中文 English

Does EDA Sell Fear?

There are plenty of complaints across the industry. Most of them are misguided.

popularity

I worked in the EDA industry for over 30 years and a common lament I heard was that the EDA industry survived by selling fear. Your new chip will fail if you do not buy the latest tool offering. There always seemed to be a natural dislike for the EDA industry and many users thought the industry overcharged and was unable to innovate.

I never quite understood the reasoning. A recent comment, while doing research for one of my articles this month, reminded me of the basis for it. It made me rethink and to bring together both sides of this issue.

First, lets get a few things straight. The EDA industry has been a very stable industry, and Wally Rhines, the CEO for Mentor for many years, used to publish numbers that showed a remarkable flatness in the curve for how much EDA contributed to the cost of development of a chip. It was around 8%. Given the increase in complexity of chips over the past few decades, that indicates an incredible increase in success and development productivity.

Did EDA companies gouge their customers? Far from it. The amount of revenue that was plowed back into R&D by EDA companies far exceeds almost every other segment of the high-tech industry. It was the true back in the early days and remains true today.

Do EDA companies leave all of the testing to their customers? Well, this one is partially true. I can remember releasing new tools back in the ’80s that had barely run successfully on a few concocted test cases. Today, higher quality metrics are in place, but EDA still suffers the same challenge that we had back then – EDA does not have much in the way of cutting-edge designs that are equivalent to those that the tool will be used to assist with. It is thus impossible for them to fully test software before release.

EDA companies cannot innovate. To address this, we have to look at where innovations come from and the role of EDA companies. EDA companies have tools that fall into a couple of major categories — those that provide productivity and those that ensure success. The roots of innovation for these two categories are different.

Productivity. This comes from automation, and that can only happen when a task has become well understood and stable. Otherwise, the return on investment for creating the tools is too low. The number of potential buyers for an EDA tool is low to start with, and so automation that focuses on a small subset of that community will not succeed. This is why every attempt to create system-level tools has failed. Each company, or project team, has a single architect, and that phase of the design is over quite quickly. The returns are too small for the necessary investment.

Most of the algorithms used are generic in nature — things like simulated annealing, for example. They are developed within universities and often the applicability of them to certain problems, such as EDA, is demonstrated. The role of EDA is to take those algorithms and find the best ways to apply them to real-life problems. This often involves some form of abstraction, or simplification, to make the problem tractable. Having seen many tools that comes out of academia, there is a lot of work necessary to produce a commercial tool from that proof of concept. Is this innovation? There is certainly some in there, but little of it is ground-breaking. It’s just a lot of hard work, and that has to persist over a long period of time to keep expanding it to handle new use-cases.

Ensuring success. This is where the fear factor comes in, and in the past, it was not the area in which EDA innovated. That was indeed the role of their customers. Cutting-edge design companies would discover problems long before the rest of the industry. They would identify the problem, work out methodologies to deal with it, and often develop tools to help mitigate the problem. Ultimately, they would give, or sell, those tools to the EDA companies because they did not want the long-term maintenance associated with them. EDA companies can amortize those costs over many customers.

The EDA companies then would go out and “sell” the need for these tools to the next tier of design companies – often companies that had not yet experienced the problem. It would only be their next design that could potentially suffer if they did not incorporate this new step in the flow. In that way, it was seen that EDA companies were selling fear, but that was not accurate. Those ahead of them had really experienced those problems, and this was a way to avoid them.

Today, that dynamic has changed. Some of these problems are too complex for a single company to understand and solve. We are seeing extensive collaborations between design companies, fabs, IP developers, and EDA companies to collectively understand the emerging problems and for them to come up with solutions together. This is innovation at its finest. Without it, few chips would actually work today.

So perhaps EDA companies are the purveyors of bad news. Their message is that if you want to continue to innovate in design and implementation, you have to learn from those ahead of you. They are the ones who have been tasked with spreading the knowledge in the form of methodologies and tools. EDA companies are playing a direct role in the innovation that allows the industry to progress.



3 comments

Harry Chen says:

Having worked in EDA (10+ years) and fabless design (20+ years), I really enjoyed this article. It succinctly captures how EDA capabilities evolve and the role of EDA companies, matching my own experience. Well done!

Dean Freed says:

This was easy enough for a simple country lawyer to understand. Thanks, Brian.

Siddhartha Nath says:

Having worked in the EDA industry for 10+ years, I don’t think this is entirely a general thought shared largely among designers even in the top-5 semiconductor design companies. EDA companies constantly try to improve their tools (modulo the resources).

Leave a Reply


(Note: This name will be displayed publicly)