(2010) Variability in Nanoscale Fabrics: Bottom-up, Integrated Analysis and Mitigation
VAL |
|
Offline
Мэтр, проФАН любви... proFAN of love
Профиль
Группа: Администраторы
Сообщений: 37821
Пользователь №: 1
Регистрация: 6.03.2004
|
(2010) Variability in Nanoscale Fabrics: Bottom-up Integrated Analysis and MitigationИсточники: - https://www.umass.edu/nanofabrics/sites/def...ritish_2013.pdf- ACM Transactions on Computational Logic, Vol. V, No. N, Month 20YY, Pages 1–0??. Авторы: Pritish Narayanan1, Michael Leuchtenburg1, Jorge Kina2, Prachi Joshi1,Pavan Panchapakeshan1, Chi On Chui2and C. Andras Moritz1 (1 - University of Massachusetts, 2- University of California, Los Angeles) QUOTE | Emerging nano-device based architectures will be impacted by parameter variation in conjunc-tion with high defect rates. Variations in key physical parameters are caused by manufacturingimprecision as well as fundamental atomic scale randomness. In this paper, the impact of param-eter variation on nanoscale computing fabrics is extensively studied through a novel integratedmethodology across device, circuit and architectural levels. This integrated approach enables tostudy in detail the impact of physical parameter variation across all fabric layers. A final contri-bution of the paper includes novel techniques to address this impact. The variability framework,while generic, is explored extensively on the Nanoscale Application Specific Integrated Circuits(NASICs) nanowire fabric. For variation ofσ=10% in key physical parameters, the on current isfound to vary by up to 3.5X. Circuit-level delay shows up to 118% deviation from nominal. MonteCarlo simulations using an architectural simulator found 67% nanoprocessor chips to operate be-low nominal frequencies due to variation. New built-in variation mitigation and fault-toleranceschemes, leveraging redundancy, asymmetric delay paths and biased voting schemes, were devel-oped and evaluated to mitigate these effects. They are shown to improve performance by up to7.5X on a nanoscale processor design with variation, and improve performance in designs relyingon redundancy for defect tolerance - without variation assumed. Techniques show up to 3.8Ximprovement in effective-yield performance products even at a high 12% defect rate. The suite oftechniques provides a design space across key system-level metrics such as performance, yield andarea. |
--------------------
|
|
|
0 Пользователей читают эту тему (0 Гостей и 0 Скрытых Пользователей)
0 Пользователей: