Product Tour: CHAMP-AV9 6U OpenVPX DSP Module
April 09, 2015
Processing Huge Sensor Data Streams in Real-time
Today’s ISR sensors generate unrelenting streams of data that must be processed in real time. Curtiss-Wright’s 6U OpenVPX CHAMP-AV9 Intel Digital Signal Processor was designed to meet that need.
Market Leading Performance
The CHAMP-AV9 is powered by two Intel Core-i7 quad-core processors, each with integrated vector processing and an on-chip Graphics Processing Unit, a GPU. Including the GPUs, a CHAMP-AV9 delivers an unprecedented floating point performance of 1.3 Teraflops.
A Balanced Design
There is abundant memory and flexible I/O to match the processing power, including
- Six DisplayPort interfaces
- 16 GB DDR3 SDRAM
- 160 GB/s on-board PCIe switching - and
- Four 40 Gigabit Ethernet or InfiniBand interfaces, at an aggregate 112 Gb/s
Designed with Fabric40 Technology
Throughout the board, Curtiss-Wright Fabric40 technology is used to ensure signal integrity even at extreme rates.
Video transcript for the hearing and visually impaired
Today’s ISR applications all employ incredibly precise and wide-ranging sensors. These sensors generate huge Datastreams, with valuable information for our warfighters. To extract that information, and deliver it quickly, the datastreams must be processed in real time.
The CHAMP-AV9 was designed to meet that need. A 6U OpenVPX COTS module, the CHAMP-AV9 has two Intel Core i7 quad-core processors, with integrated vector processing and an on-chip Graphics Processing Unit, a GPU.
To match the processing power of the two Core i7’, the CHAMP-AV9 is designed with six DisplayPort interfaces, 16 Gigabytes of DDR3 SDRAM, 160 Gigabits per second of on-board PCIe switching and four 40 Gigabit Ethernet or InfiniBand interfaces, at an aggregate 112 Gigabits per second.
Throughout the board, Curtiss-Wright Fabric40 technology is used to ensure signal integrity, even at extreme rates.
For ISR, performance is key and the CHAMP-AV9 delivers a 2-4x leap over other DSP boards. A lot of that performance comes from a unique component, the integrated GPU. Until recently, GPUs existed as unique chips, discrete silicon tied to a host CPU. Now, a GPU resides within the Core i7 quad-core die. Cache memory is shared among four cores and the integrated GPU over a high speed ring bus at 300 Gigabytes per sec.
GPUs were first designed to render high speed video with a parallel architecture using multiple ‘shaders’. Each shader is a floating point math processing engine, making them ideal not just for graphics but also for executing the complex algorithms used within ISR applications.
A Core i7’s integrated GPU delivers 350 Gigaflops of processing – adding in 307 GFLOPS from the four CPU cores, the dual Core i7 CHAMP-AV9 has an impressive 1.3 Teraflops. It saves size, weight and power, with no need for a separate GPU board. And, with processing balanced by ultra-high bandwidth communications, it is optimized for the most demanding ISR applications.