Will FinFET's benefits outweigh the costs of using the new process? Here are some things that could affect the learning and cost curve.
The picture below shows a more realistic depiction of the FinFET -- with rounded features replacing the boxes and planes of the previous example. Most designers agree that predicting the electrical performance for this type of structure will require significant innovation. The electrical field around the device and the interconnect is much more complex than what they are used to encountering with traditional MOSFETs.
Additionally, FinFET devices will have higher drive strengths than similarly sized MOSFETs, which means designers will want higher accuracy in predicting the electrical behavior. To address these requirements, new techniques will be required for device and interconnect modeling.
FinFET representation with rounded features
(TEM image source: ChipWorks; simulation source: Gold Standard Simulations Ltd.)
Moreover, from an analog or IP designer's perspective, the design methodology described above (where fins are implemented by the foundry) is not a preferred use model. These designers want more freedom to reduce leakage, match drive strengths, improve frequency response, and push electrical and geometric limits in ways that fixed fin generation will not allow. This type of design is, by its nature, custom, and not being able to manipulate fins in number or size is something that will be foreign to many of these designers.
For custom, analog, or IP designers making the jump from 28 nm or above to FinFET processes, this type of design will be revolutionary, but not necessarily in a "new and improved" sense of the word. While there are tool innovations to ease the transition, the methodology for doing this type of design will probably seem restrictive compared with their customary design practices. With traditional MOSFET processes, these designers created custom configurations of transistors that included their size and orientation.
For FinFETs, designers will have to experiment with fewer options to achieve the desired electrical response. There have been several discussions questioning whether advanced analog design can be accomplished with FinFET processes. The answer is yes, but it will require significant changes in design methodology and, potentially, more experimentation.
Where do you see your company and your designs in relation to FinFETs -- evolutionary change, or revolutionary change? How are you preparing for the change, both from a tool perspective and a process flow perspective? Do you think FinFETs are worth the learning costs they will incur?
— Carey Robertson is a director of product marketing at Mentor Graphics Corp., overseeing the marketing activities for Calibre PERC, LVS and extraction products.