Following SMT’s use of Graphics Processing Units (GPU) to perform fluid flow modeling (OITJ Vol. 10 N° 8), Boston-based FineTooth was showing seismic processing and data compression on a bespoke NVidia-based system from Panta Systems of Cupertino, CA at the SEG. The hardware included 16 high-end NVidia Quadro 4500 graphics cards with Infiniband interconnect and storage for a 110GB prestack dataset. Finetooth’s software is being developed to decompress on the fly for visualization of such massive prestack datasets. Later work will extend the use GPU to seismic processing. The potential is significant. Today’s Intel Xeon processors have a theoretical compute capacity of 6 GFLOPS, the Quadro, 165 GFLOPS.
Buch
FineTooth CEO Diderich Buch told Oil IT Journal, ‘Current interpretation software doesn’t understand prestack data, in part because the data volumes are such a challenge. We plan to let interpretive processors interact with prestack data by leveraging GPU-based processing. This will provide access to multiple versions of processed data using data compression to provide rapid access. Our real time compression technology targets QC and parameter selection, bringing processing and interpretation together.’ FineTooth is targeting the major processing houses and ‘any company doing prestack interpretation’. Buch is also CEO of Norwegian Hue Space, a technology supplier for Schlumberger’s GigaViz package.
This article originally appeared in Oil IT Journal 2005 Issue # 11.
For more information or to comment on this topic email here.