Field Notes from DAC 2008: IP—The Good, The Bad, and The Ugly

The DAC Pavilion panel on purchased IP provided a great mini-tutorial on selecting IP and IP vendors for SOC designs. The panel chair was iSuppli’s Jordan Selburn and the panelists included Kathy Werner (Freescale), Joachim Kunkel (Synopsys), and Satya Gupta (Open-Silicon). Werner recently completed a stint as head of the VSIA, which created an IP quality certification process now in the hands of the IEEE. Kunkel is the VP/GM of the System Level Design business unit at Synopsys. Gupta is a Co-Founder and VP of Engineering at Open-Silicon.

Selburn set the scene by noting that, in the early days of the IP land rush during the late 1990s, there were as many as 600 companies offering various IP blocks to SOC design teams. Of those, about 100 companies were substantive ventures and the other 500 or so were a couple of guys in a garage. As a result, IP quickly got a bad name for poor quality, inadequate support, no verification, and other problems. However, said Selburne, there’s no way that today’s semiconductor industry could have reached its current $300 billion in sales; many of today’s chips are just too complex to design without using purchased IP.

“The key,” said Werner, “is to figure out how good the IP is. IP selection is based on business, technology, functionality, and quality.” Kunkel agreed: “When IP is based on standards, quality is the differentiator.” “DAC is now an EDA plus IP show,” said Gupta.

Selburn then asked the panelists to define a good IP engagement.

Werner said there were several aspects. First, it must be a partnership. The [IP] vendor must have a stake in the relationship. [That generally means a royalty arrangement.] There are always issues that come up and the vendor must have skin in the game.

Kunkel agreed. “Remember what business we’re in,” he said, “Design Productivity.” We make the customer’s engineers more efficient with off-the-shelf and specialty IP. However, said Kunkel, many prospects reject the idea of IP royalties. They prefer IP buyouts and this creates a dilemma. Buyouts essentially take the IP vendors’ skin out of the game [by reducing the economic incentive to support the customer’s continued success]. “We need to align interests,” he continued, implying that the elimination of royalties causes vendor and customer interests to diverge.

“The problem,” replied Werner, “is that lower margins squeeze royalties. Something’s gotta’ give.”

Then Gupta asked, “Who will share the risks when a mask spin is needed because of bad IP?”

“Liabilities and responsibilities are open for discussion,” replied Kunkel.

Gupta continued, “Customer relationships define a good IP vendor. About 5-10% of the time, things go bad. You need an IP vendor who will stand with the IC vendor [when this happens].

“The problem is—where is the problem?” countered Kunkel. “Is it in the IP or elsewhere in the system? We’re all in the same boat. IF the chips don’t sell, no one makes money.”

Finally, the panel tackled the issue of the smaller IP vendor. “Most IP vendors have good technology,” said Gupta, “but they can’t participate in an IP selection if they’re small.” They can’t afford the pre-sales effort. They can’t afford to support silicon integration. Werner commented: “For a small IP vendor to be considered, they must have highly differentiated IP.”