Many-Core vs. Many-Thread Machines: Stay Away From the Valley

We study the tradeoffs between many-core machines like Intelpsilas Larrabee and many-thread machines like Nvidia and AMD GPGPUs. We define a unified model describing a superposition of the two architectures, and use it to identify operation zones for which each machine is more suitable. Moreover, we identify an intermediate zone in which both machines deliver inferior performance. We study the shape of this ldquoperformance valleyrdquo and provide insights on how it can be avoided.

[1]  Trevor N. Mudge,et al.  An Analytical Model for Designing Memory Hierarchies , 1996, IEEE Trans. Computers.

[2]  Mark Horowitz,et al.  An analytical cache model , 1989, TOCS.

[3]  Harold S. Stone,et al.  Footprints in the cache , 1986, SIGMETRICS '86/PERFORMANCE '86.

[4]  C. K. Chow,et al.  Determination of Cache's Capacity and its Matching Storage Hierarchy , 1976, IEEE Transactions on Computers.

[5]  Harold S. Stone,et al.  Footprints in the cache , 1987, TOCS.

[6]  Dawid Pajak.  General-Purpose Computation Using Graphics Hardware for Fast HDR Image Processing , 2007 .

[7]  Edward T. Grochowski,et al.  Larrabee: A many-Core x86 architecture for visual computing , 2008, 2008 IEEE Hot Chips 20 Symposium (HCS).

[8]  Mark D. Hill,et al.  Amdahl's Law in the Multicore Era , 2008 .

[9]  Anant Agarwal,et al.  Performance Tradeoffs in Multithreaded Processors , 1992, IEEE Trans. Parallel Distributed Syst..

[10]  Uri C. Weiser,et al.  Performance, power efficiency and scalability of asymmetric cluster chip multiprocessors , 2006, IEEE Computer Architecture Letters.

[11]  Mark D. Hill,et al.  Amdahl's Law in the Multicore Era , 2008, Computer.

[12]  nVIDIA社.  CUDA Programming Guide 1.1 , 2007 .