IP in the Modern Age
******************************* Remind me. How do we feel about IP? * Is it the technical and business holy grail that's propelling us into the next great epoch of design and manufacturing? * Is it the gritty solution to the globally distributed design team? * Is it the inspired adhesive cleaving vertically integrated industry partners one to another? * Is it synonymous with system-level design? * Is it a mindset, a state of grace, a natural answer to the soulful yearnings that reverberate along the entire semiconductor design and supply chain? I don't know, do you? I do know that as wacky as some of these claims may seem, they're actually pretty dead-on to some of the stuff we've all been hearing over the last few years when the discussion comes around to semiconductor intellectual property. So presuming you can put aside whatever you've got going on today or tomorrow – including and especially all those details, deadlines and deliverables related to DAC 2005 – please step up, sit down, and read this virtual Panel Discussion from beginning to end. It'll take less time than sitting through an actual panel in Anaheim. When you're done you'll have a better answer to that eternal question: How do we feel about IP? ******************************* The Participants Mahendra Jain – President at QualCore Logic ******************************* The Panel Discussion ******************************* Question 1 ** Where are we today in the IP Industry? How many companies? What's the dollar size of the market? Are there just a few Star IP providers and then everybody else?
Hal Barbour – D&R [The Design & Reuse Center] lists about 350 companies and The VCX about 250 in their respective IP catalogs. As for CAST, we have performed at high double-digit growth rates for the past three years with triple digit growth in the 1st quarter of this year. Even though the market is greater than $1 billion, it is still embryonic and fragmented. There are multiple loosely related segments including Star IP, Standards based IP, Mixed Signal IP, and hard IP. Further there are many different kinds of suppliers, from solely IP companies (like CAST and ARM), EDA companies (like Synopsys & Mentor), Design Services companies, FPGA companies, Semiconductor foundries and freeware. All have legitimate business reasons for providing IP and most do not yet compete directly with one another for new IP business. Warren Savage – Microprocessors, memory, and cell library companies whose products feature standard functions with standard interfaces are doing well. Next are IP companies that provide higher level IP products that are built on lower level IP components and often include software as part of the product. These companies are seeing rapid growth, but also suffer from higher costs because of the design complexity of the products they are offering. Last, the IP companies that offer IP that is based on openly available standards are struggling to survive because of commoditization forces that drive the prices through the floor, and limit their ability to invest in creating a quality product. Jon Atwood – When you look into the stats, I think what you will find is a very fragmented IP industry. There are a few big players at the top (Synopsys, ARM, Rambus), then there is everyone else. Everyone else consists of a lot of firms, mostly small and private and very niche focused. The big trend shaping the IP industry is consolidation. SNPS has been on an IP company buying binge of late. ARM bought Artisan. We’ll see more of this as the industry matures. Phil Casini – The IP market for SoCs is thriving due to: 1) The continuing exponential increases in demand for processing being placed on SoC as convergence accelerates through consumer markets and, 2) The increasing rate of rapid ASP price erosion in consumer markets which force systems to contain as few chips as possible. ******************************* Question 2 ** What's the current state of mind for the IP users? Is NIH (not invented here) still prevalent? Is there a bunch of shrink-wrap IP out there to be purchased off the shelf? Are IP vendors needing to tweak their product-versus-services mix on a daily, weekly, monthly, annual basis?
Analog means that everything is customized and it’s done via a contract. When we license our IP, there are a lot of requirements that we’re required to meet and it differs product by product. Analog IP means that it is a lot more service intensive than digital. Hal Barbour – It is changing. Most ASIC designers are now quite knowledgeable in the use of and expectations of commercial IP. Still, many IP sales for ASICs require some optimization for the particular requirement. Only high-end FPGA applications (such as image compression) require similar optimization. We find that FPGA users are far less experienced in using IP than their ASIC counterparts. Warren Savage – NIH is long dead. Using IP is a practical reality with users. Shrink-wrap IP is in fact, the predominant form of IP out there: Microprocessors, memories, and cell libraries are all forms of shrink-wrapped IP. It is no coincidence that companies in these spaces are succeeding-- that’s what customers want and IP companies need. Engineers don’t want to tweak the IP but are often forced to because is often not architected from a user’s point of view. My rule of thumb is that the value of an IP is inversely proportional to the amount of time the customer spends integrating it. Thus, IP that needs tweaking offers a diminishing return for the customer and works against the IP company by turning their licensing business into a margin-based services business. Jon Atwood – The current state of mind among IP users is that it is becoming safe to come out and use IP again. Many of the early IP users were burned by poor quality IP delivered by inexperienced IP providers. NIH is going away quickly. Most fabless semi companies are on the same page that in order to compete, IP licensing and integration is something they need to do. When people refer to shrink-wrap IP, they are usually referring to the lower complexity soft IP that is available from providers like SNPS and Mentor. All IP vendors have some mix of design services and IP. Almost always, the design services are in support of customizing the IP to meet specific customer requirements. In the industry today, you can see evidence of IP providers moving closer to a design services business (GDA). And design services business moving more to an IP model. Phil Casini – Sonics can only speak to the dynamics of the intelligent internal interconnect IP market. This market has determined that the economics have shifted towards outsourcing interconnect design as evidenced by Sonics’ growing business and tier one customers. The reason for this is that nearly all complex SoCs require multiprocessing. Since all processors are not alike the data flow requirements force developers to engineer a group of intelligent data flow services, in addition to the buses they need, which are very complex tasks. Buying the solution is more economical. ******************************* Question 3 ** Is selling IP still a 'missionary' sell?
Digital designers are buying analog IP. Analog is difficult for digital designers to understand because it’s an entirely different discipline. At QualCore Logic, we’re using our analog expertise in order to close the business and explain how the analog IP will fit in a customer’s environment. Hal Barbour – No. This has changed over the past few years. Engineers are now proactively seeking out IP solutions (usually through web searches) as alternatives to either developing their own or using standard commercial parts. Warren Savage – Only if nobody wants it. Jon Atwood – Glad you asked… IP selling is a unique process. It is different than EDA, and Semiconductor selling. Interestingly, it very similar to a design services sales process. There is a statement of requirements, customization, deliverables, etc. Semiconductor reps have traditionally been looked upon by private IP providers as their sales channel with little success. The IP sale is far removed from the typical semiconductor sale in many ways. It requires a unique understanding of the technical and business aspects of a customer’s decision. This unique approach to IP selling is idea that sparked my business plan in 2001 for starting The LogicWorks. Phil Casini – No. Semico Research has published charts showing the clear economic benefits of using outside IP strategies in terms of impact on semiconductor chip development costs. Tier one company's ALL employ some level of third-party IP acquisition strategies today. ******************************* Question 4 ** Is SoC complexity and/or shrinking device size having a positive impact on the viability of the IP market?
Hal Barbour – Definitely so. That is the primary enabler of the high growth rates that are still being forecast for the IP market. Warren Savage – Design starts are going down, but design complexity is going up. Shrinking geometries sets up the demand to fill those larger chips with design content. I like to think of IP companies as SoC content providers. Imagine if your house doubled in size every 18 months, you’d be spending a lot of time at the furniture store. Note that you buy furniture, not build it. Jon Atwood – This is a really good question Peggy. It was the subject of my last month’s newsletter. Complexity is a huge driver of the industry in many ways. As geometries shrink toward 65 nanometer, the risk of blowing a design are huge. What this means is that IP providers become more important to both the end customers and the foundries themselves. This will also drive the consolidation in the IP industry. Only the strong will survive. Complexity also means that design teams can less afford to do it all. Design teams are becoming IP integration teams. They will design what they consider their "special sauce" and license everything else. The EDA industry has not caught up with this trend. Phil Casini – Again, speaking only for Sonics and the intelligent internal interconnect market- Both complexity and Moore’s Law are having a positive effect on the market. Complexities are growing rapidly as an effect of convergence, and Moore’s Law continues to provide the manufacturing economics required to deliver the complexity. The engineering challenges have shifted toward design and methodology, which fuel the need to leverage outside IP from third parties. ******************************* Question 5 ** Is the "leveraging" of the distributed design team have a positive impact on the viability of the IP market?
Hal Barbour – Yes. Distributed design teams are increasingly comfortable with superb web and e-mail based support and less demanding of expensive face-to-face interaction with the supplier. (This evolution is one of the reasons why CAST's model has worked so well!!) Warren Savage – There has to be a legitimate value in it. The advantage of off-shoring design has yet to show a pay-off. Design is different than manufacturing. IP design is really the highest form of design that involves taking into consideration the requirements of a large diverse set of users. Compounding the issue— design is not a static activity, the skills needed today are not going to be sufficient tomorrow. So the challenge with distributed teams is keeping the skill sets in sync and growing. It’s a serious organization learning issue that is not well appreciated (yet). Jon Atwood – Hmmm. If by distributed you are referring to an offshore model, I’m not sure that this is a real driver to the IP industry. It could be in the sense that distributed teams need commonality of IP across teams, but here are other factors that are putting the winds in the sails of the IP industry. Phil Casini – Again speaking only for Sonics and the intelligent internal interconnect market – yes this too has a very positive effect. This is due to fact that as SoCs become more complex, the architectural exploration has shifted from the data processing elements (like microprocessors, DSPs, or multimedia engines) to the data flows. Data flow is a critical path to achieving performance in a multiprocessor SoC. Sonics has benefited greatly from this shift by offering solutions that not only meet the customer needs in terms of IP deliverables, but tools that enable them to do more architectural exploration than they have ever been able to do before. For distributed engineering teams, setting the framework for the chip, via the interconnect, and then connecting the IP is the only way to decouple the work and have it efficiently come together as a chip. The shift to a platform style strategy then is essential to the success of the business unit. ******************************* Question 6 ** Who creates the IP that customers use? In-house guys? Other business units? Third-party providers?
Hal Barbour – It's a combination of all three. There was a trend in larger companies to have their own central IP groups develop and manage IP corporate-wide. But we see less and less reliance on these centralized in-house IP groups, and more emphasis on third-party providers and in-house project-based IP development. Warren Savage – It’s a mixture, and will probably stay that way for the foreseeable future. Different companies continue to focus on their core competencies therefore IP will have many sources as the chips get bigger. Jon Atwood – It’s changing. This was once the sole domain of internal IP development groups in top-tier fabless companies. These groups still do some internal IP development, but that activity is decreasing. The growth of the commercial IP industry has taken up the slack and met the increasing demand by IP consumers. My experience is that these internal IP development groups are now more focused on providing value by acting as a "corporate IP clearing house". They will assist internal business units by evaluating external IP providers and entering into standard business terms. They may still do some internal IP development, but this is typically only done today where there is no commercial alternative available. Phil Casini – All of the above. The benefit of OCP as an open standard, which Sonics supports, is that it does not matter who creates processing elements. They all connect together in a de-coupled way using a protocol based scheme. The IP elements can be anything. ******************************* Question 7 ** Is managing the IP incorporated into a design a "differentiating" factor in the success of a design team/company/product?
Hal Barbour – The critical factor -- now more than ever -- is time to market. If managing your IP effectively helps reduce this, than it is indeed a major differentiator. Whether you're designing with a star IP processor or a USB port core, you have to beat your competition to market to win. This is probably one of the biggest implications of Jim Tulley's predictions about the consumer starting to drive the whole electronics market of the future: if you miss getting out there in the consumer space, you just lose the whole market. Would you want to be the last company to produce a new cell phone with a camera? Warren Savage – Some are saying that. Personally, I think it’s a bit overstated. We are transitioning from crawling to walking when it comes to using IP. Eventually we will get to running, where secondary factors like managing the IP will become a differentiator, but we’re not there yet. Jon Atwood – Is it ever! Key competence for a design team is evolving quickly from creating a lot of RTL from scratch to integrating complex blocks of IP into a successful design. Some fabless companies have moved rapidly along this path. Others are progressing more slowly. EDA companies have not kept up with the trend. I made a presentation while at Synopsys in 2001 that was based on this the theme that IP integration was something that we at Synopsys needed to be ready to help our customers be successful at. Phil Casini – In the case of intelligent internal interconnects, I would say yes. Autoscoring the tasks and leveraging a Sonics solution has proven benefits with respect to higher design productivity and increased engineering efficiencies. Sonics supplies solutions to the leaders in the target markets, who maintain their lead in part by adopting Sonics methodology. ******************************* Question 8 ** How does a potential IP user find out what's out there? Is IP "mining" within a large organization an effective process these days? If somebody's looking to third-party IP providers, how do they find out what's out there?
Hal Barbour – Our inquiries come from:
Warren Savage – "Google" seems to be pretty popular answer among people we ask. Beyond that they depend on reputation and looking in the online IP listings such as Design & Reuse, VCX and IP Trading Center in Japan. Jon Atwood – The top tier fabless companies typically have a central team that assists and coordinates IP searches. Smaller fabless folks have to go it alone. Usually an IP search will start with a search engine lookup like the Design and Re-Use website. Design and Re-use has created a very useful repository for information on IP providers. But this only scratches the surface. Today, IP consumers want to know about interoperability among IP providers. Often a customer will find that the best hard macro PHY provider will not be the same as the soft digital core provider. Customers want to know how these players can work together. This has been the founding principal of The LogicWorks, meaning that IP consumers need a hands-on approach to helping sort out IP alternatives. A search engine just can’t offer this. Phil Casini – IP mining within an organization has problems sometimes in that the business reasons to cross use are not necessarily the best for any one division. In the case of processor subsystems, there seems to be more sharing because software defines how the subsystem is used. In the case of internal interconnect, Sonics offers flexible solutions that can be shared in a very straightforward manner. Market research firms all have focused coverage on IP today. So contacting them and buying their reports are a straightforward way of understanding what is available in the general market place. OCP has its own web site and posts its member information. ******************************* Question 9 ** How does a potential IP user verify the reliability of the block? (I think this is the over-arching question in the entire business/technical discussion about IP.)
The IP vendor can only test for certain conditions. It does not go through the same silicon validation as a silicon chip goes through because that takes about 24 months. The IP vendor usually relies on a customer to provide further product qualification information. Design teams need to understand that not all conditions can be tested, especially analog IP where they want to make modifications that aren’t silicon proven. For example, if a design team is targeting the design for 90nm, proving the IP works at 90nm is not possible at the same time the design is being implemented. That’s because the IP vendor will never be ahead of the process technology. The IP vendor gets access to the technology at the same time as the customer. Sometimes, it’s even later than the customer. Design teams want high-quality, silicon-proven products, but don’t understand the high cost of getting silicon-proven IP and the investment required to make the IP silicon proven. Hal Barbour – The more complex the design, the more effort our customers put into higher level verification (you just can't afford re-spins at 90 nanometers). They use tools from all sorts of suppliers to do specific tests on timing, buses etc., including verification IP for cores such as MAC, USB and PCIe. We also find a lot of our ASIC customers doing FPGA prototypes to verify the entire design, not just the IP. Most assume that if there is a problem with the IP, it will be found through simulation and FPGA prototyping. (And that's one form of licensing we offer, FPGA prototype followed by ASIC production.) Warren Savage – The answer is an old one: number of uses, number of units in production that use it. New IP is still risky as hell, but ultimately the question comes down to the design team that created it. EDA companies would like us to believe that new verification technology can magically fix bad designs. I think improved design methodology is the closest thing to answer we can hope for. I like to say that if you don’t want bugs in the code, then don’t type them in. I’m personally very excited about new methodologies that are built around agile, lightweight, common sense approaches to design like Extreme Programming. Jon Atwood – This point has seen a lot of attention and rightfully so. The early days of the IP industry were a "license and pray" period for IP consumers. Quality of IP back then was very variable. Barriers to entry into the IP industry were so low that seemingly everyone jumped in. Design services firms tried to dabble in IP, found that it was not as easy as they thought, and subsequently got out. In many cases, the IP consumer was the one left holding the bag. So the IP quality/verification topic is still an important one today. Quality is a driver, and complexity is as well. As a result, a new segment of the IP industry is emerging. Verification IP is a rapidly growing segment. Companies like nSys Design Systems that focus on providing Verification IP to help IP consumers prove out complex IP-based designs are well positioned for growth. Phil Casini – Sonics eases much of this because the methodology behind data flow centric SoC design is that IP evaluations can be executed in parallel with chip performance evaluations and interconnect configuration. This can all happen at the architectural exploration phases of the design, long before physical design. The emergence of System C models and ESL also helps both from a hardware and software perspective. The key element is being able to model the data flows for the complex SoC early. OCP ensures interoperability of the IP elements, which gives architects the ability to vary the different IP elements and determine which is best for them. In the past, lack of data flow modeling would render any of these attempts as not worth while. ******************************* Question 10 ** Should there be industry standards bodies, "Good Housekeeping Seal of Approval" bodies that will benchmark and/or vouch for the quality of IP? Do such organizations already exist?
Hal Barbour – There have been good efforts -- culminating perhaps with the VSIA/FSA QIP Quality IP metric -- but we don't see customers caring that much about them. And there's the SPIRIT consortium, but that's aimed more at easier IP use than the quality of the IP itself. What customers seem to care about the most is simply if a given core has been "silicon proven," i.e., already successfully implemented by some other customer in some actual product, or at least implemented in a reference design by us. Also critical is achieving certification where that's applicable: undergoing conformance testing and displaying official logos for USB cores (USB-IF), Ethernet MACs (UNH certification program), and PCI and PCI Express (PCI-SIG) have recently become much more important, again because customers are looking for real-world validation of the core. Warren Savage – I think the GHSOA [Good Housekeeping Seal of Approval] is a great example of why these things are rather meaningless in the grand scheme of things. I can’t remember ever buying anything because it had this. The benchmark/vouch idea is fundamentally what’s wrong with the IP industry today. The problem is not about getting the bugs out, it’s about not putting them in there in the first place. Customers soon remember and respect any reputations for quality that are built. Jon Atwood – Sort of. But there is a lot more work to be done. The FSA has taken on an important role here. I was recently asked to be part of a body, headed up and organized by the FSA, to look into IP practices with respect to licensing. IP License agreements vary so much from provider to provider that there really is no standard in the industry. This body will focus on developing some common best practices for IP providers to follow in their license agreements. There really is no overseeing body that offers a "AAA" approved rating to IP that I know of. There are attempts at standardization to a specification like the USB and PCI certification standards. Most consumers want to dig beyond this themselves to understand what the silicon track record of the IP is. Phil Casini – Having one body is ambitious but in theory would solve a lot of the problems. Most often architecture teams have to take on some of this role. A unified de-facto standard interconnect could help a lot (similar to the TCP/IP steps taken by the communications industry to unify low level connectivity) and I believe we are on the path to get there soon. ******************************* Question 11 ** How do you deal with the issues of patents and/or concerns about IP theft when it comes to protecting your product? Is this not a concern? How do you encourage your customers to treat your IP with great care when they purchase from you? Have you ever found one of your products has been reverse engineered and is being used by unauthorized persons?
Hal Barbour – Regarding the specific subject of patent protection, most standards-based IP is designed around an open standard so patents are not an issue. There are other proprietary functions that are invented by third parties, and royalty payment may be necessary in those cases. In the greater scheme of things, IP protection has not been a problem for us. Most companies are honest, and, provided the rules of use and protection are clear, most abide by them. While we are not aware of a single misuse, it is possible that it has occurred but even then it would probably be a case of inadvertent misuse rather than maliciousness. We believe that it would cost a lot more in the form of lost opportunities and product delays if we were to implement more restrictive procedures. A final point is that one of the most important aspects of the IP business is support. If the IP is stolen or reverse engineered, technical support would of course be a much more difficult problem and something that would quickly become painful to the users. So, in other words our (private!) position on this topic is, if it isn't broke, don't fix it! Warren Savage – Some of the techniques we use are proprietary, but low-tech, and we can’t talk about them. However, today IP protection is still very much something that is protected by contracts and lawyers, and not by technology. I’m sure there is the odd case of piracy out there, but generally speaking, I think engineers are a pretty ethical about these sorts of things. Jon Atwood – Since I am not an IP provider myself, I can only speculate based on my experiences. IP protection is of course a critically important issue to all IP providers. It’s one of the reasons why license agreement negotiations with customers can take a long time. What I have seen is that IP consumers do live by the restrictions placed on them in the IP provider’s license agreements. I’ve actually never seen a situation where there was a question of a violation. Phil Casini – Aside from all the standard legal protection, how IP is valued has a lot to do with how complex a problem the IP solves. The more complex the problem, the less likely most people will be to spend the time to reverse engineer a solution, because it does not make economical sense. ******************************* Addendum... Peggy, After researching to answer your recent IP industry questions, I thought it might be interesting to look closer at the numbers of companies listed in the respective catalogs of The VCX and Design and Reuse. A company has to actively submit IP core data to be listed here, and so inclusion in one or both of these catalogs might be some indicator of a base level of seriousness in the IP business. (Marketing-wise, it's probably the next thing to do after getting one's own web site up.) The attached shows the approximately 250 at The VCX and 350 at D&R. I didn't analyze for overlap – there's plenty – but did sort them by alphabet so one can compare at a glance. ( IP-catalogs-comparison.pdf)Though all these companies appear in a site's search-by-company pull-down menu, they don't all get identified as "partners" elsewhere on that site. I'm not sure what that means, but probably the catalogs of non-partners (who haven't paid a fee) are less complete or perhaps not kept up to date. (It is, of course, in the catalog's best interest to list as many providers as possible!) Regards, Paul Lindemann
D&R Catalog Search http://www.us.design-reuse.com/sip/
Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com
|