HueSpace, Lenovo, Nvidia, Magma team on seismic workstation (July 2014)Lenovo moves upscale with acquisition of IBM's x86 business, unveils teraflop system.
Oslo, Norway-headquartered HueSpace has partnered with PC manufacturer Lenovo, Nvidia and hardware specialist Magma on an appliance for high-end seismic visualization development. Hardware comprises a Lenovo ThinkStation D30 workstation, Nvidia Tesla K40 GPU accelerators and Magma’s GPU expansion technology. The HueSpace API is used to render large data volumes such as pre-stack seismics. The 64-bit Linux/Windows environment is capable of handling terabyte ‘and even petabyte’ data sets.
HueSpace has been using GPUs, ‘even before Cuda was invented,’ for ‘general purpose computing on graphics processing units’ as opposed to their use for display. Connectivity between the ThinkStation and the Teslas is provided by Magma’s ExpressBox 3600 GPU expansion system. The ExpressBox can house up to 9 K40 GPUs providing some 13 teraflops of double precision compute bandwidth.
HueSpace exposes a plug-in environment for in-house and third party developers of upstream data visualization offering a range of domain specific functions (such as automatic parallelization across different accelerators), data management and I/O. The solution runs on Linux and Windows. Linux is preferable as it is better at handling a larger number of GPUs.
HueSpace’s core engine runs optimized, multi-threaded C++ throughout the pipeline, using GPUs wherever it can, for instance while transparently compressing and decompressing seismic data. General purpose computation is performed on the HueSpace compute plug-in. The HueSpace API is offered as .NET/C++/Java on Windows, and C++/Java on Linux. The announcement is a foretaste of Lenovo’s move upscale with its acquisition of IBM’s x86 business. More from HueSpace.
Click here to comment on this article
If your browser does not work with the MailTo button, send mail to firstname.lastname@example.org with OilIT_1407_6 as the subject. Web use only - not for intranet/corporate use. Copyright © 2014 The Data Room - all rights reserved.