The Community for Technology Leaders
RSS Icon
Subscribe
Issue No.04 - July-Aug. (2013 vol.33)
pp: 16-24
Hao Wang , University of Wisconsin-Madison
Nam Sung Kim , University of Wisconsin-Madison
ABSTRACT
It has been reported that carbon nanotube (CNT) devices are faster and consume less power than CMOS devices. However, current CNT devices exhibit a higher defect rate than CMOS devices. To reduce the defect rate of CNT devices, a device-level redundancy technique can be adopted. However, more device-level redundancy in turn increases area, delay, and power consumption of integrated circuits (ICs). In this article, the authors propose to use slightly less device-level redundancy than required for all processor cores to be defect-free for a yield target, which makes cores smaller, faster, and more power efficient. Although some cores can be defective with less device-level redundancy, many-core processors can tolerate some defective cores by design. Under the same power and yield constraints, the authors demonstrate that a CNT processor with less device-level redundancy can provide 1.75 times higher throughput despite also being nearly 2 times smaller than a CNT processor that has more device-level redundancy and that also makes all cores defect free.
INDEX TERMS
Program processors, CMOS integrated circuits, Redundancy, Multicore processing, Inverters, System-on-chip, Carbon nanotubes, Power system reliability, power constraint, carbon nanotube, many-core processor, reliability
CITATION
Hao Wang, Nam Sung Kim, "Improving Throughput of Power-Constrained Many-Core Processors Based on Unreliable Devices", IEEE Micro, vol.33, no. 4, pp. 16-24, July-Aug. 2013, doi:10.1109/MM.2013.69
REFERENCES
1. N. Patil et al., "Digital VLSI Logic Technology Using Carbon Nanotube FETs: Frequently Asked Questions," Proc. 46th ACM/IEEE Design Automation Conf. (DAC 09), IEEE CS, 2009, pp. 304-309.
2. J. Guo et al., "Performance Analysis and Design Optimization of Near Ballistic Carbon Nanotube Field-Effect Transistors," Proc. IEEE Int'l Electron Devices Meeting, IEEE CS, 2004, pp. 703-706.
3. A. Javey et al., "Advancements in Complementary Carbon Nanotube Field-Effect Transistors," Proc. IEEE Int'l Electron Devices Meeting, IEEE CS, 2003, pp. 31.2.1-31.2.4.
4. N. Patil et al., "Automated Design of Misaligned-Carbon-Nanotube-Immune Circuits," Proc. 44th Ann. Design Automation Conf. (DAC 07), ACM, 2007, pp. 958-961.
5. J. Zhang et al., "Carbon Nanotube Circuits: Living with Imperfections and Variations," Proc. Design, Automation and Test in Europe Conf., IEEE CS, 2010, pp. 1159-1164.
6. L. Nougaret et al., "80 GHz Field-Effect Transistors Produced Using High Purity Semiconducting Single-Walled Carbon Nanotubes," Applied Physics Letters, vol. 94, no. 24, 2009, article 243,505.
7. A. Lin et al., "ACCNT: A Metallic-CNT-Tolerant Design Methodology for Carbon Nanotube VLSI: Analyses and Design Guidelines," IEEE Trans. Electron Devices, Sept. 2010, pp. 2284-2295.
8. S. Chaudhry, "Rock: A SPARC CMT Processor," IEEE Micro, vol. 29, no. 2, 2009, pp. 6-16.
9. N. Patil et al., "Circuit-Level Performance Benchmarking and Scalability Analysis of Carbon Nanotube Transistor Circuits," IEEE Trans. Nanotechnology, Jan. 2009, pp. 37-45.
10. J. Deng et al., "Carbon Nanotube Transistor Compact Model for Circuit Design and Performance Optimization," ACM J. Emerging Technologies in Computing Systems, Apr. 2008, pp. 7:1-7:20.
11. S. Borkar, Major Challenges to Achieve Exascale Performance, tech. report, Intel, 2009.
12. SRC/NSF/A*STAR Forum on 2020 Semiconductor Memory Strategies: Processes, Devices, and Architectures, report, 2009; http://www.src.org/calendar/e003676finalreport.pdf.
16 ms
(Ver 2.0)

Marketing Automation Platform Marketing Automation Tool