Discussions.MATLABBenchmarks History

Hide minor edits - Show changes to output

Added lines 5-8:
'''Mac Pro, 2x Quad-Core 3.2GHz Intel Xeon, Mac OS X.5.2 Server'''\\
''franklin''
    0.1270    0.2868    0.1875    0.4961    0.2160    0.0046

Added lines 14-17:
'''Mac Pro, 2x Dual-Core 3.0GHz Intel Xeon, Mac OS X.4.10'''\\
''westinghouse''
    0.1479    0.2576    0.1990    0.6584    0.2562    0.0058

Changed line 19 from:
''bell, edison, ford, otis, orville, wilbur''
to:
''bell, cochrane, edison, ford, orville, wilbur''
Deleted lines 21-24:
'''Mac Pro, 2x Dual-Core 3.0GHz Intel Xeon, Mac OS X.4.10'''\\
''westinghouse''
    0.1479    0.2576    0.1990    0.6584    0.2562    0.0058

Changed line 28 from:
''franklin''
to:
''otis''
Changed line 5 from:
'''TyanPSC Typhoons, 2x Quad-Core Xeon X5355 2.66GHz per node, Platform OCS RHEL4'''\\
to:
'''TyanPSC Typhoons, 2x Quad-Core 2.66GHz Intel Xeon on each node, Platform OCS RHEL4'''\\
Changed line 10 from:
'''Power Mac, Quad G5 2.5GHz, Mac OS X.4.10'''\\
to:
'''Power Mac, 4x 2.5GHz PowerPC G5, Mac OS X.4.10'''\\
Changed line 14 from:
'''Mac Pro, Dual Dual-Core Xeon 3.0GHz, Mac OS X.4.10'''\\
to:
'''Mac Pro, 2x Dual-Core 3.0GHz Intel Xeon, Mac OS X.4.10'''\\
Changed line 18 from:
'''Mac Mini, Intel Core Duo 1.83GHz, Mac OS X.4.10'''\\
to:
'''Mac Mini, 1.83GHz Intel Core Duo, Mac OS X.4.10'''\\
Changed line 23 from:
'''Mac Pro, Dual G5 2.5GHz?, Mac OS X.4.10'''\\
to:
'''Mac Pro, Dual 2.7GHz PowerPC G5, Mac OS X.4.9'''\\
Changed lines 3-4 from:
Each of the below bench results (LU, FFT, ODE, Sparse, 2-D, 3-D) is the last line of a @@bench(5)@@ command, to account for any lag in throttling processors. Each bench was run in a remote terminal, unless otherwise indicated, so the last 2 columns are immaterial. None of these bench tests were run in multithreaded mode; as described in the [[http://www.mathworks.com/company/newsletters/news_notes/june07/clevescorner.html | June 2007 Cleve's Corner]], many parts of MATLAB aren't currently multithreaded and a few others (the sparse bench, for instance) actually perform slower with multithreading enabled, while most single-threaded tasks run as fast running one copy as running N copies simultaneously on an N-core machine. Given our typically large number of computing tasks and ability to run in [[Code/ParallelMATLAB | parallel]], we will usually leave MATLAB's multithreading disabled for now. Also, remember that, as per the documentation, the benchmark only compares performance of a particular Matlab version installed on different machines (tasks change with version).
to:
Each of the below bench results (LU, FFT, ODE, Sparse, 2-D, 3-D) is the last line of a @@bench(5)@@ command, to account for any lag in throttling processors. Each bench was run in a remote terminal, unless otherwise indicated, so the last 2 columns are immaterial. None of these bench tests were run in multithreaded mode; as described in the [[http://www.mathworks.com/company/newsletters/news_notes/june07/clevescorner.html | June 2007 Cleve's Corner]], many parts of MATLAB aren't currently multithreaded and a few others (the sparse bench, for instance) actually perform slower with multithreading enabled, while most single-threaded tasks run as fast running one copy as running N copies simultaneously on an N-core machine. Given our typically large number of computing tasks and ability to run in [[Code/ParallelMatlab | parallel]], we will usually leave MATLAB's multithreading disabled for now. Also, remember that, as per the documentation, the benchmark only compares performance of a particular Matlab version installed on different machines (tasks change with version).
Changed lines 3-4 from:
Each of the below bench results (LU, FFT, ODE, Sparse, 2-D, 3-D) is the last line of a @@bench(5)@@ command, to account for any lag in throttling processors. Each bench was run in a remote terminal, unless otherwise indicated, so the last 2 columns are immaterial. To my knowledge, none of the bench tests are threaded, and it is my experience that they run as fast running one copy as running N copies simultaneously on an N-core machine. Also, remember that, as per the documentation, the benchmark only compares performance of a particular Matlab version installed on different machines (tasks change with version).
to:
Each of the below bench results (LU, FFT, ODE, Sparse, 2-D, 3-D) is the last line of a @@bench(5)@@ command, to account for any lag in throttling processors. Each bench was run in a remote terminal, unless otherwise indicated, so the last 2 columns are immaterial. None of these bench tests were run in multithreaded mode; as described in the [[http://www.mathworks.com/company/newsletters/news_notes/june07/clevescorner.html | June 2007 Cleve's Corner]], many parts of MATLAB aren't currently multithreaded and a few others (the sparse bench, for instance) actually perform slower with multithreading enabled, while most single-threaded tasks run as fast running one copy as running N copies simultaneously on an N-core machine. Given our typically large number of computing tasks and ability to run in [[Code/ParallelMATLAB | parallel]], we will usually leave MATLAB's multithreading disabled for now. Also, remember that, as per the documentation, the benchmark only compares performance of a particular Matlab version installed on different machines (tasks change with version).
Changed line 5 from:
!!!!TyanPSC Typhoons, 2x Quad-Core Xeon X5355 2.66GHz per node, Platform OCS RHEL4 
to:
'''TyanPSC Typhoons, 2x Quad-Core Xeon X5355 2.66GHz per node, Platform OCS RHEL4'''\\
Changed line 10 from:
!!!!Power Mac, Quad G5 2.5GHz, Mac OS X.4.10
to:
'''Power Mac, Quad G5 2.5GHz, Mac OS X.4.10'''\\
Changed line 14 from:
!!!!Mac Pro, Dual Dual-Core Xeon 3.0GHz, Mac OS X.4.10
to:
'''Mac Pro, Dual Dual-Core Xeon 3.0GHz, Mac OS X.4.10'''\\
Changed line 18 from:
!!!!Mac Mini, Intel Core Duo 1.83GHz, Mac OS X.4.10
to:
'''Mac Mini, Intel Core Duo 1.83GHz, Mac OS X.4.10'''\\
Changed line 23 from:
!!!!Mac Pro, Dual G5 2.5GHz?, Mac OS X.4.10
to:
'''Mac Pro, Dual G5 2.5GHz?, Mac OS X.4.10'''\\
Changed line 27 from:
!!!!Thinkpad X41 Tablet, Intel Pentium M 1.50GHz, WinXP Tablet 2005 SP2
to:
'''Thinkpad X41 Tablet, Intel Pentium M 1.50GHz, WinXP Tablet 2005 SP2'''\\
Added lines 1-29:
!!MATLAB Version 7.4.0.287 (R2007a)

Each of the below bench results (LU, FFT, ODE, Sparse, 2-D, 3-D) is the last line of a @@bench(5)@@ command, to account for any lag in throttling processors. Each bench was run in a remote terminal, unless otherwise indicated, so the last 2 columns are immaterial. To my knowledge, none of the bench tests are threaded, and it is my experience that they run as fast running one copy as running N copies simultaneously on an N-core machine. Also, remember that, as per the documentation, the benchmark only compares performance of a particular Matlab version installed on different machines (tasks change with version).

!!!!TyanPSC Typhoons, 2x Quad-Core Xeon X5355 2.66GHz per node, Platform OCS RHEL4
''deere, tesla''
    0.1324    0.2851    0.1938    0.4290    0.1480    0.0044  (head node)
    0.1308    0.2795    0.1481    0.3923    0.1181    0.0034  (compute node)

!!!!Power Mac, Quad G5 2.5GHz, Mac OS X.4.10
''bell, edison, ford, otis, orville, wilbur''
    0.1376    0.3068    0.2800    1.4812    0.4366    0.0107

!!!!Mac Pro, Dual Dual-Core Xeon 3.0GHz, Mac OS X.4.10
''westinghouse''
    0.1479    0.2576    0.1990    0.6584    0.2562    0.0058

!!!!Mac Mini, Intel Core Duo 1.83GHz, Mac OS X.4.10
''mini''
    0.5316    0.4718    0.3550    1.2764    0.4309    0.0106  (in terminal)
    0.5308    0.4722    0.3555    1.2817    1.0587    0.4323  (in Command Window)

!!!!Mac Pro, Dual G5 2.5GHz?, Mac OS X.4.10
''franklin''
    0.2324    0.3923    0.4382    1.9575    0.6346    0.0233

!!!!Thinkpad X41 Tablet, Intel Pentium M 1.50GHz, WinXP Tablet 2005 SP2
''tablet''
    0.8862    0.5437    0.3143    0.9704    0.5612    0.4379  (in Command Window)