By Ed Sperling -- Electronic News, 8/11/2006

Electronic News sat down with Sanjay Dave, director of ASIC engineering in ATI’s digital TV group, and Kevin Walsh, director of marketing for intellectual property in Synopsys’ Solutions group, to talk about what’s working in the intellectual property space and where the problems are. What follows are excerpts of the discussion, which was held in front of a live audience at the Design Automation Conference.

Electronic News: What’s worked and what hasn’t in the IP world?

Dave: When we’ve done our due diligence and made sure the IP we get is already proven and certified with either an Internet or a USB certification, those have proven good for us. The horror stories are when we go through the verification, we’re told they’re certified, but when we get the chip back something isn’t quite right and we don’t know what it is. We have to spend a lot of our time debugging someone else’s IP and prove to them it’s their problem. We’ve spent numerous engineering hours and lost markets we’re trying to get into because there’s something wrong with our chips.

Electronic News: This is an inexact science with a lot of finger pointing. How often does Synopsys get caught in the middle of this?

Walsh: There are a lot of barriers to succeeding at this. As an IP provider, we often are called upon when there’s an issue with the IP to try to do better. In the face of a tapeout, where the customer is under enormous market pressure, we have to be prepared to dig in for the customer.

Electronic News: Is the market comfortable with IP yet?

Dave: There’s still some trepidation in the market because there’s never a guarantee. Even if it’s certified, when you put it into an ASIC you don’t know how it’s going to work with other parts that are around it. For example, there are noise issues. At ATI, we are always pushing the limits of technology, so there are issues that haven’t been uncovered yet. I think some of the best IP we’ve gotten was also tied to the service we got with it. Especially in the analog side, they’ve come in and looked at the other parts of the chip around it and said, ‘Here’s what we suggest.’ They need to feel their product is just as much at stake as ours.

Walsh: As a supplier we’ve learned that customers really require a level of transparency. We have to understand where they are in their design so we can align the appropriate resource to help them through that process. If you look at the industry and how it’s maturing, the buyers are a lot more sensitive to understanding the entire package of what they’re getting from the supplier. It’s not just the GDS (graphic data system). How does that company react under pressure to do a design?

Electronic News: We keep hearing about standard IP. Does it really exist?

Walsh: There are standards, and we build IP to standards.

Dave: If it’s standards-based and they say it’s standard, it may be. We’ve had cases where it wasn’t.

Electronic News: So the word standards should be in quotes?

Dave: Yes.

Walsh: There are variations in standards. If you’re putting a USB on a chip, if you go through that design using a certification you get a level of certification. There are some standards which are much looser, so you really have to be careful about what’s in compliance.

Electronic News: Okay, so a chip designer is looking at a slew of competing IP that’s all supposed to be standardized. How do they go about choosing the best product?

Dave: We went through this about a year ago. We put a spreadsheet together, had technical discussions with each of the vendors and went through what technologies had been certified and how best it would fit into our design. Size is important. We also check how many products are out in the market using their IP. Put all that together and then look a how we’re going to use it technically. Sometimes how the data is formatted is important. They’re all going to be the same general function. So what do they do beyond that, which is going to help us out?

Walsh: We have a checklist we go through systematically. Is it certified, what’s the performance. Ultimately you want a piece of IP that’s going to be of value to the customer. Beyond figuring out the basic functionality, you have to figure out if it has the kind of robustness that you need. Typically that is a matter of asking how many designs this IP has been used in. If there is a reference design, what kind of hardware and software verification is out there? You do need to look at whether the company has a methodology in place to build a robust piece of RTL. You do that by testing the engineering veracity of the company itself. What kind of expertise do they have if they’re laying down a particular design? Have they looked issues of cross-talk?

Electronic News: So what you’re saying is that brand is vital?

Walsh: Brand in the sense that you can deliver IP in a consistent fashion. You can get the brand satisfaction from the same vendor again and again.

Dave: Brand is important. We will consider other factors, because cost is always a factor. Sometimes the bigger players want to charge more. But we also consider whether they’re a credible source and whether they’ll give us the support that we need.

Electronic News: Do established players have an advantage?

Dave: If it’s established IP such as USB, yes. If it’s brand new IP, such as in the wireless space, then I think it’s a little bit more open. In that case, we’re more willing to look at new players because they may have done something more innovative.

Electronic News: Let’s drill down on that. Do you trust unknown IP in a market that is untested?

Dave: We do and we have. ATI is a very aggressive company. That means taking risks.

Walsh: With a new protocol, it’s a much more level playing field. The issue is really one of alignment. If ATI is on a particular mission, new IP has to be in sync with their market windows. You need to look at the mix of IP providers. A company that has invested in IP and has an IP methodology and infrastructure does have an advantage.

Electronic News: You keep talking about a methodology. Does that really matter to the customer?

Walsh: Yes. A lot of our customers really understand the structure and what verification is needed for the designs we do. Some customers actually evaluate the source code itself. They read it for their interpretation of how well it’s constructed, how well it’s supported. As the market evolves and a company loses some of its specific expertise about USB, for example, we become even a larger component because we’re often explaining the function of USB to systems.

Dave: For us it’s not one of the top priorities we look at. If it’s established IP, as long as it’s been proven it’s probably okay. Everyone has their own way of writing RTL. Even with ATI, there are factions with different opinions about how RTL should be written. You only get so much credence for that. It’s more about the innovation in the part itself, how stable is it and how proven is it.

Electronic News: If you have to buy more than one IP component, is it better to get it from the same vendor?

Dave: Yes. It’s a lot easier to deal with one customer. Two different customers have their own manufacturer’s test vectors, their own methodologies, there’s a lot of material to wade through and one is in one format and another is in a different format. You don’t have to deal with these kinds of things if you have one vendor. But that’s still not the highest priority. If one has a better technology, that’s what’s important.

Walsh: If you buy more than one thing from a single vendor, you typically get a better deal. In addition, we build on our IP. We offer layer one and layer two mixed signal together, so we negate some of the risk. That’s important for customers as they deploy standards like PCI Express and USB.

Electronic News: Assuming price is fairly similar and this is standard IP, how do you judge one piece of IP against another?

Dave: Assuming price is the same, we look at size, the number of gates being used, performance characteristics, functionality—whether it’s identical functionality or whether they have additional functionality—and then whether it’s from a stable source. Is the company going to be around to help us debug it or integrate it, if we need to do that? How good are their [applications engineers]?

Electronic News: It seems as if the IP market has become attractive in the past couple years. Why now?

Dave: We’d rather have someone else go through the pain and get it certified and wait to see which one wins, and then license it from there. Unless it’s something near and dear to our hearts, then we’d consider doing it ourselves. With standard stuff, it doesn’t make sense to spend the research dollars.

Electronic News: But what shifted to make this attractive recently?

Dave: I think it’s a cost-benefit analysis. What are the costs, what are the risks, and what does it cost to integrate it versus develop it from scratch?

Walsh: In the case of companies that are in markets that are changing every 6 to 12 months, we can develop that and amortize it over a number of different chip instantiations. It’s a win-win situation. What’s happened in the IP industry is that we are involved much earlier in the process of emerging standards, so we produce IP at the same time a final spec is being certified. If we can do that, we actually reduce systems integration cost.

Electronic News: In the consumer market, hasn’t it always been a case of ‘good enough’ versus ‘mission-critical, no failures allowed?’

Dave: Our customers are pretty demanding, and they go through a vigorous testing for our chips before they decide. There’s a lot of competition out there. I’d be surprised, even in the consumer market, if anyone is allowing anything to come out that has an error.

Walsh: Regardless of the market segment, no one wants errors.

Electronic News: There’s a difference between standards-based IP and general IP. Which one is more reliable and easier to work with?

Dave: The standards-based IP is much easier to work with. They can prove it, and you know the glue that’s required around the IP. For more general IP, you’re filling a niche. You have to do a lot more testing to make sure it works the way you want it to work.

Walsh: We would look at that as a services-based engagement, where the customer is involved in defining exactly what they want. We’re not looking to commercialize that, because it may not be replicable in other chips.

Electronic News: Given the growth of the IP market, is there an advantage to being an integrated device manufacturer versus a fabless company?

Walsh: An IDM has a captive market, but I don’t think from an IP perspective there’s an advantage. The difficulty of creating IP inside an IDM is the same level of difficulty as creating it at external sources. We work with IDMs on assembly technologies. They struggle with making their IP portable between business units.

Electronic News: We have yet to see anyone successfully challenge an Intel or AMD or IBM. Is it the technology they develop or the business model that works?

Walsh: It’s difficult to take market share that already exists.

Dave: The bigger players have good technology. It’s tough for a smaller player to come in and unseat them. It’s hard to develop technology that will leapfrog them. To show a little bit of a granular difference is not enough.

Electronic News: In the software industry, there was a lot of hype over objects and everyone predicted startups would challenge the market leaders by cobbling together less costly solutions. That never happened. What’s different in the IP world?

Walsh: You can get disparate pieces from different parties, but historically there have been no good standards in place for interoperability. Certain things are happening in the industry now, such as the SPIRIT consortium, which are a first step toward being able to do that. But without that it is difficult to find all these disparate parts and put them together and compete with someone who’s been doing it for a long time.

Dave: It’s very difficult, if not impossible, for a small company to put pieces together and compete with a big player. They have to come up with technology on their own. If all they’re going to do is put IP together, unless they can break the cost down to half the cost of other competitors, I don’t see them succeeding. That’s what’s happened in the Asian market with boards. They’re using cheaper components and reducing the costs, and that’s what’s helping them win the market. But it’s not because of innovative ideas.

Electronic News: Cost is one of the big reasons the EDA market hasn’t grown significantly, and why complicated chip starts are down. Will IP reduce development costs?

Walsh: If you look at what creates a successful market, it’s the right combination of features. It’s not just whether you have a USB port on the product. The marketing guys want to make choices as late as possible in the design cycle. If you can merge the right functionality so you can put it on board so it makes sense, it will improve the cost. But if the IP isn’t available and you have to invest internally in that IP, then it becomes a feature in the product.

Dave: To answer your question, absolutely. We use standard IP to lower the cost dramatically. If there’s a standard, everyone has to apply the standard.

Walsh: You get this phenomenon where the customer feels they have to develop the IP for the first design. That advantage disappears for the second and third design. If you take a standard and intend to put it in several products, that’s an enormous cost.