Which comes first, 3D standards or market adoption?

The more I talk to people involved in bringing 3D integration using TSV to market, the more it’s clear that the standards issue is another chicken-and-egg, cart-before-the-horse conundrum. Yes, eventually the industry will need standards, but can they really be set before the processes are established?

While some say that standards need to be in place for market adoption to occur, others say that market adoption will happen, and standards will follow suit to bring the technologies to volume production. What's important to understand here is that 'market adoption' and 'volume production' are not synonomous. The former hopefully leads to the latter, but doesn't have to. A technology can be adopted to the market, but only serve niche applications, correct? In any case, I’m told that first and foremost, there needs to be clear definitions of 3D integration schemes and roadmaps before standardization can even be considered.

Most agree that standards development for certain elements is necessary for 3D integration to reach volume production, they also note that not everything can be standardized. For example, I’ve heard several comparisons with flip chip technologies, noting that although flip chip has been used as a method of interconnect in volume production for some time; there are no standards for the process. Point taken – but then flip chip cannot be compared to the entire 3D integration scheme, but rather only to TSV as a comparable method of interconnect. TSVs exist in many forms, as we discussed at length last week. But it’s only one element. Therefore, it’s not the processes themselves that will be standardized; only the interfaces and interface dimensions, for example the interface of memory to logic in a 3D IC configuration. Additionally, design rules will be needed to address the gaps between front-end and back-end players.

Speculation as to who will set the standards varies from industry organizations like JEDEC, Jisso, ITRS, and SEMI; to the major foundries and IDMS like TSMC, Intel, and IBM. Ultimately, the task will fall to whoever stands to benefit most by investing the necessary resources. Really, it appears that developing standards will be an evolutionary process, and is likely to occur in parallel as 3D integration moves from market adoption to volume production. – F.v.T