The Future Of IP

By Ed Sperling
The rapid consolidation of the IP business is raising big questions about who will be left, whether new companies will join, and what it means for chipmakers looking to buy IP.

In a period of one month Synopsys bought Virage Logic, which had just finished a buying spree of its own with the acquisitions of ARC and the IP business of NXP, and Cadence bought Denali. So what exactly does this mean for chipmakers? There are fundamental questions that need to be answered to fully understand this market.

Why do companies choose one vendor’s IP over another? The answer to this question is more driven by performance, power, cost and ecosystems than it is by who develops the IP. Processor IP cores from ARM, for example, are dominant in the broad smart phone world because of the ecosystem of embedded and other software surrounding it. MIPS is more associated with the burgeoning Android market and specific industrial and commercial applications. And ARC has developed strong ties in the audio world.

“If you pick a problem to solve and develop a complete solution you can solve a customer’s problem,” said Mark Throndson, director of marketing at MIPS. “Virage/ARC has focused very vertically on problems and been successful with that. When it comes to ARM and MIPS you have to go into the nuts and bolts of RTL and our coherent multicore vs. the ARM Cortex A9. It’s architecture vs. architecture, the whole software stack and the marketing side.”

The decisions reach deeper into the ecosystem than just the IP vendor, though. Foundries certify based upon manufacturability, and both the Common Platform and TSMC have their own programs. Within each category there are at least two vendors whose IP has been certified, said Dan Kochpatcharin, deputy director of IP portfolio marketing at TSMC. “We have about 40 companies in our IP alliance,” he said. “We don’t pick and choose the IP. That’s up to our customers. But we do have a service that qualifies it.”

It also adds another check point for IP. In addition to price, performance, power, interoperability, ecosystem partners and licensing arrangements, you have to be sure it can be manufactured.

“We recently sponsored a perception audit to evaluate what customers consider important from both a business and a technical perspective when choosing standards-based IP,” said John Koeter, vice president of marketing for Synopsys’ Solutions Group. “From a technical perspective the most important criteria by far is that the IP has passed compliance and is certified. For standards-based IP cores such as USB and PCI Express, this gives designers confidence that the IP is functioning precisely to the specification and interoperable with other products in the market. An IP vendor’s engineering knowledge and having IP in volume production are the second and third most important aspects, respectively. Having engineering expertise gives customers assurance that there will be someone who has the technical know-how to address their issues whenever needed. IP that is in volume production means that the IP has been widely used and deployed, which lowers their integration risk.”

He noted that from a business perspective the top three factors are responsiveness of the IP vendor, a broad portfolio and cost.

Is consolidation good or bad for IP customers? The answer to this question is multifaceted and complex. The most accurate answer is, it depends. While the big IP vendors get acquired by big EDA vendors, new ones are forming in India and China. They are expected to begin showing up over the next 12 to 24 months en masse. And it’s not just India and China, where the bulk of IP is being created by startups.

The large IDMs in Japan are planning to make some of their IP commercially available in the Japanese market over the next year, with global ambitions for selling that IP after that.

Kochpatcharin, for one, believes that consolidation is a healthy trend because it creates an opening for new startups, which can’t always rise above the din of dozens of midsize IP vendors to effectively market their IP. But the transition isn’t always smooth. “When it goes to fewer vendors the business bundling gets easier, but in a customer tapeout that’s usually not the deciding factor. With smaller IP vendors it’s easier to get IP that’s customized to their needs.”

But not all IP is outsourced, either. Synopsys estimates that only about 65% to 70% of standards-based IP is outsourced. The rest is developed by companies themselves, which is still an option.

And despite all of this, there appears to be no slowdown in IP growth. The whole market is growing. Ron Moore, director of PIPD strategic accounts marketing at ARM, said that growth is likely to continue for several reasons. “First of all, the explosion of mobility devices that support multiple radios, multi-processing, and apps ecosystem will rely more on commercial IP for the system. Second, many semiconductor companies are focusing their flat (or shrinking) hardware design resources on SoC integration of high-value parts rather than building reusable IP. And third, the continued growth of the fabless or fab-lite business model means that the deep knowledge of the process technology required to co-optimize the IP will be concentrated in the leading IP companies.”

Is integration easier when large EDA companies or IP vendors control more of the pieces? Answer: Sometimes no, sometimes yes.

“Integration is not the main driver of this consolidation,” said Yankin Tanurhan, vice president and general manager of Virage’s processor, SoC infrastructure and NVM solutions. “But the advantage of consolidation is that everything is under one roof so you can make sure it’s compatible. On the other hand, monopolies don’t work. You need a combination of open standards so you can mix and match as much as possible. Integration will be easier, but there is still strong demand by companies not to rely on a single vendor.”

Not everyone believes integration is easier if it all comes from a single vendor, however. Kochpatcharin said that in some cases it’s better—particularly when it involves a controller and PHY from the same company—but two different types of IP do not necessarily work better together.

Moreover, those IP blocks may have been developed by different teams, in different locations and at different times. In fact, they may have been developed by different countries that were brought together for economic reasons. But as Tanurhan noted, it’s still easier to address interoperability issue in a big development organization than in 100 different companies.

The most effective way to ensure interoperability isn’t unifying development under a single corporate umbrella. It’s establishing standards such as IP-XACT, which has emerged as the most prominent standard in the IP industry. “With hard IP it will be different, but with soft IP standards will drive the ability to integrate,” said Tanurhan.

Will the role of EDA companies change when it comes to IP? Answer: Yes.

John Bruggeman, chief marketing officer at Cadence, has been quite consistent in saying that the challenge of the EDA industry is integrating IP and software as a way of getting the cost down in chip development.

“The role for EDA is where we enable OEMs to define and specify a system that optimizes system verification, design and modeling,” Bruggeman said. “Once you specify the system model you should be able to pump it into the supply chain and have it deliver against the spec.”

IP is a big part of that strategy, which is why Cadence bought Denali. The company builds memory models, which provide that spec. “What will change for EDA is that we will have three customers in the future vs. one today, which is the semiconductor company. In the future the OEM will become a customer because they will buy the platform, and the foundry will become a customer because the system model will be handled from OEM to foundry.”

Mentor and Synopsys are both taking similar routes, even though they are less vocal about the changes. Synopsys is working on software prototyping and standard IP, while Mentor is providing the tools for creating larger models and ensuring the manufacturability. And while there is divergence in the approaches among all three of the big EDA vendors, there is also cohesiveness in their recognition that big changes are afoot and various different approaches will be necessary.

How will IP change in the future? The biggest change for IP at future nodes may be around liability and flexibility rather than the IP itself. In a 3D chip, for example, IP that may be perfectly good on a single die may not work. In fact, an entire chip may not work, which likely will result in some nasty finger-pointing comparable to when the IP industry first got going.

“This is one of the ugly sides of 3D,” said Tanurhan. “The level of qualification might have to change.”