Guest blog: Sandeep Srinivasan

I love that last posting by exEDA_guy! Before as a customer and then in EDA, I had always wondered why the large EDA vendors (S, C, M) offered a collection of point tools that were able to talk to each other most of the time (after some effort) but lacked a common base, a platform if you will. Common infrastructure like a GUI. A good part of the reason why, is that the big EDA guys are organized in silos, with each product division responsible for their own P&L. Each pays for their own GUI team. On the other hand, with a common resource like a shared corporate GUI, each divisional VP will wonder if they're getting their fair share of the pie from their contribution. (Any resemblance to politics is strictly unintentional!) The good news is that like a lot of other software, EDA has moved towards standard platforms like Eclipse. In theory, different divisions of an EDA vendor could look into adding their product as an Eclipse plug-in. EDA users should make more noise about having a common entry point, like a single GUI where they can enter a design and then run simulation and verification, access IP in ESL or RTL, run HLS or RTL synthesis, DFM, DFT, floor-planning, etc.

Let's look at how other software companies are able to fund complex software development and still make a decent profit. In IT software development, software stack is the term that's used a lot. The top of the stack is a presentation layer, often Web-based. The middle tier is a business model layer, implemented in C/C++ or Java. The bottom layer is the database. There're similarities with an EDA product, e.g. a P&R tool. The presentation layer is Qt and Tcl. The business layer has all the algorithms. The database is either OpenAccess or proprietary datastructures based on LEF/DEF. You'll notice in the IT stack, all 3 layers have many vendors providing either opensource or proprietary solutions. Even when a solution is opensource it is not zero-cost. Example, HTML is a free standard but many companies buy Adobe software to build their web pages. Apache web-server is free but you still need to hire engineers to run it. Because of these multiple offerings, you can find cheap solutions with very rich features. Similarly, there are many database vendors, some non-free, like Oracle, others free, like MySQL. So a company mostly pay for the R&D cost of the mid-tier and integrate the rest. Why is this possible in IT but we don't see it in EDA? Database vendors (even MySQL guys) make a lot of money without any complex revenue sharing schemes. Adobe has no problems charge a lot per seat with their web development suite. On the EDA side, every tool has to have its own GUI and scripting. Many tools have poor GUI and scripting interfaces that make the tools unnecessarily painful to use. Every tool parses LEF/DEF and build essentially the same datastructures in memory. With OA this is better, but still lack of clear semantic standard tie up engineers' time to continuously hack the datamodel in response to broken DEFs/OAs from other tools. EDA companies perceive little added value in GUI and databases, just as a necessary expense to get a deliverable product. This is ok as long as process technology is evolving as Moore's Law predicts. No one cares about usability if this is the only tool that can tape-out my chip. But as more designs are falling off the bleeding edge the IT model becomes more attractive. If it turns out that a product can win by a smooth design data flow and friendly scripting and GUI, there'll be successful companies building just a GUI or providing consulting and tuning services to OpenAccess or something more 'GNU'.

You all should take a class in supply & demand. EDA gets what the market is prepared to pay. Customers get the prices EDA vendors are prepared to sell at. It's like saying beer is too expensive. Well, don't buy it if you think it's too expensive. Think EDA is too expensive? Build your own tools, outsource your design, use cheaper tools (Tanner EDA, Altium etc...) EDA - don't think you're getting paid enough? You could increase your prices, get paid less, reduce your expenses. Expand into other segments. Come up with new ideas. For goodness sake - you are not creating the IP. You are a tool that companies use to create IP. You are like a hammer that gets used to build houses - should the hammer maker get paid more because the houses are sold for such high prices? Nope.

I'm afraid that (dick_freebird) has mostly 'hit the nail on the head'. A very large portion of EDA code, maybe 90%, is just the foundation code to present the user interface for schematic capture, board layout, project management, and file format conversions. If the EDA companies could support an open-source EDA development platform then they would be freed of most of that 90% code maintenance expense and be able to concentrate on what makes them unique. EDA companies should consider a business model that reduces their costs by 80% (common core code) so they can concentrate all their efforts on the high-end customers.

The author says that EDA vendors are underselling and designer (dick_freebird) says that they are overcharging!! You are saying that there is no cost of production: who pays those engineers, the tech pubs, the software engineering guys, the software tools that EDa guys use, the machines they need? People often have a very rosy notion of GNU projects. Good luck about that! I am preparing for GNU chips :)

As an IC designer I have always rebelled against the greed of the CAE majors. I see continual churning of the tool set in an attempt to extract more rent, despite that for the kind of circuits I do there is not any value added by new releases. The fact is that for anybody working south of the bleeding edge, tools are priced out of proportion to value added. When vendors add more features, my design cycle does not compress because CAD and design management come up with new tool-centric tasks to take back any time savings I might have gained. I keep waiting for a GNU framework to arrive, and a few stable back-end verification point tools, so that we can walk away from these commercial CAE parasites. Unfortunately there is one area where the CAE companies have perfected a skill - that of sowing FUD in the minds of corporate CAD directors, so that they will sign any purchase order to avoid blame for not-the-absolute-best tool selection. My needs are simple and better served by stability than innovation for the sake of market share, because the open spaces for innovation that remain, are not close to the heart of the matter. The more appropriate question is, why do semiconductor design tools not follow a semiconductor pricing profile? The answer I see is, software companies will do anything they can to churn the user base and prevent price compression despite that 99% of this release, equals last release and there is roughly no cost of production. And any cost of maintenance revolves on their continual release of not-entirely-debugged new products so things can never get to zero maintenance. Keep looking for love, because it's going to take y'all a while to find.