Xilinx Demonstrates Industry's First QPI 1.1 Interface with FPGAs at Intel Developer Forum

This solution achieves the best possible overall system performance and power consumption by leveraging Xilinx FPGAs high performance processing and flexible I/O capabilities.

With today's demonstration, Xilinx now has the first FPGA-based low-latency, high-bandwidth interconnect intellectual property core for Intel Xeon processor-based systems, said Nick Possley, senior director of wired communications at Xilinx. This means Xilinx solutions can now play a major role in reducing CAPEX and OPEX for Data Centers by enabling higher compute performance within a given server form factor either by directly accelerating applications or offloading I/O intensive operations.

Hardware designers can start developing a QPI-based solution today using the Xilinx QPI development platform complete with IP core and a development module designed to plug directly into an existing Intel Sandy Bridge CPU socket.

The QPI 1.1 full width link between the Xilinx Virtex-7 FPGA and the Intel Sandy Bridge CPU operates at 6.4 Gb/s per lane over 20 lanes. Xilinx has created a custom development board that mounts a Virtex-7 FPGA directly in an Intel Sandy Bridge Xeon CPU socket. The demonstration leverages the Native Loopback example hardware and software to validate the data exchange between FPGA and CPU. The QPI interface enables the Intel Xeon CPU to take advantage of the Xilinx FPGA's ability to do parallel processing and speed up high computation applications for coprocessing and/or application acceleration. Developers can also use the Xilinx solution to enable a high-performance, low-latency network interface controller and I/O connectivity expansion taking advantage of the cache coherent nature of the QPI protocol to more efficiently handle packet processing.