Loading...

A Novel STT-RAM Architecture for Last Level Shared Caches in GPUs

Samavatian, Mohammad Hossein | 2013

783 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 44553 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Sarbazi-Azad, Hamid
  7. Abstract:
  8. Due to the high processing capacity of GPGPUs and their requirement to a large and high speed shared memory between thread processors clusters, exploiting Spin-Transfer Torque (STT) RAM as a replacement with SRAM can result in significant reduction in power consumption and linear enhancement of memory capacity in GPGPUs. In the GPGPU (as a many-core) with ability of parallel thread executing, advantages of STT-RAM technology, such as low read latency and high density, could be so effective. However, the usage of STT-RAM will be grantee applications run time reduction and growth threads throughput, when write operations manages and schedules to have least overhead on read operations. The purpose of this thesis is propose and evaluate a STT-RAM architecture for last level cache (LLC) in GPGPUs which uses circuit and architectural level techniques for managing access operation to LLC. First by reducing retention time of STT-RAM cells hybrid architecture introduced, then with characterization of GPGPU workloads, cache parameters such as cache micro-architecture, data retention time, latency and energy consumption were calculated. Finally by simulating target architecture with different design explores latency and power consumption of cache was measured. Proposed architecture result in Performance gain 16% in average and 100% maximum with 20% power saving. On the other hand with techniques used for data searching in cache, power consumption reduced 40% with performance improvement degradation from 16% to 15%
  9. Keywords:
  10. Energy Consumption Reduction ; General Purpose Graphic Processing Units (GPGPU) ; Spin Transfer Torque-Magnetic (STT-MRAM) ; Last Level Cache (LLC) ; Data Retention Time

 Digital Object List

 Bookmark

No TOC