Product category:
VMEbus Boards and Assemblies
News Release from: Force Computers | Subject: Champ-AV II
Edited by the Electronicstalk Editorial
Team on 08 May 2002
Big bandwidth boosts DSP performance on
VME board
New from Dy 4 Systems, the Champ-AV II quad 500MHz PowerPC 7410 board uses Dy 4's novel QuadFlow nonblocking architecture.
New from Dy 4 Systems, the Champ-AV II quad 500MHz PowerPC 7410 board uses Dy 4's novel QuadFlow nonblocking architecture QuadFlow provides for exceptionally high total node-to-node data flow of up to 2Gbyte/s, which combines with the power of the AltiVec DSP enhancements of the 7410 to offer the highest performance of any VMEbus DSP product
The new board embodies Dy 4's vision to build hardware platforms that simplify software development.
The QuadFlow architecture is both faster and less restrictive, enabling software developers to focus on algorithms rather than on optimising systems originally developed for lower bandwidth boards.
The new board is available in both air- and conduction-cooled versions and is intended for complex defence and aerospace DSP applications that demand the use of large numbers of processors and a high-performance switch-fabric interconnect, such as sonar, radar and signal intelligence applications.
It is also ideal for the real-time data needs of medical imaging, seismic surveys and other commercial purposes.
The Champ-AV II, Dy 4's second-generation DSP board, employs an independent-node architecture.
Each of the four 500MHz processors has 2Mbyte of L2 cache and its own bank of SDRAM, up to 256Mbyte for each processor, ensuring that memory-intensive applications will not suffer performance degradation compared with alternative shared memory architectures.
The board provides each processing node with dual 66MHz, 64bit PCI connections to adjacent nodes, resulting in peak bandwidth of 1056Mbyte/s into each node, and up to 2Gbyte/s peak aggregate bandwidth.
Instead of using PCI buses to relay interrupt messages, the nodes incorporate a separate hardware interface to accelerate inter-processor messaging.
This significantly reduces system latency, as interrupts are routed independently of data traffic on the PCI buses.
In addition, the Champ-AV II supports a synchronised timebase across multiple boards for time stamping of data and messages in complex data processing applications.
The Champ-AV II has two PMC sites that support 64bit/66MHz transfers, with a peak rate of 528Mbyte/s.
Each PMC interface is mapped to all four processor nodes, enabling the transfer of data between any node and any PMC interface.
Both PMC sites support high-speed low voltage differential signalling (LVDS) I/O connections between the PMC site and the backplane connectors, so that the latest generation of high-performance digital interfaces can be employed.
These include Dy 4's new StarLink switch-fabric PMC or the PMC-643 Fibre Channel module.
PMC I/O is available on both the front panel and the P2/P0 backplane connectors.
When coupled with two StarLink PMCs, the Champ-AV II can provide up to 3.2Gbyte/s of I/O bandwidth between the PMCs and the I/O connectors, enabling the switching functions to be integrated onto the card.
Software support for the Champ-AV II includes Wind River Systems Tornado II/VxWorks and VSPWorks.
Customers can choose Dy 4's IXLibs-AV library of fully optimised DSP functions designed for the board's AltiVec instruction unit.
VSI/Pro VSIPL libraries are also available from MPI Software Technology.
Steve Blackman, director of marketing and business development at Wind River Systems' Aerospace and Defense Business Unit said: "The Champ-AV II is an excellent platform for customers using Tornado II VxWorks and/or VSPWorks in radar/sonar/and signal analysis applications.
The technology is a great fit with VxWorks and VSPworks, since multiple Champ-AV II cards can be connected together with Dy 4's StarLink switch-fabric.
The environment simplifies the writing of applications for distributed computing systems and provides a truly deterministic solution for hard real-time applications that can be easily scaled with the addition of multiple boards and multiple virtual processors".
• Force Computers: contact details and other news
• Email this article to a colleague
• Register for the free Electronicstalk email newsletter
• Electronicstalk Home Page
Search the Pro-Talk network of sites
