Mercury Computer Systems Unveils Multi-GPU Development Platform for Embedded, High-Performance Sensor Stream Computing and Explo

SAN DIEGO , Nov. 17 /PRNewswire-FirstCall/ -- Mercury Computer Systems, Inc. (NASDAQ: MRCY), a leading provider of embedded, high-performance computing solutions for image, sensor, and signal processing applications, unveiled the GPU-based Sensor...


SAN DIEGO , Nov. 17 /PRNewswire-FirstCall/ -- Mercury Computer Systems, Inc. (NASDAQ: MRCY), a leading provider of embedded, high-performance computing solutions for image, sensor, and signal processing applications, unveiled the GPU-based Sensor Stream Computing Platform today at MILCOM 2008 in San Diego, California .

Graphics processing units, or GPUs, are the computing industry's most powerful, programmable floating-point graphics-rendering engines primarily used in personal computers, workstations, and gaming consoles. With recent architectural advancements, the algorithmic scope to which GPUs can be applied has grown dramatically. For traditional signal processing algorithms like the FFT (Fast Fourier Transform), they provide unprecedented performance, particularly performance per watt. GPUs can now be leveraged against the processing and exploitation requirements of growing markets like signals intelligence and oil & gas exploration, with outstanding results.

Historically, the availability of embedded GPU solutions suited to the stringent requirements of high-performance signal processing has been scarce. However, industry performance benchmarks on implementing GPUs in high-performance signal processing applications have shown that GPUs can obtain 20x performance improvement and more over other processors (see the Georgia Tech Research Institute report at http://gpu-vsipl.gtri.gatech.edu/*). With the Mercury Sensor Stream Computing Platform, embedded stream computing customers can benchmark and evaluate application performance in their choice of GPU environments, and then migrate to a larger deployed solution.

Built on in-depth, unparalleled expertise in algorithm and processor optimization, the VXS-based Mercury Sensor Stream Computing Platform offers unsurpassed scalability in compute power, performance, and thermal management, and allows for much greater, tunable performance for a variety of commercial and defense applications. The Platform leverages a dual dual-core Intel Xeon-based VX6-200 single-board computer (SBC), which offers unprecedented levels of compute performance and a wide selection of I/O interfaces. At the heart of the Platform is the VXS-GSC5200 dual MXM GPU module, which delivers very-high bandwidth performance to each GPU from the host, as well as between GPUs. Each MXM GPU module can drive up to 3 display monitors (1 analog and 2 digital).

"The Mercury Sensor Stream Computing Platform enables customers to deploy the power of GPU image processing in a rugged form factor for defense applications; and in this case, VXS," said Greg Tiedemann , Product Line Director for Advanced Computing Solutions at Mercury Computer Systems. "GPUs are widely used today for a number applications targeted at research and scientific discovery. With the Sensor Stream Computing Platform, our customers can begin to evaluate various compute-intensive applications, obtain peak performance out of their application, and then migrate to a deployed solution

that enables them to put the data exploitation processing on the platform -- and closer to the sensor."

Moving data exploitation processing on the platform and closer to the sensor is especially important, as the time it takes to get information out of today's dissemination architecture is not aligned with the tempo of market requirements -- whether the market is semiconductor inspection or ISR (Intelligence, Surveillance, and Reconnaissance). For ISR in particular, ground-based exploitation takes tremendous amounts of time; data processing bandwidth; and resources, or information analysts, who need to apply multiple looks (sensors) to accurately survey a potential threat, and subsequently exploit the data manually.

This content continues onto the next page...