The Mercury Computer Systems ILK8X Interlink Module is a high-performance switching fabric that transforms VMEbus chassis from single transaction buses into scalable real-time interconnects. Built on RACE++ Switched Interconnect technology, the ILK8X delivers deterministically low-latency communication between processing nodes through eight-port crossbars deployed onboard and across multiple boards. The module connects each compute node to the RACEway switch fabric via an interface FPGA, enabling simultaneous multi-path data transfer optimized for real-time embedded applications.
Technical Specifications
• Interconnect Technology: RACE++ Switched Interconnect (second-generation RACEway switch fabric)
• Operating Frequency: 66.66 MHz
• Bisectional Bandwidth: Over 2.1 GB/s
• Per-Crossbar Peak Bandwidth: 267 MB/s per connection
• Simultaneous Paths: Four concurrent communication paths per crossbar; total peak bandwidth exceeding 1 GB/s
• Aggregate VME Chassis Bandwidth: Exceeding 1 GB/s
• Latency (Connection Establishment): 75 ns per crossbar at 66.66 MHz
• Latency (Steady State): 15 ns per crossbar
• Memory Support: DDR-266 133 MHz Double Data-Rate SDRAM
• Routing: Adaptive routing within RACE++ interconnect
• Power Design: Low power, field-tested architecture
– Key Features
• Multiple simultaneous high-speed communication paths via RACEway Interlink Standard
• Eight-port crossbar topology for flexible node interconnection
• Interface FPGA bridges compute nodes to RACEway fabric
• Two-stage crossbar architecture: primary set on VME motherboard, secondary set on Interlink modules
• Converts traditional VME bus topology to deterministic real-time fabric
– Typical Applications
• Real-time processing systems requiring sub-microsecond latency
• Data-intensive embedded computing tasks
• Multi-node systems demanding high aggregate throughput
• VME-based avionics and defense platforms
– Compatibility & Integration
The ILK8X integrates with RACE++ VME motherboards and connects through the RACEway Interlink Standard to form larger switch fabric deployments across multiple chassis.
