There was, not surprisingly, significant overlap in the commentary that emerged from the various panels, but one set of statistics presented at the IP business impact panel was of particular interest. VSI Alliance President Kathy Werner moderated that particular session, and led off with a slide that referenced an EE Times survey of customer attitudes about IP.
Per Werner's slide:
Reasons People Use IP:
The point was clear. People want to use IP, but continue to have trouble trusting IP. This was not news to anyone in attendance at DesignCon, even if the slide did provide a more quantitative set of data points.
The idea that people want IP, but are ambivalent with regards to quality, would also not have been news to participants in a roundtable discussion I moderated on Tuesday, January 16th. The topic of that discussion was IP as well, in particular: Is IP Really that Bad?
Warren Savage, President & CEO at IPextreme, IP licensing facilitator
Joel Silverman, Vice President of Marketing at Kawasaki, IP provider & consumer
Graham Allan, Director of Marketing at MOSAID, an IP provider
Ian Mackintosh, President & Chairman of OCP-IP, a standards body
In reviewing my notes from the roundtable, I was struck by how closely the comments of panelists (see below) mirror the general sensibilities of the many panelists that comprised the IP conversations at DesignCon – with one caveat.
Star IP providers MIPS and ARM each appeared on (separate) IP panels at DesignCon. Representatives of both companies were adamant that IP quality is not an impediment to market growth. In response to both MIPS and ARM, at their respective panels, the comment was then made by other participants that microprocessors, with all due respect, are not as tough a nut to crack as other categories of silicon IP.
Consider that an informative and/or whimsical segue to the following conversation.
Roundtable Discussion – January 16, 2007
Peggy Aycinena – Can you address the recent 'controversy' over the EDA Consortium's inclusion of IP revenue with EDA revenue? I'm not sure I see why this is controversial. What do you all think?
Warren Savage – From my standpoint, I would be one of the detractors, [arguing] that IP is not an extension to the EDA business model. It's been proven that EDA and IP are separate. IPextreme is a technology licensing company working with the large IDMs, helping them to monitize their internal IP to the global marketplace. We serve as a licensing agent, supporting the agents and outsourcing the IP function for vendors. I'm quite bullish on IP on the premise that IP is the fastest-growing aspect of the semiconductor industry, outpacing growth in semiconductors themselves, and EDA.
Joel Silverman – I also disagree with including IP revenue with EDA revenue. From Kawasaki's standpoint, we're a pure-play ASIC [house]. IP is critical [to us]. It was an early decision for us to develop IP internally or purchase IP from an external [source]. The only stuff we do internally today is key IP related to areas we already have expertise in, like SerDes [serializer/deserializer]. In many other areas where we don't have the expertise [we purchase IP].
Graham Allan – I would agree with Joel and Warren that the IP/EDA combo camp [is inappropriate]. That was more appropriate when IP was in its infancy, but now I think there's much more IP being sold. One of the reasons that the IP market is attractive is its high growth rate. Interestingly, you can have an end market that's in decline with respect to growth – for instance, ASIC starts are no longer as numerous as they used to be, and the number of semiconductor companies is declining – however, there are still a tremendous number of companies who need IP. One of our IP products has over 175 users. [There are] many companies out there that have 100, 200, or [even] 500 people who cannot develop the internal expertise – for instance, in SerDes or processors cores – and that's a tremendous opportunity for the IP providers.
Ian Mackintosh – I also believe IP revenue should be [kept] separate from EDA revenue. I represent OCP/IP, which is a standards organization [working towards the] sharing of IP. I also believe IP revenue should be kept separate from EDA revenue. I think there really is a future for IP. Today, it's a successful and growing market, [plus] there are variations on a theme of the IP business model that are all being used successfully. Although the general quality of IP is not good across the board – it surprised me [to learn from a recent poll], for instance, just how negative users and buyers are – this [situation] can be helped considerably with the development of standards.
Warren Savage – There's something that can be said for Ian's comment about standards. We're involved with a number of standards – Bluetooth, FlexRay [automotive network communications protocol] – that have standards bodies that not only control specifications, but also facilitate testing labs that IP can be run through and certified. You can't use the Bluetooth symbol without a stamp, and the same [is true] with USB where there are certification laboratories that must be used. The requirements for standardization is getting pushed back [to the IP providers] from the semiconductor companies. The [users] have to insist that the IP providers provide IP with a [certification] stamp.
Peggy Aycinena – So there is a way to certify IP? This is a topic of great interest.
Joel Silverman – IP certification is not very simple any more. It used to be simple, but now with big, complex boxes the whole process has gotten much more sensitive. At Kawasaki, we build up all the blocks. Everything we license, we build into a test chip. If there's analog, especially, we send it through internal testing. If it's a USB or PCI Express, we take them to compliance workshops to verify them. Some of the blocks have other standards, as well. For instance, there might be USB on one side and ULPI [interface standard for high-speed USB 2.0 IP systems], or some other macro-cell interface on the other side. We also build those up [for testing]. By the time we're through, even if we've verified the IP, we still build a test chip.
Graham Allan – I want to echo the importance of chip verification. We verify on a test chip, but the critical issue is that it doesn't end there. It's analogous to the chip vendors themselves. Chip vendors [today] can't just send out a chip with applications notes. Now the chip gets integrated [in to a system]. Documentation and implementation guidance all require notes [that ship with the chip]. That's the difference between a printed circuit board and a chip. Our approach is to test it in test silicon, to create a comprehensive test report, and then to accompany that with very detailed information for how to integrate [the IP]. It is the customer's responsibility to work with [the IP vendor] in that regard. We find that we have areas where a customer might find a problem and then gets in touch with the vendor. IP vendors are very responsive. If a customer has a question, it will be answered in hours, if not less. It's important for the customers to ask [questions]. [Of course], the integrators must be sure to follow the guidelines [provided by the vendor].
Ian Mackintosh – IP today is a lot more complex that it used to be. We have the situation where a lot of IP is really hard to verify, particularly given the growth of soft IP. Previously, problems might have been weeded out, but the bottom line is that a lot of the IP today has a new class of problem, and it's not just in the verification. Today there's a huge inconsistency between the message from the providers and the reality. Unfortunately, "the IP just doesn't work" is often the report from the users. There's just not a strong enough following of design standards. It's true, the VSIA is producing views that can be better standardized, but what people are really dealing with is inconsistency. Using [vendor] support is a good idea because it helps users better integrate the IP. But that [promotes] a more strategic hold from the provider on the user. The fact remains that IP is not sufficiently standardized.
Peggy Aycinena – Should there be an independent, third-party entity that is certifying IP?
Joel Silverman – We're in the middle. We're selling our IP to customers, and we have in-depth knowledge of how our IP was designed and built. We can close the loop in that ongoing process. But when we license IP from somebody else, we have to get more involved in the design itself. It's usually not just – hand off the package and you're done. We have to have just as many meeting with the external providers of our IP as we do with our internal groups. They all have to follow our rules, and we push [those requirements] to the providers to assure the quality. When things gets very complex, it takes just one mistake for it to become very expensive to figure out what went wrong.
Warren Savage – I think Ian touches on a good point. You do see in some IP providers that the support is sort of a crutch for the IP, especially IP that is poorly documented and doesn't integrate well. The value of IP is inversely proportional to the time it takes to integrate it, but customers want to just plug in IP and use it as soon as possible. Everybody is moving in that direction as standards are starting to dictate [quality]. As far as having an independent center for UIP [Universal IP] testing, that's not thinking about IP in the proper way. IP is not a 2-letter thing. It's very complex and the verification technology [involved] has to be well understood.
Graham Allan – I can't imagine a way that any one particular block of IP can be validated in every possible way. There are so many different flows. You just need to get as many miles as possible under the belt [with a particular provider]. You [the user] need to shop for a provider who is well known and who passes the "large support organization" test, a provider who has sold IP to lots of customers. If there is a standard [for the IP], even that is not enough. You're going to come across problems [nonetheless]. As an integrator, [you have to accept] that the IP was invented somewhere else, in somebody else's shop. They used a different flow, and you're going to have to make compromises to understand it. You need to take responsibility.
Peggy Aycinena – Is it, therefore, Buyer Beware?
Graham Allan – It's Buyer be Informed, not Buyer Beware, and for good reason if the IP was not developed by your internal engine. Every chip, every flow, every organization is different. For an IP provider to address all of that would be prohibitively expensive. [But], IP providers want to be successful. They want to develop products and sell them many times. They need to know that they are providing IP plus design services, and they have to support specific [teams of integrators].
Peggy Aycinena – How much of an IP provider's revenue should come from the product and how much from the design services portion of the business?
Graham Allan – 15-to-20 percent should come from the services an IP provider offers for integration or small amounts of customization. However, each customer has unique requirements, and an IP provider [must adapt].
Joel Silverman – I agree with Graham, but it's hard for us. We're the middle person between the IP and the customer. Customers come to us and say they want to license IP from a third party, but we don't want our customers to go to a third party because they're doing their design through us. So we try to partition the IP we've got in such a way, so that we don't have to go back to the providers. Or we have an agreement with the provider up front and expect [them to provide] modifications up front. But for us, and IP providers, [it varies with each project].
Ian Mackintosh – That's the very problem you're alluding to here. It's all because of a lack of standardization. I agree you can't certify every design flow, but you can center around 1 or 2 main flows. But the packaging alone isn't consistent. Now I see very large consumers of IP are actually changing their behaviors. They're paring down sharply the number of IP providers [they working with] and are developing intimate relationships [with just a few]. If the consumers don't do that, they can't rely on an internal group. It's the lack of standardization [that's inhibiting] plug-and-play forms of IP, even with the best IP providers. I was literally shocked recently to learn how negative customers of IP are today. It's universal, [and it's causing] the buyers to change their behavior.
Warren Savage – What I'm seeing in the customer base is not a lot of requests for customization to build on SIP. The [consumers are] loath to have anyone touch the IP. Instead there are changes, as Ian said, where the customers themselves are starting to change their behavior. They're focusing on the integration exercise as the IP is [increasingly] silicon proven. Anything customized is outside [the ideal of using third-party IP].
Graham Allan – I certainly see that for IP, it's getting a bit easy to define things quite rigidly. Some of the products we develop – fractional PLL, for instance – we do in silicon to prove it and pretty much don't touch it after that. Our DDR IP, however, is very different because a lot of customers have specific ways they use external RAM. We also have a lot of requests for specifications that aren't standard. There are a lot of niche memories that are like standard memories, but with a few more buzzers and bells on them. I agree, however, that trying to keep it as close as possible to the original [design] is vital.
Warren Savage – I think maybe IP is interpreted too broadly. The DDR is an example that is not really reusable IP. It's a reused design, versus thinking of IP as a virtual component.
Joel Silverman – So where do you draw the line between IP and a design services block? Internally, for instance, a lot of our blocks have OCP-IP, but externally they may not be so standard. In between, there may be functions that the customer wants to change. If it's a USB, we may want to support different end points or transfers, or might want to do different types of packages. Even as much as we may try to make it standard, we may change it from 8 to 16 bits so the users can use it specifically to their application or use model. [When these changes are needed], it's not usually done by the customer.
Peggy Aycinena – So I'm hearing a variety of things are going on, and the situation is still somewhat fluid. Where then are we in the IP industry, if 10 is great and 1 is awful?
Warren Savage – The safe answer is that we're somewhere in the middle, at about a 6.
Joel Silverman – I would say we're at around a 6 or a 7.
Graham Allan – The IP industry, in general, is at about a 7, and we're there because a lot of IP is pretty well defined before you buy it. On the other side, in the soft IP area, customers may still want other customers [to use the IP before they do].
Ian Mackintosh – Based on the people I've been talking to – the big customers – we're at around a 3. The truth is that [we only have] customized IP today, and there's still a lot of handholding that has to go with it. We're not anywhere close to plug-and-play. The very best IP being used today, the very best quality available, is not considered quality by the customers. It's just not of a high enough quality.
Peggy Aycinena – Are the standards bodies helping to change that?
Warren Savage – I don’t have anything wildly popular to say about any of the standards bodies. We're still searching for the recipe as an IP industry that will make sense. VSIA is making some noble efforts pushing [for quality], but I haven's seen [the VSIA QIP Metric] adopted as widely as it could be.
Joel Silverman – I'm actually attending a standards meeting this week. It's important to be involved with standards as they develop. There are a lot of standards, but they're for [internal use]. OCP-IP is close, but there really isn't a strong following yet. So when we get IP, it could be AMBA, it could have a general-purpose parallel interface, or any number of things. We often wind up making it OCP-IP [compliant], but out in the world not everyone is following that. Every silicon provider is making their own IP and using their own interfaces. Without a standards body [that everyone accepts], we won't ever get a shrink-wrap approach to IP.
Graham Allan – I'm very much aligned with Joel on this. [However], we would struggle to deal with standards that would certify IP. Standardizing the interface is one thing, but [standards won't support] how it's integrated into the larger design. To define standard for IP blocks, you'd have to have a lot of standards for all of the small markets. But if you were to subdivide the industry into difference niches, the work would exceed the logistics of working in each market. I agree that standard may make a lot of sense, but we must recognize that there are a lot of unique instances where the IP vendor gets to a certain question [in the process] which cannot be answered by the standard.
Ian Mackintosh – I think Graham is right. We could use some generic IP metrics. There are generic IP metrics today from VSIA, and some supported by FSA, where people are providing a consistent set of views and data for functions. Similarly, OCP is providing a valuable service. We've got the largest SOC designs today which are OCP-based, and have moved a long way forward there, but I'm not talking about things [like integration] being done in a certain way. I'm talking about generic IP metrics. I'm extremely optimistic about the IP industry. There's a helluva opportunity today for IP providers. They've got a natural progression path. They can start out with an idea for a product, and then can drive to value ad or provide services to support an application, or embedded software, or systems. There is loads of opportunity for SIP. I just have to take a dour view of the quality.
Peggy Aycinena – Ian, you have referenced a poll of some sort that you say reflects negative attitudes from some IP consumers. Is that poll available publicly?
Ian Mackintosh – There are two of them, but [they are not available publicly]. We actually did a mini-detailed survey at an IDM with some fabless guys, basically all types of huge consumers. They were all just very negative. I was expecting that. But [as I said], as a result they are changing their behavior about what they buy and how they buy it. If you're an IP provider, this is actually good news.
Joel Silverman – I'm following what Ian is saying and I agree. Whenever I go to buy or license IP, I can always find many providers. But I usually don't pick the cheapest. Instead, I go through a systematic selection process. If we always used the cheapest or the provider with the quickest delivery, [we would end up] with a number of providers who are not very good. And, we sometimes advise customers to not use an IP provider just because they're the cheapest. [Again], it's really up to the users to qualify the IP. A customer may tell us they want to use this [specific] IP from that company, and maybe we'll have to respond, but we won't necessarily take responsibility for that decision.
Graham Allan – I tend to agree that a company's reputation is exceedingly important. Anyone can be an IP provider, and with the semi market getting so expensive – bringing just one design to market is so costly – even a small company with under 10 people can participate. But I believe it is Buyer Beware. You can't just look at the IP. You have to look at the reputation of company you're dealing with, at the length of time they've been in market, at how long they've been in the design market. For someone who is just getting into market, you have to look at their customer based, who have they provided product to, and if the customer who is providing public endorsement [for the product] actually has enough experience to know [what they're talking about].
Ian Mackintosh – I should point out that in the poll I referenced, it was a large IDM that said, "You have to know the kind of complexity you're dealing with. Just because you've got a good IP provider, it might still be junk they're providing to you. Just because you're getting IP that's from a world class provider, it doesn’t mean that the rest of the stuff they're providing is good."
Peggy Aycinena – So I'm hearing that the users are getting more savvy?
Ian Mackintosh – It's a fact that they're taking more services from the IP providers to get things integrated. That's a good thing if you're a provider because it's like providing heroin. But if the IP users are not working efficiently as a result [in their integration effort], then they're resorting to behavior that's not good. And none of this will drive the IP market to expand.
Peggy Aycinena – So it's not a good idea to plan on a long-term relationship with an IP provider because it's like an addiction?
Ian Mackintosh – Users are not doing it because it's a good thing to do. It's actually a bad thing. If you're a provider, and the quality were there in your IP, from the user's side [they wouldn't need to rely on a long-term relationship].
Graham Allan – I actually don't think it's great from the seller's side either. I don't want any customers expecting poor IP from me so that we're always saying, "We'll fix any problem you have going forward." Verifying IP is very complex these days if you have to imagine every switch that your customer can throw at it – particularly as there are some very imaginative users out there. As a provider, you do the best you can, but if there's a problem, the responsiveness of the provider and the time it takes for the customer to get back up and running [are all on the line]. You could fill out a scorecard [on the IP or the provider], but that doesn't tell the next user what their experience will be.
Ian Mackintosh – I agree that rating IP on a score card isn't material, but the issue is that there's a scorecard at all – even if every line item isn't a known positive or negative.
Joel Silverman – I understand what Ian is saying and yes, even within a specific provider, there is some good and some bad. We pick an IP provider for specific blocks, but I don't think that dismisses us from doing due diligence on those blocks. For good providers, the block will come and it will work, but we still have to get an understanding of how that block is working. We have to be able to figure out, if there is a problem, is it the block, is it the software, or did the [integrators] lay out the board wrong? So, it's not just does the block work – it's does the chip, the software, and the board work, and have they wired it all properly? Problems are not also just a specific thing related to just the block itself, they can be related to the whole system.
Peggy Aycinena – So using IP is still a complex process and/or proposition?
Warren Savage – I'm absorbing all of this discussion, and maybe I'm just a little more positive, but I'm not seeing the negativity that I saw 5 years ago in the IP industry. It's my experience that a lot of the bad IP players went out of business, and thankfully so. [Today], if you look at the IP market – and we did $1.6 billion last year – if you look at the revenue numbers, it's growing, by and large from the high-quality IP providers. It's the smaller IP companies which are at the root of these quality issues, and people always tend to remember a negative experience a lot longer than a positive experience.
Peggy Aycinena – Is the industry reaching a point where there is no more room for small players?
Joel Silverman – No, we use a lot of small guys. Are they contributing to the specifications, to the IP standards bodies? We've worked with small 5-to-10 person companies [and found that] a lot of times, these providers have the same if not better capabilities than the larger companies. As long as they are providing quality IP, we will continue to work with them.
Warren Savage – I would say that in certain vertical domains, there's still room for small guys like that. But becoming a larger IP company is a challenge today. We're [working] in a different era where larger IP providers [are becoming prominent].
Ian Mackintosh – There's still loads of room for small guys to come into the industry. It all depends on the [level] of innovation. Is there specific support for a series? Is there embedded software technology. They'll need all of that to succeed. And in some markets, even innovation is enough, no matter how raw what they're providing is. Also, it helps to have a lot of specific knowledge of the consumers to get going. Certainly, an IP provider must understand that.
Graham Allan – Again, if you're going to be a small IP providers, pick one market and one product. Once you've had success there, and have lots of customer engagements under your belt, [then you can expand]. Certainly there are several very reputable IP providers with fewer than 15 or 20 employees, companies that have been very successful in small market niches. And certainly, there is still innovation going on in IP with several vendors trying to change the world with their products. But some have bigger mountains to move than other. IP is driving innovation, but innovation is not just building what other people want.
Peggy Aycinena – By the way, did any of you go to the recent IP-SOC conference in Grenoble hosted by the Design & Reuse Center? Is that a unique destination for the IP industry?
Warren Savage – It is similar to other shows, although the conference this past December had quite a large emphasis on network-on-chip.
Graham Allan – It's an [important] conference, but I'd like to see a similar type of show [perhaps] in a different location.
Ian Mackintosh – I did go to the show, and this time here was something different there. They had 15-minute status updates that were useful with lots of discussion. [It was] very productive for the attendees. There were lots of senior guys there. Yes, it would be nice to see something analogous to the show here in North America, if not in several other destinations as well.
Peggy Aycinena – Does the location of the IP-SOC conference suggest that the bulk of the IP providers are in Europe?
Joel Silverman – It's not all in Europe. There are several providers in India and several in the U.S. The IP industry is pretty much worldwide, perhaps with a slightly higher percentage in Europe, but not a significantly larger number.
Graham Allan – IP providers are global, and the users are as well.
Warren Savage – There are IP providers everywhere, although I'm seeing more sophistication in Europe compared with the rest of the world. They're slightly ahead of the rest of the world with regards to integration issues, system-level modeling, and specification issues.
Ian Mackintosh – I would agree with Warren, for the immediate sophistication of integration and specification. But in terms of proliferation of IP companies, there continue to be companies coming out of India. Also in India, we're seeing companies focused on adding value-add elements, building on the basis of lots of design-services work already captured there.
Peggy Aycinena – Would you each like to make a closing comment?
Warren Savage – I think that the market for IP companies is going to do nothing but grow. The IP business is in its infancy stage. It's our feeling that IP is going to be much more loosely associated with the semiconductor companies because of the nature of the industry and the technology. The IP industry is more than EDA – it is becoming a market in its own right.
Joel Silverman – Addressing the original question, is IP that bad? No, it's not that bad, but we have a ways to go. The IP industry is going through the same transition we're all going through. It's becoming better and more reusable, and it doesn’t have to be redesigned for every application. We're going with the idea that standardized interfaces [are important] on everything we do, and that's making it much easier for customers to use IP that's strictly plug-and-play.
Graham Allan – I know that IP has a bad reputation historically, just as any product has that's intellectual by its nature. You can't taste, touch or hold it, but you can do generic metrics. And some IP blocks are easier to design those metrics for. But, at the end of the day, the quality is improving, the market is consolidating, and the large IP vendors are looking to acquire the smaller IP vendors. If any IP provider has standards in mind, they must provide [help in pushing that process forward].
Ian Mackintosh – I think growth for the silicon IP industry looks good. The quality is not yet as good as we would like, but that's a sign of potential for growth. Standardization is the path to fueling growth in the industry and anybody who can leverage that growth will do really well going forward.
Peggy Aycinena – So I'm hearing you all say that progress may be a bit slow, but it's steady nonetheless. Providers and consumers may wish it was all happy news, but despite the concerns, there's certainly plenty of good news.
Editor's Note: I had a chance to speak with D&R's Gabriele Saucier at DesignCon in Santa Clara. Mme Saucier told me that she had been asked by numerous people at the conference to explore the idea of having an IP conference in other venues in addition to Grenoble. She said that the idea was interesting and that if such a conference were to happen, it would probably take place somewhere in the central region of the U.S. There's not enough involvement in the IP industry on the East Coast, according to Mme. Saucier, and it's too crowded on the West Coast. She postulated that Austin might be a good choice, although the decision is far from finalized. Mme Saucier said there is also interest in having an IP conference in India and/or Asia as well.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Peggy Aycinena, EDACafe.com Contributing Editor.