A discussion at Embedded Systems Conference with Dave Tokic and Mark Jensen of Xilinx crystallized a feeling I’ve had since Xilinx released the first of its Targeted Design Platforms two years ago (http://www.fpgagurus.edn.com/blog/fpga-gurus-blog/further-tdp-adventures) . Xilinx, as well as Altera and Lattice Semiconductor, are emphasizing IP software and subsystem hardware partners for new architectures as much as the companies talk about the capabilities of the FPGA itself. Microsemi, since it acquired Actel Corp., hasn’t gone quite as far, but is emphasizing mixed-signal markets where its own power and analog ICs can be used alongside the Actel FPGAs.

The startup community, meanwhile, emphasizes device capability above all else, which is why an FPGA from SiliconBlue, Achronix, or Tabula may be chosen for its logic efficiency or power dissipation, but rarely due to the relationships the companies have assembled. But as Tokic stressed, if you’re not establishing a tiered partner program with a very high level of trust defined for the premier partners, you may find some customers are reticent to go beyond a single design. And now that FPGAs are being used by OEMs to move into full system production with a proprietary design, that decision can involve large numbers of highly-integrated SoCs.

Tokic is senior director of Xilinx’s partner ecosystems and alliances program, and Jensen is director of processing platform marketing. Their job of pulling together system alliances in vertical markets already was made more difficult when the Xilinx-7 family of new FPGAs split into three (Arrix, Kintex, and Virtex), but moved up a notch in complexity when Xilinx offered the high-end ARM cores in a dedicated family called Zynq-7000. Anyone who thought that Zynq might require a proactive marketing program to develop a fan base did not realize the pent-up demand for RISC-style cores at the heart of FPGAs. Tokic said that “we need to control the rush of folks wanting to get on the platform.”

Jensen, with experience in vertical marketing into sectors like automotive and industrial, said that Xilinx could enhance its mixed-signal offering by expanding support for analog cores such as pulse width modulation controllers and sensor support logic. But just as important as the FPGA library is the addition of software specialists in these realms, similar to the software partners Xilinx has added in DSP realms such as image processing and parallel signal processing.

The companies with system-level platforms must use caution in letting their closest partners know the precise stage that next-generation core development is at. Xilinx, for example, worked with ARM to create a new generation of the Advanced Extensible Interconnect, called AXI-4, and it standardizing that link for all Zynq devices(http://www.fpgagurus.edn.com/blog/fpga-gurus-blog/xilinx-christens-zynq-7000-family-embedded-arm) . But partners say off the record they are still a little worred about the arrival of AXI-4 cores in a timely fashion.

In any event, the established players in the FPGA industry have made the perceptual transition of realizing they are selling system platforms extending across multiple generations. FPGA startups who care about becoming major players need to follow this path of being less concerned about the individual device.