EDAC was kind enough to step into the void generated by Dataquest's exiting the EDA industry and organized the Sunday evening executive reception. The highlight of the reception was a panel combining representatives from four user companies and five vendor companies discussing different types of new chip design challenges to be encountered over the next two years. The panelists dealt with the subject in a polite, non-threatening manner, giving safe and known answers that did not challenge the audience to think differently about the known problems.
(See coverage here). The event was, unfortunately, predictable: it delivered a review of the problems facing the industry, a realization that designers face greater challenges than before, and that EDA vendors have yet to find the "silver bullet" that will provide a turkey solution to the issues.
The problem is that no one is willing to face the problem: it seems that recognizing its complexity it is too emotionally difficult.
In reality designers face a very complex problem, one whose existence most EDA vendors are unwilling to even admit, let alone capable of solving. We have reached a major inflection point in our industry and yet almost everyone is spending significant amount of resource denying its existence.
Developing a leading-edge design today requires the successful concurrent solution to five major issues:
- The impact of process physical variables on the accurate and reliable functioning of the design,
- The consequences of the design characteristics on wafer manufacturing (yield)
- The need to reliably integrate third party IP into a design
- The architectural tradeoff of software and hardware implementation blocks, and
- The financial analysis required to determine whether or not the product will be profitable
Engineers know that the best way to solve a problem is to deal with one unknown at the time, and in the past we have enjoyed such luxury. But now all five issues must be solved concurrently, and we are not prepared because in general we lack the education, the organization, and the determination required.
Most designers are trained in digital logic design but now they must deal with challenging analog characteristics of digital design: they must learn on the job, and learning is expensive.
We are just now beginning to realize that a successful development of a leading edge product requires a degree of interaction between designers, managers, EDA tools vendors, and foundry that introduces not just a new technical way of doing things, but, more importantly, a new business method.
We have all talked too much about the problem of third party IP qualification, verification, and validation. It is time to deal with the problem honestly and realize that, for the most part, we must solve the legal constraints that limit the free flow of information and that we must be ruthless in weeding poor suppliers from the industry. We must define "trusted partners" and find a way to equitably compensate them for their contribution to lower development costs.
I fully understand that doing some of these things is not only hard but that it will drastically change the way our industry does business. The signs that these changes are beginning are all around us, yet few people are willing to talk about them. The major EDA vendors, for example, have realized that they must form teams with their most important customers, that they must involve a foundry, or consortium of foundries in the way they develop their back end tools, and that it is time to admit, publicly and loudly, that the definition of ESL we have used since the beginning is fundamentally flawed and unworkable. We must stop to redefine "the system" to fit our parochial needs to put silicon as its most important component: silicon is one of the vehicles that take a design team toward a profitable solution, not an end. To pretend that system design is only "electronic system design" biases the problem to the point that it makes it more complex, not simpler.
And finally let us recognize that if we just continue to give conference attendees the same answers they have heard before, they will stop coming. Our industry must find the courage to see things differently, to admit that change, with all the indeterminism it brings, is required, and that not all system companies will find the latest process node the most efficient implementation vehicle for their products.