Gabe Moretti, Contributing Editor
The introduction of commercial synthesis tools is responsible for the creation and growth of at least two industries. The ability to synthesize an electronic design allowed engineers to produce Application Specific Integrated Circuits (ASIC) which in turn fostered the foundry industry as we know it today. Of course even before synthesis some companies that did not have an internal foundry designed semiconductor ICs and contracted with a semiconductor company to have the devices manufactured. In the early 1970s for example, Compucorp, my second employer, had its microprocessors built by American Microsystems, Texas Instruments, and General Instruments. But that was an exception, because most companies purchased commercial semiconductors. The foundry business grew in order to serve ASIC designers need for fabrication facilities for their products.
The second industry enabled by synthesis is the Intellectual Property (IP) industry. The IP business started in the late 1980′s and early 1990′s. ARM in England, HDL Systems in the US, and Synopsys were the first companies to offer IP products. The term IP was not around then, and the business of licensing modules to third parties in the form of synthesizable or at the RT level was not well understood. I remember a meeting with a VC company while at HDL Systems. At the time we already were selling a product and their question was: “So what comes next?” Our answer that we would build more such products left them quite skeptic, and of course we did not get the funding we requested.
Now, although still in need of some improvements, the IP industry is a multi-billion industry that continues to grow. Major EDA companies, Synopsys and Cadence above others, have a strong position in the IP business. ARM has grown to be a significant player that competes with the likes of Intel in the licensing of microcontrollers, and HDL Systems ended up being acquired by Philips Semiconductors for their internal architectural needs. So too bad for those venture capitalists, who shall remain nameless to protect the shortsighted.
The IP industry has contributed to growth in the EDA industry as well. Tools that helped designers integrated the 7400 class components in their designs were no longer sufficient. Design architecture and verification tools have changed to the point to be complete new breeds. Standards have grown in order to integrate the IP as well as in support of new verification techniques. The significant enthusiasm for the Internet of Things is the result of widespread availability of IP modules. With integration came standards. We are not quite there as far as standards, we must confess. But ARM, for example, has done a very good job in defining a bus standard that is being used even in non-ARM centric designs.
Accellera and the IEEE have also produced robust standards to facilitate the use of IP by third parties, especially in the area of verification.
The continuing progress in semiconductor process technology is giving new challenges to the IP vendors and integrators alike. Design rules are so stringent today and so foundry dependent that is no longer possible to create an IP that is independent of the foundry used by the licensing company. IP must be tested and approved by each foundry, together with the rules used in synthesis and Place and Route. The natural question that comes to my mind is: will foundries become IP vendors? Or even more stringent will foundries have to become IP vendors? Will Synopsys and Cadence, for example, find it in their interest to sell their IP business to one or more foundries? Will the overhead involved in qualifying their IP become so onerous for EDA vendors to put into question the rationale for being in the business, as margins shrink due to NRE growth?
Will TSMC, for example, become the early twenty-first century Texas Instruments? And if not, why not? The internet has certainly made selling easier and large licensing agreements can be negotiated with reasonable ease. Or maybe the availability of IP modules directly from a foundry will be a significant differentiator among foundries. I can see that the direct availability of significant IP modules can be a differentiator for Intel to compete with ARM, for example. One less contract to negotiate, and one less source of potential integration problems. It would certainly be less costly for Intel to become an IP supplier than for ARM to get its own internal foundry.