Сдам Сам

ПОЛЕЗНОЕ


КАТЕГОРИИ







The Bridge Between Manufacturing and Design






By Ann Steffora Mutschler -- Electronic News

Walden C. Rhines, chairman and CEO of Wilsonville, Ore.-based EDA tool provider Mentor Graphics Corp., sat down with Electronic News to discuss methodology shifts, foundry relationships and the growing speculation of foundries buying EDA firms. What follows are excerpts of that conversation.

Electronic News: Based on speculation by a variety of industry sources, do you believe foundries will begin acquiring EDA companies to protect their competitive leverage, as well as provide a more complete solution to customers?
Rhines: No, for several reasons. One of the things that is valuable about foundries is that they are EDA-independent. And while they have golden tools and specific design flows they support and so on, their first loyalty is to supporting the good customers who will do designs that end up in high-volume. This same issue came up quite a bit in the late 1980s and early 1990s as foundries were emerging as a significant factor and some of the early people in the foundry business - Morris Chang [founder of TSMC] and Don Brooks who worked for him - were quite aware of the impact of ASIC companies with design support versus the customer-owned tooling business. Don Brooks is one who, at TSMC, as the ASIC support things started to grow, he squashed those. And he squashed them because he said, "Look, we need to be really good at what we do and we need to be able to produce wafers and have best-in-class processes. To get into the design tools business or design support business or routing chips for customers or doing other services is a diversion that will raise our overhead and make us look more like an ASIC company and remove our differentiation."

 

Electronic News: How have the foundries been successful with their business models?

Rhines: One of the great successes that TSMC had was they focused on the job of manufacturing and they worked with the third party EDA industry and they worked with service companies, like eSilicon, that could assist customers in getting pattern generator tables released and every foundry in the world worked closely with Mentor to be sure that Calibre was their golden standard that all of their IP was verified with Calibre. They didn't do that with any other tool, they never have, they probably never will but they still, if a customer came in with some other physical verification tool, they would accept the design. The only limitation was, if there was a dispute, everyone understood the common communication will be Calibre. But if there were no disputes, they would let you use other tools. All the things the big foundries have done have been mirrored elsewhere and I think the reason the foundry business is so successful is that they are largely EDA-independent, they don't try to get in the way of the third parties adding value, either their customers, the design support infrastructure, the embedded software infrastructure even the IP business, they minimally involve themselves. In fact, they do everything to try to encourage those people to release physical layouts that will be compatible with their design rules.

Electronic News: What about the fact that designs at 65nm and below are much too complicated and more data from the foundries is needed? How will this be accomplished without some consolidation?

Rhines: There are two ways to solve the problem. It is either the consolidation of foundries with EDA tool providers which is to hand customers a toolset so you can keep everything proprietary or why not work with the leading EDA suppliers to provide the information they need so the tools will work? Although this has never been much of an issue with the smaller foundries, there were some leaders in the industry who were reluctant to provide the information and still today are reluctant to provide as much detail as the tool people would like. But they have come a long, long way and they are basically supporting at least the leaders with the information needed to tie together design with manufacturing in areas like litho-friendly design. I believe competitive pressure will force them to provide even more.

Electronic News: What do Mentor's foundry relationships look like today?
Rhines: Mentor works closely with all the foundries. They are customers as well as the vehicle through which we work with our other customers. We have to work with them to support our libraries, our tools, our flows and to make sure that our customers will be taken care of and make sure that their customers will have the option to use the best design methodologies. We have a lot of dedicated people who do nothing but work with the foundries both for that support as well as customers. The sales team is assigned to them because most of the major foundries use Mentor for their resolution enhancement, absolutely every foundry in the world uses Calibre as the golden verification standard - no exceptions. As we work more closely with them, a lot of them are using us for their mask data preparation, their interaction with the photomask shops. And more and more foundries are supporting analog and mixed-signal designs where being the largest mixed-signal design tool provider we have to get our libraries supported there and that's a big part of the relationship, also. I've never had a year when I didn't have dinner with Morris Chang since 1985.

Electronic News: How is the EDA landscape shaping up as the industry makes the next major methodology shift?

Rhines: The whole methodology never changes at once - pieces of it change. We are ready to move to the next higher level of abstraction and those only occur on the average every 15 years and we have been, by most measures, probably 18 years since the last change. In the late 80's we went from gate level design to register transfer level [RTL]. Before that we went from transistor level to gate level. Complexity has gotten to a level that we are overdue, and so it is moving to electronic system [ES] level. Methodology change is what causes growth in EDA and the whole methodology doesn't change, just a piece of it does.

Electronic News: What are the different pieces of the ES level?
Rhines: For ES level, you are moving to the system level so it means that you describe the system in a higher level language that presumably includes both hardware and software. For most people that is C++, but there are lots of other approaches. UML for example is the next higher level system language and is used in many cases to generate C++, which can then be used to generate Verilog and VHDL. Another step that some view as an implementation step is from Verilog to SystemVerilog, which take us from the language to incorporation of testbench and verification methodology, which is a sort of half-step in moving to the next higher level of abstraction. Then there are a whole variety of companies pursuing various tools to do higher level things, mostly system-level trade-offs. We've had lots of products that do various things to test out embedded software or test out embedded IP in a larger system design and the ultimate solution in moving to the next level will involve multiple ones. We focus a lot on the algorithms that get implemented in silicon and in direct synthesis from C++ to Verilog and VHDL and ultimately SystemVerilog.

Electronic News: What is the role of the FPGA at the system level?
Rhines: There is no question that system level designs are increasingly incorporating FPGAs for lots of reasons. One is that the gap is closing between FPGAs and custom designs for what you can achieve in terms of performance. It's not closing as quickly for what you can achieve in terms of power, although there are special purpose FPGA providers like Quicklogic that do low power designs for FPGAs. They are trying to close that gap and I think they'll make progress. The big motivation is really the difference between a wafer fab cycle limited development process versus the trade-offs you have to pay to avoid that. With an FPGA, you can customize without going back to the wafer fab and that has a lot of value because you can try more alternatives and because the cost of tooling is going up, but more than anything, it's the time associated with doing multiple design spins, so if you can avoid that, you are willing to pay a price for performance or power or something else. If you want the ultimate in low power, high performance or lowest cost, it still is a custom integrated circuit, but not everyone needs that and so the percentage that can get by with FPGAs is increasing and the thing that's been really unique is that through innovative architectures, performance is no longer necessarily a disadvantage. Sometimes with an FPGA, you can implement parallelism that you couldn't in a reasonable time implement with a standard product and for which you might not be able to do with enough flexibility even in a custom product.

Electronic News: What else has the ability to change the competitive landscape of the EDA industry?

Rhines: Design for manufacturing because you are bringing people into the party that weren't here before. DFM brings manufacturing and design people together and the bridge to that are EDA tools and methodologies. 2005 was the first year we saw millions of dollars of revenue for pure DFM tools. Before that, we had lots of tools but most of our DFM-related revenue was in peripheral fields such as resolution enhancement.

Electronic News: Are we likely to see more use of standard products combined with software as the customization piece?

Rhines: Our industry has a long history of cycling between custom and general-purpose. The manifestation that is usually written up about this is called Makimoto's wave where you go from full-custom to standard product. In the late 60s as I became involved in semiconductors, the standard product was in logic, TTL and in memory, dynamic RAMs were just coming out so it was mostly shift registers. The standard product [custom logic] gave way to a lot of custom designs in that period for calculators and other emerging consumer products that needed full-custom designs. The microprocessor came along and became a general purpose product that you could use in lots of applications so custom became much less important in the 70s and in the early 80s. In the mid-1980s, ASIC came along and gave people customization capability either around a standard product or full-custom. But the things that are missed in between are the gradations. For example, Intel, TI and a bunch of companies made microcontrollers that had EPROM ROMs on them so you could reprogram them. You could customize a product you had already bought, in the field - essentially field programmable. There was an era of output programmable logic arrays on bipolar logic where you could program the outputs in one of several modes. The way the real cycle works, you've got a problem that is solved with a custom chip, then someone else solves it with a different custom chip and eventually someone comes along and creates a general purpose chip since it can serve a broad range of applications.


Sounding a New Cadence

By Suzanne Deffree -- Electronic News

Two years into his president and CEO position at Cadence Design Systems Inc. Mike Fister sat down with Electronic News to discuss the EDA market's history of acquisitions and its future. What follows are excerpts of that conversation.

Electronic News: From the business side, what's the future look like for EDA?
Fister: We are optimistic about it. We're optimistic about showing value to our customers. It sounds like a cliché, but when we show more value to our customers we will expand the relationships that we have and therefore our ability to not only impact them, but make money for our shareholders.

Electronic News: What's been holding back the growth?

Fister: The industry has been in kind of a no growth fashion for years now, maybe almost five years. Some of that in my mind has been rooted in niches or nuances of technology with little demonstration of how to help customers be more productive, get their products to market faster, and deal with the incredible complexities that they have. Those three mantras, productivity, time to market and managing complexity, have been there ever since I came into the industry, and we showed people "what" they were trying to do and how to accomplish it, opposed to concentrating on "how" they do it. Sometimes I use those what and how words as a demonstrable of how the EDA industry got so stagnant. We got rooted into a commodity fashion, niched into places where we couldn't demonstrate a lot more value. It was about sheer replacing somebody else. We've really taken that as the primed to-do. The demonstrables are good to our customers because those are their three big problems. And I think in some respects we really have resonated with the sentiment for them and it's one of the reasons that as we added added some people that have chip experience to our company. Those are people who really understand the what, not just the how. Tools are a means to an end, not necessarily the end itself. Sometimes when you are in a business you can get so focused on what you do, you can miss the forest for the trees. That point of resonance, that admission, I think has done a lot for our creditability.

Electronic News: EDA historically has had a very big focus on the tools, on the "how". Part of that is the EDA model of innovating through acquisitions to gain other companies' technologies. Will that continue?

Fister: As companies were very deeply niched, there was almost a kind of revolving door of talent that would create something then leave, start a new company, then get acquired again. That might happen multiple times. I think that was in many respects a fault of the industry to be able to keep those talented technology people innovating in a direction that was for the growth of the company. That's not a selfish response for the big companies get bigger. It's a practical response because when I was a big customer of EDA, it's very difficult to bet your next product cycles on very, very small companies, especially when you are ramping microprocessors, and that was my life. The financial stability that a company like Cadence offers, being globally diverse, those are things that are very difficult to do in the current environment. We've done things to demonstrate that companies the size of Cadence can persist and grow. It's much more efficient for us, too. There is a lot of redundancy that gets built into elements of the tool or method. Why do you want to do it over and over again? It just costs money and wastes time and effort. Those are a couple of things to think about. That doesn't mean we will never do another acquisition. Certainly, there's innovation in all parts of the industry. We openly support that with our open data initiative that we call Open Access that allows customers to innovate and competitors to innovate and interoperate with our tools and methods. In that respect, I think we dare ourselves to be very good. Where we are not, or a customer has a peculiar thing they want to do themselves, or a particular technology startup that does something interesting, we welcome that.

Electronic News: How are you keeping the people you want in your company from going through that revolving door?

Fister: A lot of it is building a culture. There are cultural examples in EDA that were much richer in the 1990s than they are today. Some of that was a common passion, an appreciation of the people. It doesn't just have to be monetary, either. People love to be recognized when they do something. They love to see their ideas realized. We've done it with financial incentive; we've done it with attention. The recognition by peers is such a cool thing. An awful lot of it is setting the targets that we do, achieving them together and rewarding that in all localities. And it doesn't hurt that we are a growing company in a field where a lot of the companies aren't growing. Some people are better thinkers, some are better intuitives, some are great starters, some are great finishers. We actually have fostered incubation environments inside the company where a person or small group of people can go out and test an idea without a lot of worry. It's just a practical way of walking the talk that we value the people in our company.

Electronic News: On the incubator point, do you see the EDA market changing from an acquisition and merger environment to a parent company relationship, more in-line with spin-offs?

Fister: I don't know. I suppose it's possible. I think one of the ways to demonstrate real value and partnerability, opposed to just being a supplier, is by the breath of what you offer. I think of problems from the beginning to the end, the end-to-end. That kind of a holistic view is a way to incorporate lots of elements of technology, through utilization of the targets. I think companies that either didn't have clear incubation goals and/or were very acquisitive never really had as much of a total plan of that end-to-end dynamic. I'm not being critical of anybody, I'm just observing. The IT industry went through the same kind of transition. In that way, I would tell you that I am absolutely thoughtful of the roadmaps we perpetrate. We consider the roadmap to be the end-to-end. It gives us a path to identify and develop technology inside and work with partnering elements of the world, whether they be inside efforts inside our customers, maybe smaller companies or new companies, even if we should consider OEM collaboration, distribution agreements, and even ultimately I suppose acquisition. I believe the most successful companies have that total goal in mind and then have a very big attitude that it's not just about owning something. We partner with companies, we collaborate, and ultimately maybe there are some that we choose to partner even more with and integrate together. There are all those elements to deal with. It's what the biggest and best companies in the world do and it's what we do.

Electronic News: So less buying of companies, more holistic innovation?
Fister: Yes, and I think the consequence of less buying of companies is a byproduct of demonstrating internal innovation and that compliment to a customer has a net affect. That's certainly incumbent that that will happen because it's just too difficult for companies to start and really gain a lot of momentum. As I said, the bets are bigger with customers. They can't afford to bet a lot of their success on very small entities. The stakes are very high. I think you see a lot more early innovations and companies will continue to flourish, or maybe you'll never even hear of them and they'll be integrating faster. Maybe people will never leave and the innovations will happen and that would be as if they virtually started on their own. It's a whole kind of a different environment.

CONTENTS

1. Digital Products May Drive Recovery 3

2. Cell Phones: Physical 5

3. Europe Snaps Up Camera Phones 7

4. The Fluttering of Tiny Pixels 8

5. Microelectronics Grows Up 10

6. How Many Small Libraries Can Be a Large Library 13

7. China Launches 300mm Manufacturing 15

8. Semiconductor Fabrication 19

9. A Million Points of Light 23

10. Managing Complexity 25

11. Concepts of ESD Control 28

12. Needed: An Admiral of the "Nano Sea"? 30

13. Atomic Force Microscopy 32

14. SiGe - a Basic Material for Silicon Technology 35

15. High-Performance Bipolar Transistors

with SiGe:C and Poly-SiGe 38

16. Seeking a Comprehensive Automated Wafer Inspection
Solution for 300mm 40

17. Intel Quality 42

18. Wafer Fab Profit Opportunities and Costs 44

19. Japan's Super Silicon Group Moves Toward 400-mm Wafers 48

20. KLA-Tencor Improves Patterned Wafer Inspection 50

21. KrF Lithography: Not Finished Yet 51

22. Origami Unfolds 53

23. TSMC Blows the Doors Off Q4 56

24. Turbulent Times Ahead, Gartner Says 57

25. The Bridge Between Manufacturing and Design 60

26. Sounding a New Cadence 65







ЧТО ТАКОЕ УВЕРЕННОЕ ПОВЕДЕНИЕ В МЕЖЛИЧНОСТНЫХ ОТНОШЕНИЯХ? Исторически существует три основных модели различий, существующих между...

Что вызывает тренды на фондовых и товарных рынках Объяснение теории грузового поезда Первые 17 лет моих рыночных исследований сводились к попыткам вычис­лить, когда этот...

Система охраняемых территорий в США Изучение особо охраняемых природных территорий(ООПТ) США представляет особый интерес по многим причинам...

Что делать, если нет взаимности? А теперь спустимся с небес на землю. Приземлились? Продолжаем разговор...





Не нашли то, что искали? Воспользуйтесь поиском гугл на сайте:


©2015- 2024 zdamsam.ru Размещенные материалы защищены законодательством РФ.