Historically, designing IP has not been seen as "sexy" as say, designing chips. These days though, there's less and less original design in chips as engineers rely more on reusing other engineers' designs. The edict is coming down from management to innovate at the system level, assembling design blocks from available sources both internal and external to the company. Engineers that do have the uncommon luxury of doing original design need to ensure that others are able to leverage their work and are putting their design expertise and creativity into creation of reusable IP. System design is sexy. IP design is sexy. Chip design is now "manufacturing". Why the sudden shift? The reason is simple: chips are getting very complex, driven by convergent devices like cell phones.
No longer is IP only basic building block components like phase-locked loops (PLLs), standard cells, and library generators; nor is it only larger functional blocks like central processing units (CPUs) and digital signal processors (DSPs). Today, IP is needed for "killer functions" ranging from MPEG decode and audio processing to wireless data. And, it is comprised of both hardware and software like firmware, protocol stacks, verification models, and even applications. It has become what some analysts now refer to as subsystem IP. It is not just a solution for increasing productivity. IP is now a critical component of any successful product development. Bryan Lewis, of Gartner, refers to this movement toward reuse of subsystems as a key characteristic of second-generation System-on-Chips (SoCs).
Why exactly has IP evolved into subsystems and what role will those subsystems play in enabling next-generation electronic products? Consider that in the old days, companies provided part of the IP solution to customers. Customers then finished off the solution by developing or finding other sources for the missing parts, often times the software (e.g., drivers). These days, with IP functionality becoming increasingly bigger and more complex, customers no longer have the time, or vertical domain experience to assemble the pieces. They want a shrink-wrapped, ready-to-go solution that they can drop in and forget. As such, IP vendors now have to deliver complete subsystem solutions that comprise hardware as well as software and application-level expertise.
The high-fidelity Bluetooth headset offers a prime example of this new type of IP (see Figure 1). Comprised of everything required for this application, including the Bluetooth system, DSP for audio and Bluetooth protocol stack, it can be easily integrated into an array of electronic devices including a laptop, MP3 player or PDA.
Figure 1. The hardware architecture of IPextreme's CoolBlue high-fidelity Audio over Bluetooth IP subsystem utilizing NXP's CoolFlux DSP core.
The benefits of using subsystem IP are obvious. Development of such an IP requires expertise in multiple domains: wireless, multimedia, DSP, and consumer software applications. It is put together in a package that is pre-optimized, pre-integrated, tested, certified, prototyped, and ready to use as a "black box". It dramatically reduces uncertainty and risk to the chip designer that the chip will come back working improperly.
There is little doubt that the nature of IP has been forced to evolve over the years to keep pace with the needs of the electronics industry, but where will it go in the future? While the role of IP will undoubtedly grow in the future, it seems likely that the IP industry will segment into two types of companies. The first will focus on delivery of horizontal, low-level building blocks which can be used in a wide variety of applications, such as a Bluetooth core. The other type of company will deliver solutions that pull many of those building blocks together to create market-specific subsystems; a prime example of which is the high-fidelity audio Bluetooth subsystem mentioned earlier.
Ten years ago, it wasn't clear whether advances in EDA or adoption of IP reuse would be the solution for what was called the "productivity gap". The market has voted and has shown that IP is the winner, outpacing EDA and semiconductor markets in its growth with the IP industry expected to reach a respectable $2.5B by 2010.