I would like to clarify the comments you attributed to me as a result of our conversation at IMAPS International. I do not believe I said that through silicon vias (TSVs) were a pipedream, nor did I doubt that they would be adopted. What I said was that they would not be adopted at the rate projected. That is why I referenced the flip chip vs, wire bond issue as an historical reference. I said that they would have their areas of adoption, but that I doubted it would be as widespread as everyone was projecting. I even talked about some of the roadblocks to their becoming fully mainstream.
I believe that the pipedream comment was related to the general belief that the TSV was going to allow for wafer-level stacking. Technically this is true, but my comment was related to the overall aspects of wafer-level stacking including the technical hurdles that must be overcome. To do this requires that each die on a wafer have the same size and pitch. This is a highly inefficient use of silicon since different devices have different sizes. If all the devices are to be stacked in wafer format, then every wafer will have to have the same die-to-die pitch on the wafer. This would mean that production would be driven to the largest device size to be stacked instead of the highest density possible on the wafer. That goes against the best cost model the industry uses at every turn. There will be a market for TSV type processes for device stacking but I doubt that there will be much in the way of wafer stacking. Even device stacking will require a considerable amount of effort to move beyond a niche application area in the near to middle term future.
There are lots of developments in this industry that have claimed to be the new beginning and arrived to great fanfare, both in product and process fields. Remember TAB bonding? What about bubble memories? There are many other processes that have been developed that are great for specific areas but have tremendous costs associated with them. They are accepted only because they have performance advantages that outweigh the cost impact. If someone can come along with an equal performance process at a better cost point, the industry would fall all over itself to adopt it quickly.
Don Ream, ASM Pacific Technologies