Upcoming Hurdles For The Semiconductor Industry

Experts at the table, part three: Going vertical; consolidation; partnerships; playing nice; little-understood security.


Semiconductor Engineering sat down to discuss upcoming challenges and hurdles to overcome for the semiconductor industry with Vic Kulkarni, senior vice president and general manager, RTL Power Business at Ansys-Apache; Chris Rowen, Fellow and CTO, IP Group at Cadence; Subramani Kengeri, vice president, Global Design Solutions at GLOBALFOUNDRIES; Simon Davidmann, CEO of Imperas Software; Michael Buehler, senior director of marketing for Calibre Design Solutions at Mentor Graphics; and John Koeter, vice president of marketing, IP and prototyping at Synopsys. What follows are excerpts from that discussion. For part one, click here. For part two, click here.

SE: How has the manufacturing side of the industry evolved and what does that mean for the industry today?

Kengeri: Looking 15 years back, there were probably 30 manufacturers at advanced nodes, where there are only a handful today.

Buehler: There are companies that are making good money being the established node instead of the emerging node, but the emerging node slows down, they have no business.

Koeter: In my mind there’s no doubt that the EDA and IP industry has to go vertical to survive. EDA tools can stay maybe largely horizontal, but from an IP vendor perspective, you have to go vertical. One thing I always say is that if all you do is slap an automotive label on your IP, or an IoT label on your IP — which people try to do — you will fail. You have to optimize those IPs for that vertical market and more importantly, it’s not just about the IP, it’s about the software that goes around it, it’s about security, it’s about prototyping it, it’s about making it successful in the system context — it’s all of that and that’s clearly where the IP industry is going.

Buehler: I absolutely agree but they are ‘ands.’ You don’t get to drop the basic stuff we’ve been working our tails off on — and that costs more if you are going to add to these markets, but you’re not getting rid of the other core engineers that are building the tools, you’re adding to it.

Kulkarni: How do we get people to work together? We talk about ecosystem forever, from 30 years+ we all have in the industry. In reality, very few people really work together and customers are showing that already by acquisitions and consolidation. You see the Intel/Altera to Qualcomm/CSR to Micron/Elpida and so on. They are telling us something that you cannot avoid — you may want to really be serving particular mammals, as you put it [to Rowen], but the mammals themselves are combining their species — it’s because they see they cannot do it single-threaded. The laptop as a single threaded has a very limited use until you mix in a tablet and a pc/tablet. The PC concept is pretty much going, so the Altera acquisition is brilliant in that going to the programmable world of IoT application because the volumes are very small, as it were, but the technology is so deep. So how do we add value?

SE: Yes, we talk about co- everything but is it reality or is consolidation the answer?

Kulkarni: Or standardization, which is more than just a handshake. For example, the Veloce emulator with our RTL Power — this happened because customers told us. It was customer driven.

Koeter: These kinds of things only work if you have aligned self interests. If you don’t have aligned self interests between companies, it’s never going to work so you might as well buy somebody.

Buehler: That’s where you’ve got to get a customer to drive.

Rowen: There’s no question that acquisitions are hard; that you have to really invest deeply in figuring out everything from product rationalization to company culture has to line up, and you have to find ways to exploit the things you have. For the IP business of Cadence, I think the great opportunity and hence the great challenge is Cadence has a tremendous field channel, has tremendous contact with customers, how do you start enabling that channel to really leverage all of this rich IP portfolio very well. And we’ve made great progress on that but it’s an ongoing process of retooling a lot of people’s thinking as well as retooling the product to make it work in that world.

Kulkarni: The IP I consider to be so critical to the future of all of us, so you don’t even have to think about Cadence as your channel, we all could be your channel for some common customers, for example. With standardization, with good business agreement, driven by customers, not necessarily just because we are trying to work together.

SE: Although EDA doesn’t always play so nicely.

Kulkarni: That’s why we have to stop that because the market is telling us to stop it.

Kengeri: Chris, you brought up the point that integration is painful and challenging, but then what are the options? If you look at the other options, I’m trying to compare that with co-existence in an open ecosystem and trying to make it work. Both are equally painful, so it’s all about what is the better route. Maybe in some cases it does make sense, just to integrate, versus continuing to work in an open ecosystem. But, an open ecosystem doesn’t really mature that much so there has to be an aligned interest — in I don’t know how many different ways you can get there. The aligned interest doesn’t happen just like that — somebody has to take the initiative and drive the whole system, drive the whole solution, and then make it work.

SE: Is the open ecosystem working?

Buehler: Sort of. It depends. However, it’s better than it ever was because more people are concerned about it, and want to do it. But is it where we need to be to handle all of the derivatives that were built for sensors? We’ve got work to do, and we haven’t come up with a good model to do that yet. You do it if you have relationships.

Rowen: The best evidence that it is working is really the success of the fabless model, and the foundries over the IDM model. Clearly, in the past we did say, ‘It has to be all under one roof, and one logo.’ It has evolved, and it is successful, and it will go on being successful. I think people go on being very committed to the open model but there also are dimensions in which some of those interfaces or connections between things are particularly critical, because there’s particular opportunity to optimize across the boundary. In the typical collaboration model, there is adaptation on both sides to meet a standard but there are some cases where you don’t want to be constrained by an open interface standard because you want to be able to do things across that boundary much faster and more deeply, and to achieve technical results in terms of power, cost, area, and functionality that cannot be done using an existing standard. And in those cases then, to get those teams to collaborate deeply, it may take years of work, it may take hundreds of engineers on each side of a boundary to get the right thing, than integrating the teams much more deeply.

Koeter: From my perspective, the open ecosystem is clearly working with regard to IP — I don’t know what the exact numbers are with GlobalFoundries, but I know from talking with GF’s competitors that somewhere between 70 and 80 percent of their wafer volume is enabled by third party IP — that’s a big message to the foundry industry that the foundry industry is not going to survive without the IP industry.

Kengeri: That IP is close to 100 percent in our case.

Koeter: So it’s at least 80-ish percent. There is an example of two ecosystems — foundry and IP — working closely together and for clear mutual success.

Kulkarni: Another idea I think we can all work from is the system level side, and that is security. Take for example the 6LoWPAN model, the new standard emerging for sensor networks, then Zigbee, then ARM Trust Zone — and thanks to ARM for creating that ecosystem model. The number of people that come to, let’s say, ARM Tech Con, I see people who have nothing to do with EDA. EDA and foundries are only like 7 partners out of 700 they have, and those companies have put in a lot of effort which we can learn from by creating the TrustZone (embedded encryption from IP to the SoC to the 6LoWPAN) and then 811.4.1 — that, and then going into cloud, and cloud has slightly newer standards coming up under the IEEE umbrella. That is very important otherwise everybody will throw up their hands and say, ‘It is your problem, Mr. Hospital Care. What you do with so-called wearable computing? How do you make it into healthcare?’ Healthcare suddenly becomes a priority security issue while the wearable computing nobody cares how many miles you ran. These ecosystem partners of ARM are really collaborating to create the Trust Zone.

Koeter: The reality is that today, that’s a very small segment. It’s emerging and security is going to be critically important, but today, it isn’t. Accelerating five years from now, it will be the most dominant.

Buehler: In answer to your questions, it’s working for today. But when we look at what else we need, it needs to evolve and scale just like we have. If you look back to IDMs, and went in and presented GF’s ecosystem to LSI, for example, 15 years ago, they’d kick you out of the room. That’s evolution. It’s ok for today, but I wouldn’t call it 100% successful.

Kengeri: It was not as fragmented as what we are seeing today so I believe there is more compelling value now to come together.

Rowen: The thing that is most striking about security is just how little understood it is. Of all of the blind men, and all of the elephants, the blind men are blinder with regard to that.

Koeter: People want it, they need it, they don’t know how to do it.

Rowen: And there are many dimensions of it. One person will stand up and say it’s all about encryption, another says it’s all about protected execution, and another it’s all about generating unique root keys, and another it’s all about standards for discovery so you have this very diverse set of things and because in general, both the customers and the vendors have very narrow views of it, you get people standing up and saying, ‘I’ve solved the security problem because I offer X,’ and it doesn’t quite work that way. And there’s somebody in the audience who thinks that it’s all about Y.

SE: Do you think we can all agree security is a very important ‘AND,’ and a foundational piece of the system going forward, especially in IoT?

Rowen: Yes, and I think it has to start with education because before we can even talk about solutions, we have to get some agreement on which problems we’re solving.

Davidmann: I think we are just starting to be educated about the scope of the problem and we can just see the tip of the iceberg.