Mercury Systems debuts embedded industry’s broadest bandwidth OpenVPX processor blade
Mercury Systems, Inc. of Chelmsford, MA announced on March 22 the rugged OpenVPX™ Ensemble® LDS6526 processing blade that seamlessly integrates the Intel® Xeon® Processor D-1500 system-on-a-chip (SoC) product family (formerly codenamed “Broadwell DE”), the versatility and performance boost of software-defined off-load processing with built-in, double-bandwidth sensor I/O capability into a powerful 6U form factor blade for streaming signal processing applications.
Mercury’s software-defined, FPGA-based protocol offload engine technology (POET™) combined with Altera’s latest Arria® 10 FPGA effortlessly deliver twice the sensor I/O bandwidth of any other OpenVPX blade with four channels of I/O that can be routed to either the processor or data plane. This innovative embedded technology is switch fabric-agnostic and runs 40Gb/s Ethernet or many other protocols at full speed.
“Ensemble LDS6526 blades are the highest-performing sensor processing blades available in the embedded industry today, with a maximum theoretical processing capability of 576 single-precision GFLOPS, supported by an enhanced sensor I/O bandwidth of 5 GB/sec per channel. These blades are opening the door to a host of new possibilities for sensor chain architects,” said Shaun McQuaid, Director of Product Management, Mercury Systems’ Embedded Products Group. “This new capability comes with SWaP, reliability and affordability bonuses as the LDS6526’s native I/O capabilities eliminate the need for four-channel high-speed mezzanine-based I/O modules,” he added.
LDS6526 blades leverage Arria 10 FPGAs to bridge an additional four channels of external I/O to the standards-based four channel OpenVPX data plane while augmenting the blade’s built-in private and personalized security features. Native I/O capability maintains a direct low-latency signal channel, reduces chassis volume and makes both cooling and software development easier and therefore lower-risk and more affordable. Alternatively, the four sensor I/O channels can be utilized as additional data plane interfaces, vastly increasing high-speed bandwidth and flexibility of configuration. Further, the LDS6526’s built-in sensor I/O preserves the LDS product line’s standard dual embedded mezzanine sites, making these blades exceptionally versatile and scalable.
Both the Xeon D-1500 and Arria 10 FPGA are rugged, extended temperature devices that utilize the latest manufacturing technology to draw less power contributing to the LDS6526’s remarkable SWaP performance. LDS6526 blades are powerful signal ingestion engines that are ideally suited to streaming low-latency signal processing applications including next-generation radar, complex image intelligence (IMINT), multi-functional sensor chain and advanced situational awareness applications.
“Combining an Arria 10 FPGA with a Xeon Processor D-1500 server-class processor make the Ensemble LDS6526 processing blade ideally suited for industrial applications,” said Umar Mughal, director, broadcast, military and medical business group within the Programmable Solutions Group, Intel. “Arria 10 FPGAs provide a low-SWaP floating point processing and acceleration capability that drives POET- applications, giving the LDS6526 its signature processing, I/O bandwidth and deterministic performance and ability for end-users to customize the platform.”
Arria 10 FPGAs supports main processor off-loading, private and personalized security and enables micro via radial interconnect-optimized (MVRI) switch fabrics to be updated, making the LDS6526 exceptionally versatile, capable and fast. Embedded MVRI technology is OpenVPX-compliant and enables rugged OpenVPX subsystems to run switch fabrics at speeds of 40Gb/s and higher.
The Ensemble LDS6526 leverages Mercury’s fourth generation of highly SWaP-efficient packaging technology to securely package the Arria 10 FPGA and Xeon D-1500 devices for reliable deployment right to the tactical edge. The Ensemble LDS6526 secures and reliably cools the best commercial technology to produce a dense, powerful combination of server-class, low-latency front-end FPGA processing with advanced switch fabric management in an open systems blade that is designed and made in the USA.
The LDS6526 as other Ensemble modules supports open data movement middleware, including OpenMPI and OpenMPI/OFED™, next-generation VITA 46.11 system management, and standard optimized math libraries. Production units will become available later this year for air-cooled, conduction-cooled, Air Flow-By™ and Liquid Flow-By™ systems, with Xeon D-1500 options of up to 16 cores. Additional configurations featuring fiber optic I/O capability and VITA 66 compliance are expected to be announced by Mercury later this year.
Source: Mercury Systems, Inc.