Reliability-performance tradeoffs between 2.5D and 3D-stacked DRAM processors

Three-dimensional DRAM stacking has emerged as a vehicle for scaling system densities and performance improvement. The two design choices for interfacing to processors are - i) a separate core die connected to the DRAM stack via a silicon interposer (2.5D), and ii) DRAM die stacked on top of the core die (3D). These alternatives have different performance, power, and reliability behaviors. Specifically, 3D designs realize higher performance but operate at higher temperatures and thus exhibit lower lifetime. On the other hand, 2.5D designs provide lower bandwidth between the core die and the DRAM stack, but exhibit significantly longer lifetime due to less thermally-induced degradation. This paper explores this tradeoff between reliability and performance of 3D and 2.5D stacked memory systems. Our results indicate that, in general, lower voltage and frequency operations with 3D stacked systems may achieve balanced reliability-performance tradeoff.

[1]  Jun Wang,et al.  Manifold: A parallel simulation framework for multicore systems , 2014, 2014 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS).

[2]  Luca Benini,et al.  Energy optimization in 3D MPSoCs with Wide-I/O DRAM using temperature variation aware bank-wise refresh , 2014, 2014 Design, Automation & Test in Europe Conference & Exhibition (DATE).

[3]  Muhannad S. Bakir,et al.  A Silicon Interposer Platform Utilizing Microfluidic Cooling for High-Performance Computing Systems , 2015, IEEE Transactions on Components, Packaging and Manufacturing Technology.

[4]  Kai Li,et al.  The PARSEC benchmark suite: Characterization and architectural implications , 2008, 2008 International Conference on Parallel Architectures and Compilation Techniques (PACT).

[5]  Yogendra Joshi,et al.  Compact Modeling of 3D Stacked Die Inter-Tier Microfluidic Cooling Under Non-Uniform Heat Flux , 2012 .

[6]  Sudhakar Yalamanchili,et al.  Managing performance-reliability tradeoffs in multicore processors , 2015, 2015 IEEE International Reliability Physics Symposium.

[7]  Eduard Ayguadé,et al.  Another Trip to the Wall: How Much Will Stacked DRAM Benefit HPC? , 2015, MEMSYS.

[8]  Jung Ho Ahn,et al.  McPAT: An integrated power, area, and timing modeling framework for multicore and manycore architectures , 2009, 2009 42nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO).

[9]  Seth H. Pugsley,et al.  USIMM : the Utah SImulated Memory Module , 2012 .

[10]  Norbert Wehn,et al.  Omitting Refresh: A Case Study for Commodity and Wide I/O DRAMs , 2015, MEMSYS.