High Performance Interpretation Computing

TerraSpark's concept of Interpretation Intelligence™ makes use of every computational resource available in the user's workstation or laptop. The powerful computations displace tedious mouse-pushing activities and support advanced technologies to condition and analyze seismic data, in order to automate the interpretation process under control of the geoscientist.

Because interpretation activity takes place away from computing centers, it was not realistic to pursue CPU cluster computing capabilities for activities driven by interpreters. TerraSpark software makes optimal use of multi-core CPU architectures, but the more powerful among these systems are quite expensive, generate a lot of heat and fan noise, and are most probably not going to be aligned to our customers' standard workstation configurations.

Multi-core CPU optimization of existing code, in effect a re-write of exisiting parallelized software to leverage the latest chip designs such as the Intel I5 and I7 architectures, delivers tremendous boosts in performance. This makes it possible to run intensive calculations on a mainstream PC (e.g. 8 core).

GPGPU (General Purpose Graphic Processing Unit) architectures are plug-compatible with recent Windows and Linux workstations, and for a low cost and minimal implementation efforts add massive computing resources to any workstation. Working with market-leader NVIDIA®, TerraSpark is implementing most of its compute-intensive processes for operation on one or more GPGPU cards, should they be detected when Insight Earth is run. Compared to an 8-core CPU system, processes ported to date show 10x acceleration for a single graphic card, with an almost scalar improvement if additional GPGPU cards are added.

 

This performance boost will redefine the perception of Interpretation Intelligence-based systems. While the current CPU-based execution times are offering high-accuracy results in a fraction of the time needed to interpret manually, users are concerned about long compute times. With GPGPU execution taking only a few hours, the whole workflow becomes even more attractive, for example by enabling experimentation around parameters and settings.

 

 

Commercial availability of the GPGPU-enabled processes will be announced soon.