Spectraseis Wins High Performance Computing (HPC) Award

Recognition for fast wave propagation algorithms for microseismic fracture imaging enables Spectraseis to move from 2D to 3D at 100x previous speeds.

Spectraseis’ Software Programmer Igor Podladtchikov was recognized by IDC (International Data Corporation) with doubling the speed of both acoustic and elastic solver throughput, at the same time improving code size and maintainability, while harnessing the massive parallel computing capabilities of Graphical Processing Units (GPUs). On an average usage condition, using a pure CPU approach, it would take 3 to 5 hours for average solvers to execute a program; however, with improved efficacies obtained by code redesign for Fermi and Kepler GPUs, the time necessary to solve the same program was reduced to seconds.

Under normal conditions, Spectraseis’s Time Reverse Imaging (TRI), driven by a wave equation solver, would take approximately 4 hours running on 7 CPUs; using our new approach and running on 7 GPUs, this process is now reduced to 90 seconds! In addition, for this 3000 time-steps process, the TRI solution also requires an additional noise model, so the solver is solving for two domains to obtain the corrected image. Utilizing Kepler K10 GPU cards, the solvers are 1.8x faster.

The improved capability allowed Spectraseis to move from 2D to 3D and in several cases we obtained more than 100x speed-up.

We would like to congratulate both Igor and his collaborators for this accomplishment.