What's really at the heart of the ARM-Xilinx deal

10/23/2009 10:06 AM EDT

First and foremost, what took so long? I wondered for years why Xilinx never adopted the ARM architecture like many of its competitors. The answer to that question likely lies in the development of the next-generation AMBA bus specification, which Xilinx is calling "an optimized interconnect standard for FPGAs." The Xilinx designers likely felt that without this new development, they couldn't get the maximum potential from the ARM architecture, and had no desire to be a "me too" FPGA vendor.

In speaking with Xilinx about what this agreement means, it became clear that at least one the three main challenges of FPGAs—that they use too much power—figures prominently in the adoption of ARM. The potentially lower power spec came up again and again as a key feature that Xilinx wants to exploit.

What Xilinx wouldn't talk about is which of the ARM cores it plans to deploy in its FPGAs, or whether those cores will appear in its mainstream Spartan products or the high-end Virtex line. Those details won't be forthcoming for at least a few months. But again, without the correct bus architecture, it wouldn't have mattered as the impact would not have been nearly as significant as it could have been with the proper topology in place.

While the new AMBA spec will likely benefit all ARM processor developers, it includes all the hooks required to maximize the connection within an FPGA architecture. And we may see AMBA pervade FPGAs that don't even contain an ARM core, but is used as the interconnect structure within the FPGA, connecting the traditional FPGA IP, and replacing current interconnect structures.

Note that Xilinx and ARM are not alone in the creation of the forthcoming AMBA spec. It also involves some of the industry leaders, like Cadence, Denali, Mentor Graphics, and Synopsys.