GPU-based simulator targets nano-scale flow in unconventionals

Nvidia GPGPU conference presentation lifts veil on Ridgeway Kite’s new simulator.

In his presentation at the Nvidia GPGPU conference earlier this year, Garf Bowen of UK-based startup Ridgeway Kite unveiled its work-in-progress reservoir simulator. Ridgeway is working on a full physics toolset for characterizing unconventional shale reservoirs. The modeler computes multi-phase thermodynamic equilibrium in nanometer scale pore spaces including capillary pressure, vapor-liquid equilibrium, multiple porosities and a ‘pseudo-geomechanical’ approach to describe fracture growth.

Small scale modeling requires big compute resources - to provide the ‘maximum possible parallelism.’ Here Ridgeway has developed the 'extensible parallel library’ (XPL), a library for ‘massive’ code parallelization that runs across ‘all current hardware types.’ The intention is that it will also run on most future systems too.

However, coding its linear solver in XPL is a significant overhead. Ridgeway has been tempted from the pure device-independent approach by Nvidia’s AmgX library. This provides a 'free’ pathway to accelerated core solver technology on GPUs with an 'up to 10x’ acceleration for the computationally intense linear solver portion of simulations.

Ridgeway’s software is tested on two systems housed by the UK Center for Innovation. One, Emerald, is a 114 teraflop machine with 372 NVIDIA Tesla processors. The other, IRIDIS is a 12,000 core Intel Westmere-based cluster with a 108 teraflop bandwidth. Ridgeway plans to release its reservoir simulator in 2015. More from Ridgeway Kite.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.