It is questionable whether DRAM will continue to scale and will meet the needs of next-generation systems. Therefore, significant effort is invested in research and development of novel memory technologies. One of the candidates for nextgeneration memory is Spin-Transfer Torque Magnetic Random Access Memory (STT-MRAM). STT-MRAM is an emerging non-volatile memory with a lot of potential that could be exploited for various requirements of different computing systems.
Being a novel technology, STT-MRAM devices are already approaching DRAM in terms of capacity, frequency and device size.
Special STT-MRAM features such as intrinsic radiation hardness, non-volatility, zero stand-by power and capability to function in extreme temperatures also make it particularly suitable for aerospace, avionics and automotive applications.
Despite of being a conceivable alternative for main memory technology, to this day, academic research of STT-MRAM main memory remains marginal. This is mainly due to the unavailability of publicly available detailed timing parameters of this novel technology, which are required to perform a cycle accurate main memory simulation. Some researchers adopt simplistic memory models to simulate main memory, but such models can introduce significant errors in the analysis of the overall system performance. Therefore, detailed timing parameters are a must-have for any evaluation or architecture exploration study of STT-MRAM main memory. These detailed parameters are not publicly available because STT-MRAM manufacturers are reluctant to release any delicate information on the technology. This thesis demonstrates an approach to perform a cycle accurate simulation of STT-MRAM main memory, being the first to release detailed timing parameters of this technology from academia, essentially enabling researchers to conduct reliable system level simulation of STT-MRAM using widely accepted existing simulation infrastructure. Our results show that, in HPC domain STT-MRAM provide performance comparable to DRAM. Results from the power estimation indicates that STT-MRAM power consumption increases significantly for Activation/Precharge power while Burst power increases moderately and Background power does not deviate much from DRAM.
The thesis includes detailed STT-MRAM main memory timing parameters to the main repositories of DramSim2 and Ramulator, two of the most widely used and accepted state-of-the-art main memory simulators. The STT-MRAM timing parameters that has been originated as a part of this thesis, are till date the only reliable and publicly available timing information on this memory technology published from academia.
Finally, the thesis analyzes the feasibility of using STT-MRAM in real-time embedded systems by investigating STT-MRAM main memory impact on average system performance and WCET. STT-MRAM's suitability for the real-time embedded systems is validated on benchmarks provided by the European Space Agency (ESA), EEMBC Autobench and MediaBench suite by analyzing performance and WCET impact. In quantitative terms, our results show that STT-MRAM main memory in real-time embedded systems provides performance and WCET comparable to conventional DRAM, while opening up opportunities to exploit various advantages.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados