As the Baby Boomer generation gets older and medical science becomes more advanced, the possibility of longer lives for all of us becomes more real every day. We are all aging, but the ultimate price for that process is moving farther away. The difference between the process of life and that of death is pretty clear to most of us.
Why then, when we talk about the electronic design automation (EDA) industry, do we mix up these two fundamental processes so badly? I have spent most of my adult life being part of the EDA industry. I watched it be born in the 1980s and mature through 1990s until today. Industry segments are like living beings in that they are born, mature and then ultimately die, to be replaced by new industries. EDA is in roughly its fourth decade of existence; not even middle-age yet. To be clear: EDA is NOT dying " it is simply maturing.
The "older" parts of the industry do show their age first. The back-end tasks of logic synthesis, place & route and physical verification for example are entering a mature stage. There is consolidation in this part of the market as best practices converge. The shrinking number of ASIC design starts contributes to the process as well. Just because one part of the industry is older, the inductive leap seems to be that the whole industry is dying.
This line of thinking completely misses the mark. While ASIC starts are declining, the complexity of the remaining designs is growing dramatically. These chips also target a broader range of problems to increase ROI if you are spending over $100 million to design one of these chips, you had better find at least a half billion dollar market to attack. FPGA use is climbing steadily, filling the gap left by less sophisticated ASICs and ASSPs. The electronic content of virtually all consumer products is steadily growing. In many cases, the chip defines the system. If you doubt that statement, read any analysis about Apple's new iPad. It won't take long to find a spirited discussion about the merits of their custom A4 processor chip. If you look closely, you will realize a lot of EDA is needed to fuel all this. But don't look down; look up.
The physical implementation challenges for all this advanced silicon are daunting I won't deny that. But the growth opportunity in EDA is not in new techniques to implement these chips. Rather, it is in new and better ways to design and use these chips. The reuse of semiconductor IP is a fact of life for any advanced design group. Stitching together heterogeneous processing elements, myriads of compiled memories, multiple mixed signal blocks, high speed I/O, RF front-ends, all potentially from different vendors, is anything but simple. How do you ensure you've got the interfaces correct? How do you know the IP is actually suitable for the intended application? How do you verify the embedded software in context of the hardware architecture? And how does the whole thing harmonize to run at dramatically lower power consumption?
All these questions can be answered by advanced EDA software. These tools don't look like traditional EDA, however. They focus on synthesizable RTL descriptions, architectural definition and hardware/software co-design. This is the new frontier for EDA, and it is ripe with opportunity. Over the past year or two, there have been many new EDA companies formed. In spite of the horrific investment climate, somehow these companies find a heartbeat. Virtually all of them have focused on architecture advances, RTL optimization, design process improvements or IP authoring and integration strategies.
It is unreasonable to believe that the current technology revolution in the consumer market will continue while EDA dies. Let's agree to stop predicting the death of this important business and rather debate how it will mature (and grow).
Mike Gianfagna is vice president marketing at Atrenta Inc. (San Jose, California)