Five year wait predicted for 3d chips

1 min read

Stacked chips that use through silicon vias (TSVs) to talk to each other are unlikely to move into mainstream design before the middle of the decade, according to analysts and industry experts at last week's Design Automation Conference (DAC) in San Diego.

Mary Olsson, senior analyst with Gary Smith EDA, said: "It's very much like the overhyped multichip module market of 20 years ago." Olsson explained that the outsourced assembly and test (OSAT) companies have, so far, spent very little on equipment for dealing with this new generation of chip stacks in which connections between layers are made using contacts drilled through the substrate of the silicon. Currently, three fabs are capable of producing TSVs, but they are all 200mm lines, suitable for image sensors and other specialist products which use the technology today. However, there are no 300mm fabs, which would be used for high volume applications, such as devices for mobile phones. "Follow the OSATs and where they are spending their money," said Olsson. "They are spending it on copper wirebonders. The technology that we are hearing the hype about needs a whole new infrastructure." That is not being funded right now, she added. "The technology will probably not ramp up for the next three to five years. Too many other, safe, options exist." Mike Gianfagna, vice president of marketing for Atrenta, said the company is working with imec and customers such as Qualcomm on 'pathfinding' development for 3d ics. "It is an area we have chosen to invest in. But this ain't going to be mainstream until 2013/2014 at least. It will be a long time before it becomes mainstream for a lot of customers. But we think it will be a big market and we want to be first." Riko Radojcic, pictured, leader of design for technology initiatives at Qualcomm, said: "We think 3d is inevitable, especially for wide I/O and high bandwidth memory. It is becoming increasingly difficult to provide bandwidth: you can only increase the clock so far. The questions are when and how? But, one day, I will be able to say I told you so." Philippe Magarshack, vice president of design automation at STMicroelectronics, said: "We have not yet found a viable mainstream product to use with TSVs. We have been investigating when it makes sense to use wide I/O. But we will use LPDDR2 and 3 before we go to TSV. There are thermal issues and, potentially, cost issues with TSV. We produce high quality TSV wafers in Crolles for image sensors, but we have not found the sweet spot for other products." For more on 3d chips, click here