Moore's Law - the Z-dimension: A Decade Later - 2009

By Sergey Savastiouk

Almost 10 years ago (January 2000) I wrote an article for SST titled "Moore's Law -- the Z-dimension". The call was to shift focus towards Moore's Law in the z-dimension and then invest in affordable, vertical miniaturization and integration, rather than continue to invest in further feature size reduction. A new term -- through silicon vias (TSVs) -- was first introduced in that article and is now commonly used. Then, the expectation was that the acceptance of TSV technology would be faster than what has since happened. Drawing parallels with the history of flip chip technology may explain why it's happening slowly.

Analyzing the history of flip chip technology may predict the future of TSVs. Taking into account the advances made in semiconductor equipment, as well as the size of the market and number of players, the time line may be shorter, but the steps to reach the goal will be the same. There are five steps to be analyzed: feasibility, niche applications, reliability, cost reduction, and high-volume production.

Feasibility Decade

IBM developed flip chip for ultra-performance mainframe processors in the 1960s. The main purpose was to improve performance of high-end computing systems. In 1965, IBM introduced its flip chip process called controlled collapse chip connection, or C4. The next question asked was whether low cost organic substrates, like FR4, with relatively high CTE's, could be used for flip chip, which provides significant performance benefits. It seems that the 90s were the TSV feasibility decade, because many companies filed TSV patent applications and started exploring its performance benefits for 3-D applications. The motivation was to achieve greater packaging densities with increased demand for greater bandwidth and power consumption as well as improved electrical interconnects. The question here is whether low-cost thin silicon wafers with TSVs be attached to other low-cost organic substrates and/or silicon wafers to provides significant performance benefits.

Niche Decade

In the 1970s, the flip chip solder bumped chip attach process was taken up by Delco Electronics, and has since become very common in automotive applications. Others dedicated their investments to other niche applications. The 2000s may be called the TSV niche decade. Companies started delivering wafers with TSVs for some niche applications, such as MEMS, RF, CMOS image sensors, and began collecting reliability data.

Reliability Decade

In the early 1980s, Hewlett Packard Laboratories was also involved in the R&D and reliability studies of flip chip solder bumping and eventually released the technology in high-end server processors in the 1990s. I think the reliability decade of TSVs has just started, as some companies and universities are now presenting TSV reliability data. This reliability "decade" may be compressed into 3-5 years, but in this economic environment it may take longer. It is clear that there is a lack of TSV reliability data and very few companies have done sufficient studies. ALLVIA has collected enough reliability data required for development of new products.

Cost-Reduction Decade

Flip chip technology became commercially available for microprocessors in the 1990s. By the early 1990s, Intel adopted flip chip technology; its focus was R&D of solder-bumped underfilled silicon on polymer substrate technology to bring the cost to an acceptable level. The question is how long the TSV cost reduction period will take. Maybe, again, everyone will wait for a big company to drive TSV to acceptable high-volume costs. Perhaps several vertically integrated companies that appreciate TSV performance benefits will invest enough to develop cost-effective solutions. The cost issue remains the main obstacle for high volume production.

High-Volume Production Decade

A growing percentage of cell phones, PDAs, and high-speed microprocessors are assembled with flip chip technology today. If the reliability and cost reduction decades can be shortened to 3-5 years each, then high-volume production using TSVs could happen in 6-10 years, i.e., 2015-2020. If not, then high-volume production will start after 2020.

>Contact Sergey Savastiouk, PhD, CEO, ALLVIA, Inc., 657 N. Pastoria Ave., Sunnyvale, CA 94085 USA; ph.: 408-720-3333; savastiouk@allvia.com.