Lattice Semiconductor has announced that Mercury Computer, a computing solutions provider to government, defense, energy, medical and telecommunications providers, has ported its high-performance Serial RapidIO Intellectual Property (IP) product line, consisting of ×1 and ×4 endpoint IP cores, to the LatticeSC family of 90nm (nanometer) Extreme Performance FPGAs. In addition, Mercury has joined Lattice's ispLeverCORE Connections IP partner program. Working together, Mercury and Lattice will develop and deliver complete system solutions for mutual customers requiring high-performance Serial RapidIO connectivity supported by Lattice's FPGA product families.
Mercury achieved its target performance of 12.5 Gbit/s with a comfortable guardband on the LatticeSC family by taking advantage of the clock boosting technology incorporated within Lattice's ispLEVER version 6.0 SP1 software design tool suite. This feature automatically optimizes setup and clock-to-output times in register-to-register paths to improve timing performance. The Mercury core also takes advantage of the Serial RapidIO protocol support embedded in the LatticeSC devices' powerful Physical Coding Sublayer (PCS), specifically the complete PCS/PMA layers of the Serial RapidIO Rev 1.2 specification blocks.
A co-founder and steering committee member of the RapidIO Trade Association, Mercury is an IP provider of Serial RapidIO technology, as well as a system solution developer of Serial RapidIO computing platforms for networking, embedded, and storage applications. It has hardware-validated the ×1 and ×4 configurations of the core using the LatticeSC Communications Board with the Lattice LFSC25 device. These cores operate with significant timing margin in lower cost LatticeSC speed grades, so designers can optimize system cost and be confident they can easily implement the cores to meet timing requirements in the most demanding applications.
Over the last two years, Mercury has established itself as the leading provider of system-level IP for Serial RapidIO interfaces, and has successfully demonstrated IP working in a range of silicon technologies, including FPGA, structured ASIC and ASIC.
About the Mercury Serial RapidIO core
The Serial RapidIO Core is a complete high-performance, high-function core based on an industry-standard, high-performance packet-based interconnect technology. It incorporates a logical layer, a transport layer and a physical layer, and supports input/output and message-passing logical layer standards. It is completely compliant with the RapidIO 1.2 specification.
With the adoption of Serial Rapid IO endpoints in FPGAs, DSPs and microprocessors, and switch fabrics, the IP offered by Mercury is ideal for creating superior time-to-market solutions. The applications for this IP are numerous, including creation of end-to-end solutions. Designers can create endpoints and also easily implement interfaces from proprietary/standard buses to Serial Rapid IO, or create multiport switches to aggregate other Serial Rapid IO inputs.
About LatticeSC extreme performance FPGAs
Announced in February 2006, LatticeSC Extreme Performance FPGAs deliver the highest performance and most robust feature set of any programmable logic product in the industry. The LatticeSC FPGA combines up to thirty-two 3.8 Gbps SERDES channels with an innovative Physical Coding Sublayer (PCS) to provide a breadth of support for interface protocols including Serial RapidIO, PCI Express, Ethernet, Fiber Channel, XAUI and SONET/SDH. Source synchronous I/O standards such as RapidIO, HyperTransport, SPI4.2, SFI-4, UTOPIA, XGMII and CSIX, and memory standards including SDR, DDR1, DDR2, QDR2 and RLDRAM are implemented with dedicated PURESPEED I/O logic that delivers up to 2 Gbps parallel I/O performance, the fastest in the industry. The LatticeSCM devices also include exclusive MACO (Masked Array for Cost Optimization) pre-engineered IP in structured ASIC blocks for low cost and low power system-level integration.
LatticeSC device sizes range from 15,000 to 115,000 Look-up Tables (LUTs) with up to 7.8 Megabits of block RAM on a single chip. A variety of family members are currently available including the LatticeSC15, LatticeSC25 and LatticeSC80.