This interview originally ran Aug 11, 2010 in EETimes online at http://www.eetimes.com/electronics-blogs/other/4206103/Interview–Paul-van-Besouw–Oasys-Design-Systems
I recently had the opportunity to sit down with Paul van Besouw, CEO of Oasys Design Systems
, and interview him on lessons learned from his entrepreneurial efforts at Ambit and Oasys. I have added hyperlinks where I felt they would provide context.
Q: Can you talk a little bit about your background?
I am a founder of Oasys, along with Johnson Limqueco, vice president of R&D, and Harm Arts, CTO. We were part of the core R&D team at Ambit Design Systems, which was acquired by Cadence. At Cadence, the team focused on physical synthesis, connecting traditional synthesis to physical design.
I have a Master of Science degree in Electrical and Computer Engineering from the Eindhoven University of Technology in The Netherlands.
My entry into the EDA industry goes back to when I was at the University of Eindhoven. I was part of a group doing EDA research and I got an email from Rajeev Madhavan, who was starting Ambit. He was looking for software developers who had experience developing RTL synthesis front-ends. I was just finishing my Masters degree and trying to figure out what to do next, so the timing was perfect.
That was 1995. I talked to Rajeev Madhavan and he invited me to come out to California. I packed up my suitcase and came over, and that is how the whole thing started. It was initially meant to be a six-month project but it turned into 15 years.
Just coming out of University with no industrial experience, it was a great opportunity. There really was nothing there. Everything had to be built from scratch, and it was a great learning experience. We had a small core R&D team of about six that built the foundation of the Ambit technology.
Q: Can you talk a little bit about what led you to found your company, what was the problem that motivated you?
By building synthesis technology at Ambit and separately building physical technology at Cadence, we realized that we were forcing design teams to deal with a suboptimal solution. We saw how hard it was to integrate synthesis and place and route together in a way that produced the best quality of results. Traditional synthesis turns RTL code into a quick-and-dirty netlist and then runs a powerful but slow optimizer on that netlist. Our goal is to produce the best starting point for physical implementation by producing placed netlists that in turn enable the place and route tool to deliver the targeted quality of results.
It took two years to build a new chip synthesis technology from scratch, we call it RealTime Designer. The first prototype solution was qualified on an actual design and the engineer’s reaction was priceless. After they had verified the design and validated the quality of results on site they said: “Holy smokes! The run really did take less than 15 minutes.”
Q: Can you give me a brief overview of where the company is today?
RealTime Designer has proven itself in dozens of benchmarks and is in use in production flows. For example, Renesas in Japan has been working with Oasys for more than two years and its chief engineer considers it to be the most advanced synthesis technology. In fact, he thinks that it has the potential to change the usage model.
Design teams today are more sensitive about letting their company names be used as a reference. Most of the design teams we are working with have single vendor relationships, and several of these customers have EDA CEOs as their executive sponsors. That makes it tricky for them to come out and say anything about another competitive technology.
Just before the Design Automation Conference this year, we announced that Juniper Networks had selected RealTime Designer for the design of its next-generation networking chips. Juniper thoroughly evaluated RealTime Designer and concluded that it offers high-quality results and performed well in the Juniper design environment. Juniper approached us because the runtime of the existing tools was way too slow and took forever to get from RTL to GDSII. Debashis Basu, vice president for Silicon Development at Juniper Networks, gave RealTime Designer high marks, noting that it is a great tool that fits a very real performance need.
We followed this announcement with the news that we signed a multi-year strategic licensing agreement with Xilinx for our Chip Synthesis technology. Vin Ratford, Xilinx’s senior vice president of worldwide marketing, said that with programmable chip sizes growing and complexity mounting, Xilinx needed to look at a new generation of synthesis to support the needs of its customers.
Q: It has been reported that Xilinx is an investor in Oasys? Is either Xilinx or Juniper an investor in Oasys? Are any of the executives of either firm an investor in Oasys?
Juniper and Xilinx are customers and not investors. Neither has taken an equity stake in Oasys. None of the executives at Juniper or Xilinx is an investor in Oasys.
Juniper licensed our RealTime Designer software to use in its design flow. Xilinx licensed the Oasys Chip Synthesis technology and is modifying it to support FPGA designs; it will have the Xilinx brand.
Q: What are the two or three things that you have been able to accomplish that you take the most pride in or satisfaction from?
We created a new technology with a small team, and little funding. At first, we were completely self-funded. We rented a small apartment where we spent about a year just coding everything from scratch. Later, we received some seed funding from several EDA-savvy angel investors, which allowed us to move into a “real” office.
We had a working prototype by 18 months to show other angel investors, which allowed us to secure a bit more funding. I was able to attract the attention of some of the best people in the industry. It took some convincing, but I was also able to attract Joe Costello‘s attention. He is now a member of our board of directors.
Joe’s keynote speech at the 43rd Design Automation Conference in 2006 outlined three rules for building a successful company in EDA. The rules talk about how to “think like a fish,” and that you should “write your press release first,” and to “fundamentally change the rules.” It was around the same time that I started my discussion with Joe. I talked him through the technology and how we did things. He definitely saw the potential of this kind of technology where it could fundamentally change the game rather than trying to play the game, which is basically what we had done at Ambit.
Q: What has been the biggest surprise? What was one key assumption you made, perhaps even unconsciously, that has caused the most grief?
From the very beginning, the team was focused on building a better implementation technology, and on getting better quality of results through place and route. We expected to improve the capacity and runtime as well, but really did not expect such a huge improvement. I think this must have been the biggest surprise, although a positive one.
On the negative side, the biggest surprise was when the economy weakened so much. It has made the funding environment challenging, to say the least, and has also caused customers to become cautious about considering new tools.
One big challenge is that we are competing in a mature market. That sets the bar extremely high since design teams only start to become interested once the tool is better than the incumbent technology in pretty much every respect. And, not just incrementally better, but compellingly better.
If you go back to Ambit, the motivation for design teams to buy Ambit was different. Synopsys had a monopoly. Ambit had a good technology, but it was more a “me too” implementation. We had the advantage of doing it from scratch and were able to do it a little bit better than what Synopsys had done at that time. Clearly, one motivation for design teams to go with Ambit was just to have a competitor in the market.
Ten years later, it is more difficult to innovate and come up with new technologies. Nobody got fired for going with Synopsys. Adopting a new technology in a critical path of your design tapeout is a big deal.
If you start a technology company and are able to compete in the market today, it is all about building a complete solution. It is one thing to have a promising technology that shows some good results on one or two designs. You have to be able to build out a technology with all the bells and whistles and still show that advantage over a large number of designs.
Q: What development, event, or new understanding since you started has had the most impact on your original plan? How has your plan changed in response?
Venture capital for EDA is pretty much non-existent. This was a new reality and we were forced to do things differently. We are working with less money and fewer engineers on a longer development time-line than we would have if we had started Oasys 10 years ago.
Nobody embarks on a startup thinking it will take six years just to get to market, but that was the reality we faced. On the other hand, it did allow us to focus on maturing the product before announcing the product and/or company. We were able to close some significant business, which is driving the company right now. We continue to focus on building our customer base and growing the business based their success.
When we started, we had a different perception than we have now. With my background from Ambit, I knew what it would take to build things from scratch, but we were still counting on the funding to be there. In hindsight it was a blessing because we were forced to do more with less. It was a core team that worked on the technology from the ground up rather than having some ideas, build something and then have a big team work it out. In the end, it has helped the technology mature in the way it did because it was a technology completely different from what we had ever done before.
However, it definitely took longer to develop the technology because of a lack of funding. It took us three years to get a prototype to work and to get to a point where we could start engaging with some design teams. However, when we did, we realized that we had something different that was even better than we had imagined.
Q: Any other remarks or suggestions for entrepreneurs?
Starting a company sounds glamorous but it is hard work and takes perseverance. The two pillars that allow the company to go through some of the tough times are the support of your family and a founding team that works well together.
Obviously the founders have to believe that what they are doing will make a difference. They need to learn to balance out the highs and lows to keep the team motivated.
I have been very fortunate in being able to key surround myself with great people. Oasys has extremely knowledgeable people as investors and on the board of directors. It is easy for a founding team to get absorbed by the technology, so it is important to balance it out.
What makes EDA both interesting and challenging is that it is not only about the software. In the end, you are building software to build hardware. You have to start with insights into both and learn a lot more along the way. In many cases it is the experience of what does not work that really allows you to focus on the things that do work. EDA software is built on a technology foundation surrounded by algorithms. Starting out, a lot of time is spent on finding out what does not work. There are many details that need to be incorporated to enable your technology to work in an actual production flow.
Starting with a great technology is not sufficient.
Q: Thanks very much for your time.