The Mercury Computer Systems 910-08092 is a RACE Series daughtercard built around a 400-MHz PowerPC 7400 processor for high-performance embedded computing. It integrates AltiVec vector acceleration technology to handle computationally intensive image and signal processing workloads. The daughtercard plugs into RACEway backplanes, enabling multiple concurrent transactions across a switched-fabric architecture that eliminates data starvation and delivers simultaneous input, output, and computing capabilities in real-time sensor acquisition and display systems.
Technical Specifications
• Processor: PowerPC 7400 (G4) superscalar RISC with AltiVec parallel vector execution
• Clock Speed: 400 MHz
• Execution Units: Seven total—two integer units, one load/store, one branch/system, one floating-point, one AltiVec ALU, and one AltiVec Permute
• AltiVec Capability: Processes up to four floating-point numbers or sixteen 8-bit integers per vector operation
• L2 Cache: Up to 2 MB supported
• Memory Interface: 125 MHz or 133 MHz maximum (DDR SDRAM capable)
• Fabric Interface: 267 MB/s per compute node
• RACE++ Crossbar: 66.66 MHz per connection, 267 MB/s peak bandwidth per path; four simultaneous paths support over 1 GB/s aggregate
• Latency: ~75 ns per crossbar during connection establishment, ~15 ns per crossbar thereafter
– Key Features
• AltiVec acceleration for vector-parallel workloads
• RACE++ switched-fabric architecture prevents performance degradation from I/O bottlenecks
• Dedicated fabric interface with 267 MB/s throughput
• Multi-gigabit crossbar bandwidth through four concurrent communication paths
• Superscalar execution with seven independent execution units
– Typical Applications
• Real-time signal and image processing
• Sensor data acquisition and processing
• Display rendering in time-critical systems
• High-throughput embedded computing requiring sustained I/O and processing concurrency
– Compatibility & Integration
The daughtercard integrates into Mercury RACEway and RACE++ backplane architectures, providing seamless pairing with compatible motherboards and enabling multi-processor, multi-I/O configurations without performance degradation.
