A hierarchical optimization engine for nanoelectronic systems using emerging device and interconnect technologies
MetadataShow full item record
A fast and efficient hierarchical optimization engine was developed to benchmark and optimize various emerging device and interconnect technologies and system-level innovations at the early design stage. As the semiconductor industry approaches sub-20nm technology nodes, both devices and interconnects are facing severe physical challenges. Many novel device and interconnect concepts and system integration techniques are proposed in the past decade to reinforce or even replace the conventional Si CMOS technology and Cu interconnects. To efficiently benchmark and optimize these emerging technologies, a validated system-level design methodology is developed based on the compact models from all hierarchies, starting from the bottom material-level, to the device- and interconnect-level, and to the top system-level models. Multiple design parameters across all hierarchies are co-optimized simultaneously to maximize the overall chip throughput instead of just the intrinsic delay or energy dissipation of the device or interconnect itself. This optimization is performed under various constraints such as the power dissipation, maximum temperature, die size area, power delivery noise, and yield. For the device benchmarking, novel graphen PN junction devices and InAs nanowire FETs are investigated for both high-performance and low-power applications. For the interconnect benchmarking, a novel local interconnect structure and hybrid Al-Cu interconnect architecture are proposed, and emerging multi-layer graphene interconnects are also investigated, and compared with the conventional Cu interconnects. For the system-level analyses, the benefits of the systems implemented with 3D integration and heterogeneous integration are analyzed. In addition, the impact of the power delivery noise and process variation for both devices and interconnects are quantified on the overall chip throughput.