer, providing cluster management and networking services, a ‘distributed in-memory’ database, thread management and scheduling.
MI found seismics to be very different from entertainment visualization. While transparency and raytracing are common to both verticals, subsurface data is ‘sparse’ and requires a different approach.
The system is claimed to scale to petabytes of data, distributing and balancing the workload across groups of specialized clusters. Real time interactive volume horizon rendering was tested with 20 horizons with 20 million triangles each. MI’s DiCE is claimed to simplify scaling of high-performance applications, enabling developers to concentrate on the core aspects of their field of expertise.
Mental Images is a wholly-owned Nvidia brand. More from www.oilit.com/links/1010_1.
© Oil IT Journal - all rights reserved.