Semiconductor IP News and Trends Blog
Technical Trade-offs Leave Long Tail
Architectural trade-offs – typically resulting in IP – made early in a design can affect a company’s market participation for years to come.
Does the path we chose forever lock us into a specific fate? Does it severely lessen our options?
These were questions that came to mind during a recent interview with Denny Steele, VP of solutions marketing at Silicon Blue – an ultra low power FPGA company.
Steele was explaining the company’s technological advantage in the ultra-low power, niche mobile consumer market. He side that the advantage comes from architectural trade-offs that favored low-power over high performance. “Every trade-off we made was for low power,” explained Steele. These trade-offs resulted in 30 to 50 patents for low-power, FPGA architectural designs.
In contrast, companies like Xilinx and Altera tended to optimize for performance in their trade-offs, Steele noted. That’s because the big telecom market – which they service – is all about performance, e.g., higher throughputs and faster pipes.
Then came the comment that struck me a both obvious but odd. Steel observed that any engineer at Xilinx or Altera could have chosen to optimize exclusively for low power, as had been done at Silicon Blue. Instead, they optimized for performance. Now, Silicon Blue has made it very difficult for them to go back and re-optimzed for low power, due to all the low-power IP patents held by the company.
I understand that much of Steele’s comments are just market “speak.” Xilinx has technology that supports both low power and high performance, whereas Silicon Blue has a niche position on ultra-low power and medium performance.
But the trade-off choices made by each company do set their fate. At one point, Xilinx could have bought Silicon Blue for the patents. Lattice’s recent acquisition of the former now places those ultra-low power patents even further from Xilinx’s reach, since it seems unlike that Xilinx would by Lattice.
In theory, Xilinx could innovate new ways to architect even lower power into their designs – which they can do in part thanks to Moores Law. Or Xilinx my concede the smaller niche market to Lattice-Silicon Blue. In reality, that choice may well have been made at the beginning, when the first architectural decisions were being made.