March 30, 2009
How to Compete Against the Big Three
Please note that contributed articles, blog entries, and comments posted on EDACafe.com are the views and opinion of the author and do not necessarily represent the views and opinions of the management and staff of Internet Business Systems and its subsidiary web-sites.
| by Jack Horgan - Contributing Editor
Posted anew every four weeks or so, the EDA WEEKLY delivers to its readers information concerning the latest happenings in the EDA industry, covering vendors, products, finances and new developments. Frequently, feature articles on selected public or private EDA companies are presented. Brought to you by EDACafe.com. If we miss a story or subject that you feel deserves to be included, or you just want to suggest a future topic, please contact us! Questions? Feedback? Click here. Thank you!
If you are a firm in EDA industry and not a member of the Big Three club, sooner or later you will have to compete against them. Unless of course, you carve out a niche so small, that it will not hold any interest for the Big Three. In the end user community you have to compete not only for budget dollar but also for mind share, attention and resources. It takes resources for an EDA end user just to evaluate new products, much less to fold such products into their existing design flow.
The competitive struggle is not a fair one. A new player has to fight simply to get noticed, while the Big Three already have a global market presence. Any target prospect is likely to be a current user of products from one or more of the Big Three firms, even if it is a different department or different division. Their relationship may be one of long standing.
There are varying approaches for how to compete against dominate players in a given market. You can adopt the Yogi Berra philosophy of successful batting which is to “hit it where they ain’t”. You could develop a niche product in an area outside the scope of the Big Three offerings. Such innovation is challenging as is getting anyone’s attention once you have the unique product. You could also target different geographies or different end user industries. Another approach is to offer products that are complementary or add-ons to offerings from the Big Three. This approach is vulnerable to an expansion or broadening of offerings from the Big
Three. The third approach is to offer a product which overlaps offerings from the Big Three but one that possesses some combination of a) superior technology as measured by some objective criterion such as speed or capacity, b) lower price point, c) lower cost of ownership, or d) easier to use. The best approach would be a combination of all of the above. A small technical or economic advantage is easily overcome by any of the Big Three. They point to the obvious risks of going with a small firm. They suggest that their own products will soon catch up or leapfrog the new kid on the block. This approach requires a significant and sustainable advantage for end users to commit. As they
say in the VC community “What is your lasting competitive advantage?”
Competing on price only is difficult because the true cost of individual products from the Big Three are buried in complex all-you-can eat and product swaps arrangements.
One company that has adopted the lower price point, easier to use and technically superior approach is EVE, an emulation vendor that has been growing steadily for many years now. I had an opportunity recently to interview Lauro Rizzatti, VP of Marketing and GM of EVE-USA.
How are you this morning?
Actually I lost my voice. I was in Paris. It snowed and I did not have winter clothing and here are the results. There were three snow storms which is very, very unusual. The weather here is more wet and rainy than snow. They told me that it was a record year for them.
Would you give us a brief biography?
In a nutshell I graduated with a Doctorate in Electronic Engineering in Italy from the University of Trieste. Then I started to work in the telecommunication industry for a number of years; first in Italy and then in Germany. Then I moved into ATE (Automated Test Equipment). Teradyne was the company I worked for. I spent more than 10 years with them. My role up to that point had been either as a hardware designer in telecom or as an application engineer at Teradyne. Then Teradyne moved into the EDA space. They made an attempt in the eighties, which was not successful. They tried to move outside the ATE space. That experiment did not succeed. I was already in EDA at Teradyne during
those years. I understood that there was no future for them there. I ran into Mentor Graphics in 1985. T the time Mentor was emerging as the leader of the then big three, Daisy, Mentor and Valid. I received an offer from Mentor, left Teradyne and joined Mentor as an application engineer. I moved location from Boston, Massachusetts to Portland, Oregon. I spent the next 11 years at Mentor. At Mentor I moved outside the traditional application engineer work more into technical marketing and then into product marketing. Finally, in the past, I would say, 9 years I have been working for startups; first Get2Chip and then EVE, getting into more executive roles, especially at EVE. For EVE I opened
up the US operations. EVE is essentially a French company, created in Paris by ex-Mentor engineers, scientists. They worked previously in the emulation division of Mentor Graphics. I was associated with them back in the nineties, so there was a connection. I opened the office in San Jose. Since then the company has grown from one employee to about thirty people today in the US. Overall, the company has 110 people. That’s it in a nutshell.
So, you were not a founder of EVE.
I was not a founder of EVE. I started with EVE in the early, early days. I believe that I was employee number seven, the first in America. The company was founded by four engineers, French engineers that left Mentor.
Were the founders employed by Mentor in France or in the US?
Always in France. They were part of a small company called Meta Systems that Mentor acquired in the nineties when the decision was made by management to get into the emulation business. It was actually part of due diligence for Mentor. They were looking for existing emulation companies that Mentor could acquire. One was QuickTurn but that was already a well established company and very expansive to think of purchasing. Meta Systems was a tiny company with 15 employees. They had very advanced emulation technology at the time. Mentor made the move and acquired them. That’s the history behind EVE.
Where did the founders get funding?
They started by providing consulting services in the emulation business for end users of emulation systems. In particular, they worked as consultants for Broadcom, Ericsson and STMicrosystems, who were users of Mentor (IKOS) and other emulation systems. These consulting services generated the initial cash flow to begin the development of what later became EVE ZeBu emulation systems. For three years from the inception of the company, which was in 2000, the company was self funded, with of course minimum salaries and all of that. Indeed, when I joined the company, we did not have funding yet. So I worked for the company for five months without any salary. Essentially, I had stock as a
way to compensate in those days. The first round was closed in 2003 at 3M euros. We had a second and a third round, two years apart for a total of 13M euros. Now, for three years we have been sustaining growth with our own revenue. We have not had any further funding since 2006.
You have been growing at 30% to 45% per year.
Until 2007 the compound growth rate was 100%, one year higher, one year lower. The average from 2003 to 2007 was 100%. Now we are down to 30%. It would be impossible to sustain 100% growth. In 2008 we grew 30%.
Can you tell me what was the revenue is dollars or euros?
Sure, approaching $30 million, below $30 million but in that ballpark.
The company name EVE stands for?
It stands for Emulation Verification Engineering.
You web site is EVE-team.com
That is the website and email address notation. The actual company is EVE. In France it would be EVE SA. In the America we are EVE-US Inc. We also have subsidiaries in Japan (Nihon EVE KK), in Korea (EVE Korea, Inc), in Taiwan (EVE –Taiwan) and in India (EVE-DA Ltd). The Japanese office is fairly well established and large. We have thirteen to fifteen people there. The other offices I mentioned are small office with two employees, a salesman and an application engineer.
Most EDA companies are US based with offices and/or distributors overseas. EVE’s situation is the reverse with a headquarters in Paris and a sub in the US.
Correct. We have debated in the past to do a switch, to make the US operation the mother company, mainly looking for funding. But ultimately, we were able to raise money in Europe from French and British venture capitalists. So the need to do the switch was never pressing. That’s a bit of the history of the company. I agree with you until today. I see now a trend to move to Asia. SpringSoft is probably a classic example, possibly becoming the number four EDA company. If Magma is not going to be successful, SpringSoft might become the number four EDA company.
There is also another reason to keep the activities in France. The French government is really supporting initiatives, startups through reimbursement of R&D expense and through zero interest loans, multiyear loans. We are talking here of millions of dollars. Loans of $2 million are quite possible. We get back every year 30% of all the dollars or euros we spend in R&D. Out largest R&D entity is in France. We do have a few R&D people in the US. Actually, they came onboard through an acquisition we did two years ago. But because of this opportunity to be funded, to be supported by the French government, we continue to invest in France.
What was the end user problem or the market opportunity, the founders were trying to address?
Emulation technology has been around for over twenty years. The largest company in the past was QuickTurn, which was finally acquired by Cadence. The second one was IKOS which was acquired by Mentor. We worked in emulation with Mentor. The problems with those implementations were very multifaceted. They were expensive machines, in the millions of dollars range. So they were limited in terms of market acceptance. You had to be a big company like Intel or like Nvidia to spend that sort of money. The other issue was that when you were deploying the QuickTurn box, you had to provide an army of application engineers to make the system work. So it was not only expensive to acquire, it was
also very expensive to maintain. The cost of ownership was very, very high. What we did at the time was to identify an opportunity to implement an inexpensive, I would not say necessarily cheap, solution that would be easier to use and would require fewer engineers to deploy. That was difficult. That was the thinking behind our ZeBu emulators; small, inexpensive and through stacks of software (compiler and runtime software) make it easier to deploy without this need for an army of application engineers. That was the idea that came about. It was not a revolutionary idea, but without this we would not be as successful as I believe we are today. We have sixty plus companies today that have
adopted our ZeBu. These include big names like STMicro and NXP in Europe, in the US we have Qualcomm and LSI Logic, a bunch of consumer companies like Gennum, HDTV stuff like decoders, in Japan Toshiba, Panasonic and others.
In the past twelve months we have seen opportunities in China and Taiwan. That is why we have small offices there. We are already in Korea. We have started to sell in all three countries.
In a nutshell we are easier to use and less expensive than any other emulation system. When we build our machines, we make sure that they will execute, will run and will execute your design at higher speeds than the traditional emulation systems. So we run at multiple megahertz whereas the big boxes typically run below 1MHz. So we are five to ten times faster, less expensive and easier to use.
The principal use of the system is for hardware/software co-verification.
Yes! Definitely yes! You know that the problem today is that not only are the chips big whether measured in transistors or measure on gates. The latest chips from Sony are approaching forty million ASIC gates; so big, big chips. They used to be below one million gates only a few years ago. The processor and graphics chips have hundreds of millions of gates. So one use of our system is for complexity. Another important aspect of design is that that there is a lot of software embedded in memory of the chip. Simulators do not do any good there. They would run for weeks and weeks or months to move any data in and out of a chip. Processing software requires lots of power. That is actually
our main application, i.e. hardware/software coverage more than traditional debugging.
Eve’s product is a board, a box, ..?
Actually, we have two product lines. A board, a smaller system with limited capacity but high speed and of course less expensive and a box. The box approach can be expanded, interconnected to form multiple boxes. We can interconnect up to eight boxes for 800M ASIC Gates versus one box at 100M ASIC Gates.
ZeBu-XXL Extra-Large Emulator
Configurable to handle designs from 12.5M to 100M ASIC gates
You said your system was not cheap but inexpensive. What is your definition of inexpensive?
An inexpensive chip would be less than $2K. In EDA, when you have a commodity like a simulator or a synthesizer you probably want to spend $5K to $10K. That would be the cheapest in my mind. Inexpensive has to be below $100K. We have emulators from $50K, some even below $50K list price to $100K and over $100K. That would not be the big boxes for hundreds of millions of gates. But if you are designing something for five million gates, the price would be around $50K.
What does a user need to connect to box?
Physically connected? It would be a PC. Throw in a PCI Express board. In the PC you would have not only the design loaded, downloaded into the box through the PCI Express interface, but also interface to the FPGAs. Also residing in the PC would be your testbenches. It could be a very big RTL testbench. It could also be C, SystemC, C++ or SystemVerilog testbench. The uniqueness of our approach is the ability to interconnect the testbench at the transaction level to your design in the box. When you have a transactor in between the testbench and the design, the performance will be orders of magnitude higher, faster than RTL, HDL simulator type of interface. In other words, the machine
can be deployed in multiple modes of operations; driven by simulators, driven by C testbenches, at cycle level or at transactor level. Of course being an emulator, it can also be deployed in what is called ICE (In-circuit emulator) mode, where you have a physical target system driving your design matching into the box. That is the traditional ICE approach, which people were using ten years ago with IKOS and with QuickTurn. Today, we put more and more emphasis into developing a better interface to testbenches and software testbenches.
The target users for your product; are there certain applications that are better suit for emulation or is it simply a matter of equivalent ASIC count?
Typically, it would be SoC designers. The best fit would be designers who not only develop the subsystem, the silicon SoC hardware but also the embedded software, where they can deploy the emulation to begin testing the hardware and then the integration of the hardware and embedded software, which would be drivers and operating systems but also application software ahead of real silicon. Have a head start developing application software before tapeout, before having silicon available. So the ideal user is a SoC designer or SoC team (not one person but 20 people or whatever) that adopt this for testing the hardware integration with the software and debugging the software.
How far along must the hardware design be before software engineers can do something meaningful given they have the EVE product?
Qualcomm once told us that the technology will allow them to subtract between 3 and 6 months off their development cycle. This number came from a user. That is because of a combination of accelerating the verification of the hardware but also the ability to begin software ahead of the silicon. So between three and six months. In today’s competitive world, let’s say three months could be millions of dollars in a product.
The underlying technology of EVE’s products is
The hardware is FPGA. We use Xilinx Virtex devices, the largest at any given time. We started with Virtex-2, then moved to Virtex-4 and then to Virtex-5. We are developing products for Virtex-6 based machines. Obviously they are not available yet. That is the commercial device we use in our boxes and boards. Most of the innovation comes from the software. Just to give you a data point, the composition of our R&D team includes 90% software developers and 10% hardware developers. Most of the R&D effort and dollars is in software. The compiler that will take your RTL design and automatically synthesize and map it into the box is a very challenging task in today’s designs.
You have to deal with not just sheer complexity but also with clocking schemes, clocking trees, low power designs that have very complicated architectures. The mapping of a design in an automatic way without human intervention is a major challenge. The same goes for mapping memories (we support DDR2, DDR3 and whatnot). We want to do that automatically. We do not want the user to do anything manually. Lots of efforts are put into that. The other side of the coin will be the runtime software. I mentioned before that the box works with software testbenches in a multitude of forms like Verilog, C testbenches, signal level, transactional level that interface in a complicated manner. We have
teams of designers improving and enhancing constantly. In a nutshell 9 engineers in software to 1 engineer in hardware to make this product workable, deployable, and easily adoptable.
Are the end users software engineers using their own standard compliers and debugging environment?
The debugging environment, we interface to simulator so you can use your simulator of choice; VCS or ModelSim and use that for debugging when the design is mapped into the box. But we also provide means to debug your design without the need for a simulator. It comes with our software.
Are there any other firms providing products comparable to EVE’s products?
Our major competitor without a doubt is Cadence. They have this machine called Palladium, which went through four generations. Today, without doubt, it is a well established and successful emulator. We compete with them. Today what we give to a user, besides being a quite less expensive machine, is higher speed, an order of magnitude higher execution speed of your design inside the box than Palladium. But our efforts in 2009 and 2010 will be to improve everything in the box with the next generation Virtex-6 that will beat Palladium on all specifications. Our mission is to become the number one provider and take the lead from Cadence. We also compete with Mentor Graphics but we believe
that we passed Mentor already a year ago or longer. Then there are FPGA prototyping companies. We have seen that Synopsys lately acquired two companies (Hardi and ChipIt). Although they are not emulation providers, we feel that they will be competition down the road. We are preparing to compete with Synopsys as well. Another one is definitely Cadence.
Do you expect the demand for EVE’s products to grow substantially over the next several years?
Yes! Definitely yes! If you consider what has happened over the last 20 years, emulation in the nineties was only done by CPU designers and graphic chips designers. Nobody else used emulation. It was too expensive and too complicated to use. In early 2000, which is when we began, the wireless industry started to adopt emulation mainly because of the embedded software. Our first customer was TI wireless. As of today, they continue to purchase our machines, bigger machines, more powerful machines. But, in the past four years, so from 2005 to today, the entire consumer industry is moving into this space; digital TV, camcorders, digital cameras, set top boxes, and multifunction printers.
They are all using our technology, not only our technology but emulation technology because of the complexity of embedded software. I can only anticipate that the future will be more of this rather than less of this.
Does the current economic environment, the global recession …
Of course, it is a concern. What we saw for instance was a change in purchasing our machines from buying up front into leasing; one, two or three year leasing. We did not see this in the past. That is one trend. Another trend, especially with big users, is not to buy additional boxes, additional emulators, but rather to use existing machines and share them between teams. That of course has an impact on our revenue. We had two instances of big customers saying “I will postpone the purchasing of new machines for six months and I will use other machines internally, sharing them among teams.” Indeed for 2009, we do not expect much growth at all. I think either having
flat or 10% to 20% growth will be achievable in this climate. But I do not expect shrinkage of our business.
One of the alternatives to EVE’s products is for end users to do rapid prototyping using FPGAs. Why would you say that such an approach is less effective or less attractive solution for the end user?
As of today, when you take that route, you are limited in terms of design complexity. Rapid prototyping is typically limited to four or six FPGAs, no more. If you need bigger than that to hold you design, then the provider might not give you enough FPGAs. If they do, they do not have the software to help you automate the process of mapping the design and certainly no interface or a poor interface to the software testbenches. Rapid prototyping which is successful is mostly successful in-house. We see left and right everyone is building prototypes in-house but mostly it is for small designs. Very few companies have teams designing things with twenty or thirty FPGAs. Very, very few
companies. It is a major effort to do that.
A year or so ago, EVE acquired Tharas Systems.
Two years ago we acquired Tharas, which was an acceleration company. That was not too successful. The main drive for that acquisition was their compiler technology that we use today in our emulation systems. A compiler that is very fast and has some behavioral synthesis mapping capability that we use in creating transactors for transactor-based verification. We use the complier for the design. We launched something called zFAST (ZeBu Fast SynThesis) synthesis one month ago, a very fast, ten x faster than the best in class FPGA system that comes from Tharas. It is interesting. When we announced that this technology is supporting ZeBu, we had users ask us to make it a standalone product
that would work with FPGA rapid prototyping because they are not too happy with the commercial solutions. Those solutions are too slow.
Editor: zFAST is available now as an add-on feature for ZeBu. It is priced from $50,000 for the first license, and from $15,000 for additional licenses.
Customer, prospects are asking. Is EVE thinking of going forward with such a product?
Delivering the product?
No decision has been made. Until now, it has not been a thrust for the company. We wanted to provide an entire self-contained solution that includes the hardware and the software. But I can say that if we have enough demand, we will certainly seriously consider it. It would not be a major effort and a major investment to make it work as a standalone product, by the way. If the demand will be there, we will do it.
Has EVE made any other acquisitions?
Lots of internal brainstorming. We are looking at the ESL space. We can interface the emulator to a high level of abstraction of a design. We run in CoWare, Virtutech and Innovator from Synopsys and take charge of the RTL portion of the design, maintaining a high level of execution speed because of transactors. So we can stick for instance an AXI transactor between the ESL environment and the RTL block of a design mapped into the box and run at multi MHz. We have demo vehicles with CoWare, Virtutech and Innovator as well as common customer that are supporting this. So we are looking into that space and maybe something will come of it but no decision to date to move forward with any
one of these nor have any serious discussion about merging or acquiring taken place. We will see down the road. We are looking at other synergistic deals in test bench automation, test bench coverage and other companies, where maybe we could work together or maybe we could merge for that matter.
I am out of questions. Anything to add?
If you would be able to capture all that I told you in a nice article, I will be extremely pleased and happy.
The top articles over the last two weeks as determined by the number of readers were:
Silicon Valley jobless rate hits 10.0 percent for first time since 1990 (Mercury News)
The combined unemployment rate for Santa Clara and San Benito counties in February hit double digits —10.0 percent — for the first time in almost two decades, as the recession began taking a bigger bite out of the tech economy.
It was the highest unemployment rate since at least 1990, when the state started tracking the data with its current system, and the highest number of unemployed at 92,800. The valley lost 26,300 jobs, or 2.9 percent, in the past 12 months, according to the Employment Development Department, which released the figures today.
I.B.M., Looking to Buy Sun, Sets Up a Software Strategy (New York Times)
I.B.M.’s interest in acquiring the server computer maker Sun Microsystems for nearly $7 billion may seem at first to be a reversal of its recent efforts to move away from the hardware business.
But analysts say there is more to Sun than servers, which are used in corporate data centers. They say its strengths in software, systems design and research make it an attractive target.
The price tag being discussed by the companies works out to nearly $10 a share, a person with knowledge of the negotiations said on Wednesday. That is approximately twice Sun’s closing price on Tuesday. Shares of Sun surged nearly 79 percent Wednesday on news of the negotiations, to close at $8.89.
IEEE Approves New Low-Power Design Standard for Integrated Circuits In Electronic Systems
The IEEE and Accellera announced today that the IEEE has approved a new standard, IEEE 1801™, "Standard for Design and Verification of Low Power Integrated Circuits."
The standard is also known as Unified Power Format (UPF) 2.0. UPF, first developed by Accellera, is currently supported by multiple vendors and is in use worldwide. This is the first time that UPF has undergone an IEEE standardization effort.
Developed by the 1801 Low Power Working Group, the standard provides portability of low-power design specifications that can be used with a variety of commercial products throughout an electronic system design, analysis, verification and implementation flow.
Real Intent Continues Growth Worldwide, Partnering With Satris in Israel
Real Intent Inc has signed Satris Corporation in Israel as its newest distribution partner.
Satris is providing sales and support for Real Intent's prospects and customers in Israel, and is the newest addition to the company's existing sales and support teams in North America, Europe and Asia. Satris was set up by a group of EDA veterans with a deep knowledge of the Israeli market, its customers and its needs. SATRIS' team is well known for its high level of technical expertise and also offers advanced design and verification services, providing Israeli customers with a full, high quality solution.
SpringSoft Completes Acquisition of Certess
SpringSoft, Inc. has completed the acquisition of Certess, Inc., creators of the Certitude™ Functional Qualification solution. The acquisition, announced last month, resulted in SpringSoft purchasing all outstanding shares of Certess, and the transaction has received necessary shareholder and regulatory approvals.
Certess is now a wholly owned subsidiary of SpringSoft USA, Inc. and its operations have been integrated with SpringSoft’s global development, sales, support and marketing functions. Certess products have been added to SpringSoft’s Novas™ family of verification enhancement solutions.
The acquisition adds approximately 20 people to SpringSoft, primarily in France and the United States. The former Certess R&D operation in France now reports to Yu-Chin Hsu, Vice President of Logic Verification R&D. The existing Certess field sales and support personnel have been folded into the SpringSoft global channel under Bill Shepard, Vice President of Sales.
Other EDA News
Other Embedded, IP & SoC News
You can find the full EDACafe event calendar here
To read more news, click here
-- Jack Horgan, EDACafe.com Contributing Editor.