Efficient Storage of Defect Maps for Nanoscale Memory

Publish in



Please download to get full document.

View again

of 25
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Efficient Storage of Defect Maps for Nanoscale Memory. Susmit Biswas Tzvetan S. Metodi* Frederic T. Chong Ryan Kastner Tim Sherwood {susmit,chong,sherwood}@cs.ucsb.edu, kastner@ece.ucsb.edu, tsmetodiev@ucdavis.edu. *. Nanotechnology in Action. Scaling limit of CMOS Vdd ~ 1V Leakage
Efficient Storage of Defect Maps for Nanoscale MemorySusmit BiswasTzvetan S. Metodi* Frederic T. ChongRyan KastnerTim Sherwood{susmit,chong,sherwood}@cs.ucsb.edu, kastner@ece.ucsb.edu, tsmetodiev@ucdavis.edu*Nanotechnology in Action
  • Scaling limit of CMOS
  • Vdd ~ 1V
  • Leakage
  • Design of new nanoscale devices
  • SONOS, CMOL, Crossbar memories
  • CNT interconnect
  • Nanotechnology: Pros and Cons
  • Higher Density
  • 5nm<=Fnano <= 10nm
  • Faster operation
  • Fast switching
  • Low active power
  • Inexpensive
  • Reliability
  • Manufacturing defects might be as high as 10% [DeHon-NanoTech2005]
  • Solution: Reconfiguration
  • Dynamic
  • High latency in testing
  • Static: Defect Map
  • Using bit-level reconfiguration
  • High overhead in storage
  • Block level reconfiguration
  • Efficient storage techniques !Presentation overview
  • Motivation
  • Prior Work
  • Our Approach
  • Algorithms
  • Results
  • Conclusion
  • Prior Work
  • 1D list of regions [sun-NanoArch-06]
  • Bloom filter defect map [Wang-ICCAD2006]
  • Example: A Defective MemoryMap of good regions
  • List based approach
  • Ranges
  • 1D
  • 2D
  • Can be stored in TCAM
  • Good for correlated defects
  • Storing Defect Map in RectanglesEquivalent Problem:Finding optimal rectangle cover
  • NP-Complete problem
  • Greedy Algorithm
  • R-Tree as data structure
  • New point inserted greedily for least increase in rectangle area
  • Suitable for storing ranges
  • k-means clustering to decide the insertion order of points
  • Algorithm6R431R58R2R1R6752R34Algorithm 1: IllustrationRootR1R2R5R3R6R4Storing Sparse Defect Locations
  • Bloom filter defect map [Wang-ICCAD2006]
  • Supports membership queries
  • Uniform hash function
  • No false negative
  • False positive
  • Better storage efficiency than bit vector
  • D={d1, d2,…, dn}H1(d1)H2 (d1)H3 (d1)H4 (d1)Bloom filter as Defect MapCombined ApproachCombined Approach6R431R58R2R1R6752R34Algorithm 2: IllustrationRootR1R2R3R4Distribution of densityImprovement in distributionExperiments
  • Error Model
  • Gaussian distribution
  • Test data
  • Synthetic
  • TCAM: 128 Entry
  • Bloom filter: 5 times number of points
  • ResultsCoverage of ErrorsConclusion
  • Defect map storage techniques
  • Region based
  • Combined approach with Bloom filter
  • Error model
  • Need of Finer model
  • Questions?Thanks!References[sun-NanoArch-06] “Two Fault Tolerance Design Approaches for Hybrid CMOS / Nanodevice Digital Memories”,Fei Sun and Tong Zhang, NanoArch ’06[Wang-ICCAD2006]“On The Use of Bloom Filters for Defect Maps in Nanocomputing”, Gang Wang, Wenrui Gong, Ryan Kastner, ICCAD ’06[DeHon-NanoTech2005]“Non-Photolithographic Nanoscale Memory Density Prospects”, André DeHon, Seth Copen Goldstein, Philip J. Kuekes, and Patrick Lincoln, IEEE Tr. Nanotechnology ’05[Nicolaidis-JET2005] “Memory Defect Tolerance Architectures for Nanotechnologies”,Michael Nicolaidis, Lorena Anghel, Nadir Achouri, Journal of El. Test. 2005
    Related Search

    Previous Document

    TVI Express Presentation

    Related Documents
    View more...
    We Need Your Support
    Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

    Thanks to everyone for your continued support.

    No, Thanks