Besides the great fanfare, Intel's transition to 3D transistors
is another step in overcoming the roadblocks in advancing Moore's Law for at least several more process generations.
Intel has maintained a regular cadence of a new process every 24 months, on average, so advancing to the 22nm process was not surprising. The surprise came in the accompanying technology = 3D tri-gate transistors. The new transistor design is a shift from the traditional planar (or 2D) transistor to a 3D transistor using a vertical channel that stands above the silicon substrate and a gate that encapsulates the channel.
So Intel advanced one of the three key pillars of innovation in semiconductor manufacturing technology, which are the materials, the lithography, and the transistor design.
In recent years, Intel, as well as the rest of the semiconductor industry, has implemented significant improvements in the first two pillars. In materials technology, the industry has implemented the use of strained silicon, metal gates, and high-k dielectrics. In lithography, the industry has moved to multiple patterning and immersion. There have also been different experimentations in creating the transistor structures, gate-first and gate-last, but 3D transistors is the most significant change in transistor design since the introduction of semiconductor devices in the early 1970s.
Intel and others in the industry have been working on 3D transistors for over a decade. So why Intel's announcement comes today?
"The industry often works on new technologies for extended periods of time, only being brought to market if and when they are proven to improve certain aspects of semiconductor technology and can be implemented in high volume manufacturing for and acceptable cost," In-stat Chief Technology Strategist Jim McGregor explained.
Intel's annoucnement indicated that the 3D transistor will add some additional process steps to manufacturing and approximately a 2% - 3% increase in manufacturing cost, but it will not require any special tooling. This is good news for Intel and the rest of the industry, which is likely to adopt the technology in the future. Intel cites at least a three-year lead on other semiconductor manufacturers. This figure translates to a process node generation (2 years) plus Intel?s typical lead at each process node (1 year). With no other 3D transistor announcements by other semiconductor manufacturers, this is likely the case.
The industry has considered 3D transistors for quite some time because of the potential to reduce leakage current and produce more predictable transistor performance characteristics, but the benefits don?t end there. Intel?s analysis also demonstrates the potential for faster transistor transitions and lower voltage levels, which all translates into higher-performance and lower power chips. As with traditional planar transistor, Intel has developed various transistor designs to be able to optimize the transistors for power or performance.
While the new technology will eventually benefit the entire semiconductor industry, Intel is expected to see the initial benefit through lower chip costs and competitive advantages in performance and efficiency.
"While this is not a huge impact to Intel?s primary competitor, AMD, it does put additional pressure on the smaller rival, which has struggled to keep pace with Intel," Mr. McGregor added.
It will also benefit Intel in its efforts to provide low-power products based on the company's Atom processor architecture for other mobile, consumer, and embedded applications. According to Intel?s roadmap, the first products using the new technology will be PC processors and server processors slated for introduction in early 2012, which translates into production during the latter half of 2011. Intel will utilize the current 32nm process technology for the upcoming Atom-based products and utilize the 22nm process technology for the follow-on generation.
While Intel's manufacturing plans are logical from a manufacturing standpoint, In-Stat believes that Intel would be better off pulling the upcoming generation of Atom products onto 22nm even if it resulted in a short delay, likely a few months, in releasing the products. Intel's manufacturing strategy is driven by the dynamics of ramping a new process technology and the requirement to maximize the ROI. Ramping a new process is a very costly learning process. Semiconductor manufacturers try to begin manufacturing with modest yields while improving the yields with each lot until acceptable yields are reached and/or further improvements are too costly to implement (the point of diminishing returns). Semiconductor manufacturers also seek to ramp new processes as quick as possible to gain a competitive advantage in the market, reduce chip costs, and maximize the ROI on a huge investment in process R&D and manufacturing capacity. The natural way to accomplish these goals is to begin with the highest volume product, which is PC processors for Intel. However, Intel faces significant hurdles penetrating the consumer electronics markets. While the upcoming family of products may be competitive, Intel needs to leapfrog the competition if it hopes to unseat the incumbent semiconductor suppliers in these new markets, especially the billion unit handset market. While moving the upcoming Atom products to 22nm would still not guarantee Intel success in the consumer electronics markets, it could provide a must needed boost.
"Intel continues to demonstrate its industry leadership and continued focus on its core competencies: semiconductor manufacturing, R&D, and the x86 architecture. The transition to 3D transistors is another step in overcoming the roadblocks in advancing Moore's Law for at least several more process generations. 22nm and 3D transistors also ensure a continued trend in performance and power efficiency for the next generation of electronic devices," In-stat commented.