Neils Bohr, the physicist (or should that be the quantum mechanic) famously said "Prediction is very difficult, especially about the future." The semiconductor industry, despite its incredible complexity, is easier to predict than many other industries because so much money is invested in making it predictable, and the ecosystem is so complicated that it has to all move together. This time of year is traditional for predicting the year to come, so here are a few of mine. System Design Enablement There will continue to be a move towards system companies taking control of their destiny and designing their own proprietary silicon, and semiconductor companies building entire software stacks. A key driver of this will be software-driven verification especially using emulators such as the recently introduced Palladium Z1 along with fast processor models. See my post earlier this month (and my post on Palladium Z1). 3D Packaging Finally Goes Mainstream I think that this year will be the one in which 3D packaging technologies finally go mainstream and enter volume production. At the high end, this has partially happened using silicon interposer and TSV technologies. But new consumer approaches are coming and my expectation is that hundreds of millions of designs will ship next year, creating a volume ramp that will take the technology mainstream. Check out my post about this technology. Designs Spread Across Nodes It is clear that for a number of reasons, many designs are not going to move to the most advanced FinFET processes any time soon. This means that the processes used for design will become a much richer menu. The most advanced design groups in the highest volume industries are already doing 10nm designs and they will start to ship. But old processes are being re-architected to have variants for lower power, for lower cost, and non-volatile memory, based on all the experience gained since they were originally introduced. And various flavors of FD-SOI will become important due to body bias. See my post from November. Automotive It will be big. The electronic content in cars has risen explosively. The structure of the existing automotive industry (with the exception of Tesla) reflects the hierarchy of the vehicle. But that hierarchy is going to change and the skills necessary, such as image recognition and video processing barely exists in the current OEMs and Tier 1s (that's what you and I would call car companies and their major suppliers). RS26262 and the requirement that automotive electronics can run self test regularly while the vehicle is in use will drive changes in how test is done, which will probably bleed out into other industries. See my recent post on Automotive Ethernet, for example. Democratization of Formal Verification Formal verification used to be done by PhD-level engineers using the engines directly. Now many use cases have been encapsulated as Apps, introducing formal as a complement to other verification approaches, in particular simulation. See my post on the Jasper User Group meeting. CES Will Be Full of USB Type-C Equipped Products Everything you read is that the USB Type-C connector is the fastest growing connectivity standard ever. I even wrote a post about it. There are even rumors that future iPhones might abandon the current Lightning connector for it and even remove the headphone jack, as they did to all the legacy connectors on the current Macbook Air. Test Test has always been the orphan step-child of EDA. But I think we will see an increased focus on it. One big driver for this is automotive as I mentioned above, but increasingly it is costing more to test the die than to manufacture it. I expect to see a big focus on getting the cost down and coverage up.![]()
![]()
↧