Evaluating the performance of single and multiple core processors with PCMARK®05 and benchmark analysis

PCMark®05 [4, 8] is a highly popular synthetic benchmark for evaluating the performance of personal computers (PCs) with millions of downloads via the Internet. Based on open source and commercial applications, it measures the execution time of highly representative code extracts of these applications and reports scores reflecting the overall system performance, the CPU performance, the memory subsystem performance, the graphics subsystem performance, and the disk subsystem performance. In this article, we focus on the PCMark®05 CPU test suite which is composed of 8 tests to measure the performance and scalability of various Intel single- and dual-core processors. Six of these tests run a single application each. One test runs 2 multitasked applications in parallel and another test runs 4 multitasked applications simultaneously. We present the results of executing this benchmark's CPU test suite on high end Intel-based PC platforms with top of the line single processor and dual core processors, present the results of our profiling and hotspot analysis, shed some light on this test suite's prominent microarchitecture events and its active threads' distributions, and characterize this suite's workload. These results help in understanding the performance characteristics of this popular benchmark and in guiding future processor design enhancements.

[1]  John L. Henning SPEC CPU2000: Measuring CPU Performance in the New Millennium , 2000, Computer.

[2]  Milind Girkar,et al.  Tight analysis of the performance potential of thread speculation using spec CPU 2006 , 2007, PPOPP.

[3]  Fadi N. Sibai Performance analysis and workload characterization of the 3DMark05 benchmark on modern parallel computer platforms , 2007, CARN.

[4]  David Kaeli,et al.  Performance Characterization of SPEC CPU2006 Integer Benchmarks on x86-64 Architecture , 2006, 2006 IEEE International Symposium on Workload Characterization.

[5]  Shikharesh Majumdar Scheduling Strategies for Controlling Resource Contention on Multiprocessor Systems , 1999, HiPC.

[6]  Lieven Eeckhout,et al.  Measuring benchmark similarity using inherent program characteristics , 2006, IEEE Transactions on Computers.

[7]  Shikharesh Majumdar Performance scalability in multiprocessor systems with resource contention , 2000, 2000 IEEE International Symposium on Performance Analysis of Systems and Software. ISPASS (Cat. No.00EX422).

[8]  JunSeong Kim,et al.  Performance sensitivity of SPEC CPU2000 over operating frequency , 2004, ISICT.

[9]  Lizy Kurian John,et al.  Analysis of redundancy and application balance in the SPEC CPU2006 benchmark suite , 2007, ISCA '07.

[10]  Donald Yeung,et al.  BioBench: A Benchmark Suite of Bioinformatics Applications , 2005, IEEE International Symposium on Performance Analysis of Systems and Software, 2005. ISPASS 2005..

[11]  Jozo J. Dujmovic,et al.  Evolution and evaluation of SPEC benchmarks , 1998, PERV.

[12]  R. Todi SPEClite: using representative samples to reduce SPEC CPU2000 workload , 2001 .

[13]  A. J. KleinOsowski,et al.  MinneSPEC: A New SPEC Benchmark Workload for Simulation-Based Computer Architecture Research , 2002, IEEE Computer Architecture Letters.