ACCMP-assymetric cluster chip-multiprocessing

Achieving high performance within a power envelope becomes extremely difficult as feature sizes decrease. Uniprocessor architects encounter diminishing returns when trying to attain higher performance in exchange for area and power. For multithreaded programs, symmetric chip multiprocessing (CMP) offers higher throughput and power efficiency than uniprocessors. However, symmetric CMP does not scale well with technology. In this paper we explore the theoretical advantages of placing asymmetric core clusters in multiprocessor chips. All of the cores on the ACCMP die have the same instruction set architecture, but may have a completely different micro-architecture. We show that asymmetric core clusters are expected to achieve higher performance per area and higher performance for a given power envelope.

[1]  Brad Calder,et al.  Discovering and Exploiting Program Phases , 2003, IEEE Micro.

[2]  W. Dally,et al.  Route packets, not wires: on-chip interconnection networks , 2001, Proceedings of the 38th Design Automation Conference (IEEE Cat. No.01CH37232).

[3]  L. Ivanov,et al.  Modeling and verification of cache coherence protocols , 2001, ISCAS 2001. The 2001 IEEE International Symposium on Circuits and Systems (Cat. No.01CH37196).

[4]  Balaram Sinharoy,et al.  IBM Power5 chip: a dual-core multithreaded processor , 2004, IEEE Micro.

[5]  Norman P. Jouppi,et al.  Single-ISA heterogeneous multi-core architectures for multithreaded workload performance , 2004, Proceedings. 31st Annual International Symposium on Computer Architecture, 2004..