October 22, 2007
Can a Firm Prosper or Even Survive, If It Gives Away Its Product?
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.
have to do that by hand. That is going to be the case as we go to automate analog in the custom world. I don’t think we are going to replace the creative, innovative work of a brilliant analog designer. We are going to take a bite out of the stuff they have to do in order to get that brilliant idea into silicon. There is a certain amount of drudgery that has to be done. I think the industry has an opportunity to automate that.
I understand the greater productivity that PyCells have relative to drawing by hand but doesn’t Cadence have similar capabilities based upon the SKILL language for PCells.
The nature of software engineering has changed a lot in the last two decades. We are in the era of Web 2.0. Twenty years ago nobody conceived what the Internet was going to be. In the early eighties Microsoft Basic was probably the most popular language in the world although most sophisticated things weren’t done in it. When I was in college in the early eighties there was Pascal and things like that. We are just eons ahead of that in terms of modern object oriented software engineering. There is something we can learn in the EDA industry from the cutting edge work being done on the Internet and other kinds of sophisticated applications, financial engineering software and so
forth. In the EDA industry we are good at algorithmic technology but in terms of raw software engineering, the world has moved faster than a lot of parts of EDA. One of the opportunities and this is not the core focus of PyCell, relative component. We have been able to take advantage of that. If you compare and without wanting to do a blow by blow comparison of one language against another, 2007 versus 1987 is light-years of difference. When some of those early languages came out people were driving DeLorens and listening to Duran Duran. It seems like a long time ago.
Ciranova faces a couple of challenges. Whenever I talk to EDA industry executives, they tell me if the use of their product is transparent to existing customers’ design flow, just plug and play, then it is a tremendous advantage. If the prospective customer has to change its design flow, there is considerable resistance. A 10x simulator that takes the same input and produces the same output has a short sales cycle. Wide spread adoption of a new language such as SystemC or Python takes considerable time. How will Ciranova address this issue or am I overstating the problem?
I think you have articulated it well. That is one of the basic challenges that everybody grapples with. It is not different from should I buy a cell phone today or should I wait 3 months for a better one. What you have here is a situation where if I switch to something, it is going to take a certain amount of effort and learning and so forth in order to get there. The question is “Is the thing I am going to switch to enough better than what I’ve got today that it justifies my making that switch?” That is the same for every design flow, every aspect from mechanical to civil, software and so forth. That paradigm is repeated everywhere. I understand it is our job to make sure that we add greater and greater capability to the design process in such a way that we minimize the effort to move over. One of the things, for example with PyCells, is that you can have Cadence SKILL PCells in the same design. So you do not have to choose one or the other. It is really important. If you have an infrastructure you are comfortable with and it runs perfectly well on traditional SKILL PCells but you have one device and every semiconductor company has a few devices that are really hard to model in PCells, you can write a PyCell for that device and use it together with all your existing PCells. You do not have to change your whole environment overnight
to do that. You can adopt PyCells as fast or as flow as makes sense for your environment.
A different view of the same issue. People tend to have budgets for software but possibly not for the support of free software. Is this an issue for Ciranova in getting support contracts? Do you sell direct?
We do. So far we have not seen this to be an issue. It is important to put all of the stuff into context. Most estimates of what it costs to develop a 65 nm chip are many tens of millions of dollars in development. If you are going to make this kind of investment in a silicon chip, for most people support for their infrastructure is not going to be the straw that breaks the camels back. People are looking for ways to be more productive on very complex and expensive chips. If we can add value in that area, the price of service has not cropped up as an issue.
The typical consumer of electronic products, your next door neighbor, buys something. He is more likely to get a service contract or extended warranty as the cost of the product goes up e.g. large screen plasma TV versus a CD player. By analogy won’t it be challenging to get people to pay thousands of dollars to support a free product?
We have not found that to be an issue at this point. It may not be the right analogy. My wife never bought a service contract on a consumer product in her life because they all come with a year’s warranty. I’m not sure it is a good model or not. So far it has not been an issue.
In the software industry there is usually some ratio between the price of software and the price of support (upgrades, bug fixes, phone support), typically around 15%.
This is becoming an increasingly accepted model in the industry. Go back to Red Hat Linux. A lot of software companies have developed very large service organizations, all the way from small firms to Oracle and IBM. The latter have developed very large service organizations not that their software free. People are accepting that service and support as a component of the software business.
I worked for IBM. IBM Service offers to take responsibility for everything, namely hardware, software, facilities and even your IT personnel.
Without wanting to go into excessive detail, I think that’s a model you may see Ciranova doing work at some point as well. The whole problems of PDKs and PCells, most customers just wish it would go away. It is not their value added. Designing a better PCell typically does not mean the difference between winning and losing in the market. People are looking for solutions where this problem can be robust and not be an issue that distracts them from where their core value is. Basically, people look at PCells and transistor models and say that’s where our core IP is.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Jack Horgan, EDACafe.com Contributing Editor.
Be the first to review this article