You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I'm expecting a merger of the domains in the area of parallel processing and digital design, the level of abstraction above RTL is pretty much the same as what you need for fine-grained parallel software -
If that happens a chunk of the existing EDA market will become a general purpose software market.
EDA by itself is probably a shrinking market since the cost of spinning chips keeps going up, and fewer people can afford to do it. There's not enough money in EDA to keep the startup community interested in investing.
I think the problem EDA is solving is much more complex. Automating things like most commercial software versus solving complex problems. It is hard to compare. Do other software companies go through evaluations and benchmarks? EDA companies also shoot each others feet. It is fun to watch sometimes. Tell Cadence you are looking at the Synopsys equivalent tools and see what happens! Bang boom pow!
One huge difference is the EDA is always delivering software that is barely ready. It is simply not possible to test effectively EDA software for designing, say, 20nm designs without having any 20nm designs to test on, and no 20nm designs can exist without the software. This is a fundamental reason that EDA software is so buggy on the leading edge. Of course, there is plenty of non-buggy EDA software on trailing process nodes, but you can't do a 20nm design using software that was created for 90nm. This is simply not true in other enterprise software areas such as databases. Yes, databases get larger but they don't have fundamental technology shifts every couple of years.
I agree. EDA software is buggy at 65nm so it is not just new designs. My view is that we all do things differently and all variations cannot be tested properly. Even worse is when we add new software to get better design results and hit bugs. It seems by the time the flow is stable we are done with it. That is the curse of EDA.
whether it is EDA tools or Test equipment or anything else, latest generation 'tools/equipment' to produce the newest generation is always a generation behind. We have to create the future before we can use it. We as consumers are used to getting solutions that work without issues in the commercial world (enterprise sw as Paul mentions, iP* products, etc). But these are using established databases, formats, algorithms, etc and are not pushing the envelop any new changes. This is very different compared to developing SoCs, PCBs, and systems that are creating a new future. Many of my co-workers laughed at this situation...if we could only go into the future and come back with new tools/equipment that would make our jobs easier.
The problem a commercial software is trying to solve as well as the method for solving the problem is much more well defined and hence the main importance lies on the effective implmentation, marketing and customer support. But in EDA the if not the problem, the solution method is very vaguely defined and lot of effort goes into designing the algorithm before the actual implementation starts. This also make the interaction between customer (in this case design companies) and the EDA companies very crucial from the very early stage of software development.
