Universal Truths

Sometimes big ideas come from small companies

by Peggy Aycinena

Michael Munsey is Vice President of Marketing at Silicon Dimensions, Inc. The company’s a new player in the EDA space, just announced in January 2004.

I had a chance to speak with Michael by phone some weeks ago. We had such an interesting conversation that I told him that I felt I could extract some of the Universal Truths about EDA out of the chat. He seemed delighted. Subsequent to that phone call, Munsey has learned a Universal Truth himself - one of the first laws of journalism. That truth? No matter how interesting the project (i.e., article), it can take 6 to 8 weeks longer than first announced to take the thing to completion.

So delayed as they may be, here are some Universal Truths about chip design and the CAD tools that facilitate the process as expressed by Silicon Dimensions' VP of Marketing Michael Munsey.

****************************

** The number of tools in the flow is a direct result of the complexity of the design.

As the designs become more complex, the number of tools required increases accordingly.

** It’s no longer possible to do flat designs.

Designers have to partition projects into blocks; there’s just no other way. But that means you’ve got a different physical design associated with each project block. But, the reality is that there simply aren’t enough physical designers on any team to have one per partitioned block. Hence, you have a situation where logic designers are, by necessity, having to deal with physical design on their own.

** It always requires a couple of iterations from logic design to physical design.

No matter how clever the tools, there’s no such thing as first-pass success. The traditional flow has the logic design moving to the netlist, moving to the physical design, the trail placement, determinations of the routing specification files, and then circling back to the logic design.

(Silicon Dimensions provides block design and analysis planning tools that help to simplify this process - not eliminate iterations, but reduce the number of iterations. We’re dealing with reality here.)

** Traditionally, there have always been far fewer physical designers on a project team in comparison to the number of logic designers.

The typical ratio is one physical design guy for every 10 logic design guys. This situation is exacerbated by the fact that physical designers are very expensive in comparison to the costs of your typical logical designers. So there’s a maximizing of the capital required for design projects that revolves around minimizing the physical designer’s work and maximizing the efficiency of the logic designer’s work.

(Silicon Dimensions tools offer a capability that allows some of the work of the physical designer to be moved up to the task level of the logic designers. This saves on the cost of the physical designer, plus it makes the physical designers work more efficient once it is finally under way.)

** The biggest issue in design today is communication.

With time, the design process has become further and further specialized. Even as recently as a couple of years ago, you would see smaller design shops at DAC where they had the logic design and the physical design being done by the same guys. But as we’ve moved from 0.25 micron, to 0.18 micron, 130 nanometers and 90 nanometers, the functions have become more and more specialized. At the same time, all of the EDA vendors have been developing more and more tools to meet the growing needs. Now communication is the problem - due to the current complexity in the design flow. You’ve got engineers who are using specialized design tools, but they’re having a huge amount of difficulty handing off their results to the next guy in the flow.

(Therefore, what Silicon Dimensions is concentrating on is the issue of communication.)

** Intense technical dialog is absolutely essential across the entire design team in order to achieve design closure at 130 and 90 nanometers.

The logic designer doesn’t/can’t/won’t see the results of DFT circuit insertions or the results of too much fan out, unless the effects of those design choices can be made available immediately to the logic guy. But you can’t let the logic designer resort to the easy way out, can’t let the mindset be - ‘This isn’t my problem, it’s somebody else’s problem - or designs will never be completed within the most timely fashion.

(Silicon Dimensions has sat in on many postmortems with their customers. Every customer tells Silicon Dimensions that what they really need is the onset of important cross-team conversations earlier on in the design process. Early conversations seem to be the only way to avoid long-term, nightmarish problems. For instance, it’s true that DFT tools have done a lot to make diagnostic processes easier after the fact. But, there’s always the critical pass where the logic gets inserted, and immediately the DFT tools have to be constrained - just like the synthesis tools. But there’s need at that moment as well for human intervention or else things can slip through. Conversation, early and often, is the only way to avoid the pitfalls of these processes.)

** It’s all about who controls the data.

Fundamentally, the project manager owns the data on a design. The tacit responsibility of that manager is to deliver a product in a timely manner. In order to do that, they have to divide up the project. Each sub-contractor on the project owns a piece of the process. Therefore, it’s up to each engineer and/or engineers within each of those ‘sub-contractor’ teams to optimize their design process within their own section. The Logic Designer, for instance, has to be able to do the best job with the knowledge he has available to him within his own space. But, there’s a problem here. People have developed a great deal of tunnel vision - they can’t see beyond their own sector - and the design process has become awkward, ungainly, and dangerously inefficient.

(Silicon Dimensions has developed tools that allow the project manager to have the kind of control needed to optimize the efforts within each of the ’sub-contractor’ design groups, while simultaneously optimizing the overall effort of the larger design team.)

** These days, people are selling a lot more back-end tools than front-end tools.

... Synopsys, Cadence, Magma all included. That’s where their revenue is. Why would anybody making a play in the EDA space do something to circumvent that situation? Anybody wanting to make a play in EDA should be addressing the front-end tools. You can choose to invest a lot of capital in convincing customers to hire more physical designers, and then to outfit those designers with new tools, or you can offer tools in the front-end that help pipeline the job and add more design time and flexibility into the projects. Everybody’s dealing with constrained resources these days. Business and technology decisions should be made that reflect that reality.

** Among the top concerns for physical designers is that others not be allowed to encroach on what has been their traditional space within the overall design process.

But physical designers can be convinced that others are not encroaching if they can be shown that efforts up-stream to their efforts are actually beneficial to the long-term effectiveness of their efforts.

(Silicon Dimensions tools are not trying to eliminate the physical designer’s job. They are trying to make the physical designer more efficient by asking more of the logic designers. The tools are making the overall process more efficient, therefore allowing the physical designer to concentrate on what he does best - routing, clock-tree inserting, surgical corrections of the netlist, etc.)

** Tools should be intuitive, and should allow designers to stretch themselves a bit.

A company that is founded on the fact that logic designers can do their work more easily if they’re not required to handle things outside of their sphere of expertise is a company that’s founded on the false premise that more can’t be asked of logic designers, even those with skillsets that are limited within traditional boundaries.

(Silicon Dimensions tools are not asking logic designers to actually use and/or implement the full suite of back-end tools. Instead, the Silicon Dimensions tools eliminate the complexities, both in use and intuition that might make it impossible for logic designers to actually help, rather than hinder, the overall design process.)

** Universities are great proving grounds for tools.

When tools are made available to universities, not only do the students benefit from having access to software that simulates a real-world work situiaton, but the vendor providing the tools also benefits from the feedback.

(Silicon Dimensions is talking to a couple of different universities about having their tools incorporated into the educational environment. Silicon Dimensions is providing sufficient graphics data to the student-users so that they can see the output of what their synthesis should look like. Students can do logic traces and put that data right back into their logic designs.)

There are unintended consequences of making tools available on less expensive computing platforms, the principle results being that more people use the tools and the strengths and weakness of the tools are more quickly articulated by a larger population of users.

(Silicon Dimensions tools run on the most recent versions of RedHat, and will soon have a 64-bit implementation to further enhance the capabilities offered to users.)

** DAC can be a wonderful venue for small EDA companies.

Even when attendance is down, as it has been over these last several years, everybody has always said that the right people come to DAC nonetheless - people who have the power to purchase the tools. As the economy improves, the actual end-users are having, once again, the power to make the decisions about which tools are purchased.

(Silicon Dimensions had a great DAC and was delighted with the turn out at their booth. They’ve grown sufficiently at this point, also, to open an office in San Jose to augment their office in Marlboro, Massachusetts.)

When a company focuses on a particular market segment, they can develop sophisticated technology to address that one segment and succeed.

(Silicon Dimensions has been successful, per reports, in convincing customers that theirs is the unique way to go - principally by focusing on logic designers and the tools those guys have been lacking in their work. The Silicon Dimensions software offers designers the opportunity to know what they hadn’t previously known about the affects of their design choices on the process down stream.)

** Deciding what level in a customer’s’ organization to sell tools into depends on where the EDA company is in its own life cycle.

EDA vendors that have focused products should be selling to the end-users. That happens by developing internal champions for your products in each customer account. If you can, develop a champion that will act as an internal advocate for your technology and that can advance the belief among fellow employees that your technology will solve the issues facing the team and their management.

(Silicon Dimensions offers a 30-day evaluation key that potential customers can pass their own data through. In a day or two, those people become strong advocates for the Silicon Dimensions technology and demonstrate to their co-works that the tools can show them things they were never able to see before.)

It doesn't get any simpler than that.

****************************

August 5, 2004

Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com

Copyright (c) 2004, Peggy Aycinena. All rights reserved.