Benchmarks

Share icon

I compared a 2006 vintage MacBook Pro (Core II Duo, 2.33GHz, 3GB) running OS X 10.4.11 with a 2010 MacBook Pro (i7, 2.66GHz, 8GB) running OS X 10.6.3. Igor Pro 6.20B02 was used in both cases. I used the Igor benchmark experiment (File->Example Experiments->Testing->Benchmark 2.02) and ran 121 iterations.

The total time to run the benchmark on the 2006 MacBook Pro was 54.77 seconds versus 30.35 on the 2010 MacBook Pro.

Here are the details:

**** test on Macintosh OS X10.6.3 using 6.20 and 121 passes; Mac Book Pro 2100 i7
Create new graph time: 124.98ms, relative speed= 1.78
big data update time: 58.25ms, relative speed= 1.37
curve fit time: 741.47µs, relative speed= 1.46
user curve fit time: 11.68ms, relative speed= 1.36
double complex fft time: 531.88µs, relative speed= 1.72
single complex fft time: 518.60µs, relative speed= 1.60
double real fft time: 255.68µs, relative speed= 1.58
single real fft time: 242.43µs, relative speed= 1.52
5 pass smooth time: 279.95µs, relative speed= 1.47
Sort 8192 points time: 11.84ms, relative speed= 1.62
WaveStats time: 139.58µs, relative speed= 1.45
simple eqn time: 470.12µs, relative speed= 1.11
exp eqn time: 634.93µs, relative speed= 1.35
sqrt eqn time: 498.00µs, relative speed= 1.38
sin eqn time: 324.76µs, relative speed= 1.44
User fit fctn time: 419.51µs, relative speed= 1.45
MatrixOp eqn time: 15.88µs, relative speed= 1.83
**** done ****
total test time= 30.3483
System: MacBook 1.83 GHz Intel Core 2 Duo, 2GB Ram + MacOS X 10.6.3 + Igor Pro 6.12

**** test on Macintosh OS X10.6.3 using 6.12 and 121 passes
Create new graph time: 202.17ms, relative speed= 1.48
big data update time: 99.08ms, relative speed= 1.45
curve fit time: 1.41ms, relative speed= 1.34
user curve fit time: 20.79ms, relative speed= 1.71
double complex fft time: 1.04ms, relative speed= 1.42
single complex fft time: 992.81µs, relative speed= 1.20
double real fft time: 482.43µs, relative speed= 1.17
single real fft time: 448.32µs, relative speed= 1.10
5 pass smooth time: 521.51µs, relative speed= 1.04
Sort 8192 points time: 24.95ms, relative speed= 1.53
WaveStats time: 248.28µs, relative speed= 1.07
simple eqn time: 467.41µs, relative speed= 2.99
exp eqn time: 745.09µs, relative speed= 2.70
sqrt eqn time: 781.24µs, relative speed= 2.26
sin eqn time: 577.25µs, relative speed= 2.01
User fit fctn time: 722.18µs, relative speed= 1.79
MatrixOp eqn time: 35.69µs, relative speed= 0.67
**** done ****
total test time= 50.1788

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 22, 2010 at 05:23 pm - Permalink

System: MacBook 1.83 GHz Intel Core 2 Duo, 2GB Ram + MacOS X 10.6.3 + Igor Pro 6.20b2

**** test on Macintosh OS X10.6.3 using 6.20 and 121 passes
Create new graph time: 206.54ms, relative speed= 1.45
big data update time: 99.94ms, relative speed= 1.43
curve fit time: 1.39ms, relative speed= 1.35
user curve fit time: 20.61ms, relative speed= 1.72
double complex fft time: 1.04ms, relative speed= 1.42
single complex fft time: 991.39µs, relative speed= 1.20
double real fft time: 486.61µs, relative speed= 1.16
single real fft time: 445.80µs, relative speed= 1.10
5 pass smooth time: 521.13µs, relative speed= 1.04
Sort 8192 points time: 24.39ms, relative speed= 1.57
WaveStats time: 256.93µs, relative speed= 1.04
simple eqn time: 664.27µs, relative speed= 2.10
exp eqn time: 1.03ms, relative speed= 1.95
sqrt eqn time: 876.45µs, relative speed= 2.02
sin eqn time: 634.70µs, relative speed= 1.83
User fit fctn time: 761.85µs, relative speed= 1.70
MatrixOp eqn time: 34.83µs, relative speed= 0.69
**** done ****
total test time= 51.5081

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 22, 2010 at 05:32 pm - Permalink

Might as well add a recently acquired machine:

HP Pavilion p6320y running AMD Phenom II X4 820 and Windows7 64bit.

**** test on Personal (Build 7600)6.1.7600 using 6.12 and 121 passes; HP AMD Phenom II x4
Create new graph time: 100.18ms, relative speed= 2.99
big data update time: 90.35ms, relative speed= 1.58
curve fit time: 1.39ms, relative speed= 1.35
user curve fit time: 15.09ms, relative speed= 2.35
double complex fft time: 1.04ms, relative speed= 1.42
single complex fft time: 791.35µs, relative speed= 1.50
double real fft time: 408.08µs, relative speed= 1.38
single real fft time: 403.26µs, relative speed= 1.22
5 pass smooth time: 339.44µs, relative speed= 1.60
Sort 8192 points time: 14.31ms, relative speed= 2.67
WaveStats time: 630.31µs, relative speed= 0.42
simple eqn time: 534.88µs, relative speed= 2.61
exp eqn time: 824.48µs, relative speed= 2.44
sqrt eqn time: 717.41µs, relative speed= 2.47
sin eqn time: 645.37µs, relative speed= 1.80
User fit fctn time: 548.14µs, relative speed= 2.36
MatrixOp eqn time: 22.26µs, relative speed= 1.07
**** done ****
total test time= 35.2478

April 25, 2010 at 09:38 am - Permalink

I just upgraded my mobile machine, so I'm posting benchmarks here as a point of reference for anybody considering an upgrade. My old machine was a 2.2 GHz Core2 duo from 2007 or so.

Early 2011 Macbook Pro with 8 GB of Ram, 1 GB vram, 5400 RPM HDD

**** test on Macintosh OS X10.6.7 using 6.21 and 121 passes; Early 2011 MBP
Create new graph time: 118.60ms, relative speed= 2.52
big data update time: 50.81ms, relative speed= 2.82
curve fit time: 654.89µs, relative speed= 2.88
user curve fit time: 6.45ms, relative speed= 5.50
double complex fft time: 422.30µs, relative speed= 3.49
single complex fft time: 403.66µs, relative speed= 2.94
double real fft time: 205.60µs, relative speed= 2.74
single real fft time: 198.97µs, relative speed= 2.47
5 pass smooth time: 256.83µs, relative speed= 2.11
Sort 8192 points time: 11.14ms, relative speed= 3.43
WaveStats time: 134.72µs, relative speed= 1.98
simple eqn time: 267.20µs, relative speed= 5.22
exp eqn time: 397.90µs, relative speed= 5.05
sqrt eqn time: 295.25µs, relative speed= 5.99
sin eqn time: 281.92µs, relative speed= 4.13
User fit fctn time: 204.35µs, relative speed= 6.32
MatrixOp eqn time: 11.86µs, relative speed= 2.01
**** done ****
total test time= 26.1131

This laptop has two gpu's in it, and it didn't seem to matter which one was enabled at the time of testing (forced by gfxCardStatus 2.0.1). Also, I was able to decrease the total test time by about 9 seconds by hiding the application while it ran (cmd-h), reminding us that suppressing graphics is a great way to get faster performance with current hardware, without the need to upgrade.

I will post another update with a workstation running a 6-core 3.2GHz i980x cpu later this week.

April 20, 2011 at 05:32 am - Permalink

As promised, here is a dell workstation with the 6-core i7 in it.

**** test on Professional Service Pack 1 (Build 7601)6.1.7601 using 6.21 and 121 passes; i7-980X
Create new graph time: 56.38ms, relative speed= 5.31
big data update time: 57.46ms, relative speed= 2.49
curve fit time: 881.71µs, relative speed= 2.14
user curve fit time: 6.98ms, relative speed= 5.08
double complex fft time: 569.89µs, relative speed= 2.58
single complex fft time: 534.00µs, relative speed= 2.22
double real fft time: 283.55µs, relative speed= 1.99
single real fft time: 259.83µs, relative speed= 1.89
5 pass smooth time: 223.35µs, relative speed= 2.43
Sort 8192 points time: 8.93ms, relative speed= 4.28
WaveStats time: 383.77µs, relative speed= 0.70
simple eqn time: 322.31µs, relative speed= 4.33
exp eqn time: 493.32µs, relative speed= 4.08
sqrt eqn time: 429.01µs, relative speed= 4.12
sin eqn time: 443.55µs, relative speed= 2.62
User fit fctn time: 242.02µs, relative speed= 5.34
MatrixOp eqn time: 15.18µs, relative speed= 1.57
**** done ****
total test time= 20.8849

April 20, 2011 at 05:31 am - Permalink

I was also able to shave about 10s off the total time on a MacBook Pro by hiding Igor during the test.

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 19, 2011 at 04:27 pm - Permalink

**** test on Windows XP Professional Service Pack 3 (Build 2600)5.1.2600 using 6.22 and 121 passes; AMD Dual X2 3800+ 2GB RAM
Create new graph time: 201.20ms, relative speed= 1.49
big data update time: 134.42ms, relative speed= 1.07
curve fit time: 1.80ms, relative speed= 1.05
user curve fit time: 17.82ms, relative speed= 1.99
double complex fft time: 1.58ms, relative speed= 0.93
single complex fft time: 1.16ms, relative speed= 1.02
double real fft time: 550.47µs, relative speed= 1.03
single real fft time: 591.89µs, relative speed= 0.83
5 pass smooth time: 507.10µs, relative speed= 1.07
Sort 8192 points time: 19.62ms, relative speed= 1.95
WaveStats time: 287.52µs, relative speed= 0.93
simple eqn time: 856.57µs, relative speed= 1.63
exp eqn time: 1.30ms, relative speed= 1.54
sqrt eqn time: 1.12ms, relative speed= 1.58
sin eqn time: 906.11µs, relative speed= 1.28
User fit fctn time: 547.45µs, relative speed= 2.36
MatrixOp eqn time: 41.59µs, relative speed= 0.57
**** done ****
total test time= 60.3425

I've got the highest test time! (And a pretty old machine too.)

April 20, 2011 at 01:14 am - Permalink

Old laptop (dual core Centrino) running Linux, Igor under Wine. No idea why Igor thinks it's running on a G5 - I have the Windows version of Igor installed.

**** test on Professional (Build 7600)6.1.7600 using 6.20 and 121 passes; Dual 2 GHz PowerPC G5 Mac Pro
Create new graph time: 769.24ms, relative speed= 0.39
big data update time: 134.70ms, relative speed= 1.06
curve fit time: 15.96ms, relative speed= 0.12
user curve fit time: 19.53ms, relative speed= 1.82
double complex fft time: 939.57µs, relative speed= 1.57
single complex fft time: 869.26µs, relative speed= 1.37
double real fft time: 459.42µs, relative speed= 1.23
single real fft time: 416.98µs, relative speed= 1.18
5 pass smooth time: 330.39µs, relative speed= 1.64
Sort 8192 points time: 14.38ms, relative speed= 2.66
WaveStats time: 647.14µs, relative speed= 0.41
simple eqn time: 334.59µs, relative speed= 4.17
exp eqn time: 647.36µs, relative speed= 3.11
sqrt eqn time: 485.81µs, relative speed= 3.64
sin eqn time: 646.08µs, relative speed= 1.80
User fit fctn time: 369.56µs, relative speed= 3.50
MatrixOp eqn time: 123.93µs, relative speed= 0.19
**** done ****
total test time= 156.86

April 20, 2011 at 04:06 am - Permalink

The processor type info is entered by the user when you run the test. A dual G5 is the default and is what was used to establish the baseline of 1.0x speed.
I updated my posts because I mis-identified the processor I Was testing (it is a 980X)

I threw all the data here on a graph, along with some other benchmarks I've made and have some questions about consistency of the benchmark...

First of all, the first part of the test takes the longest amount of time to complete and is graphics intensive. The time to complete the benchmark isn't a real measure of only the cpu performance, but of the graphics system, which is strongly both OS and gpu dependent. Because of this, you can't really compare CPU vs. CPU based solely on the time to complete the test, especially across platforms.

The desktop machines reported(i980x, phenom) lagged behind severely in the wavestats test, while the laptops excelled... Why is this, I wonder?

My i980x should theoretically be mopping the floor with my laptop i7, as it has roughly 50% higher clock speed per core and about 50% the memory bandwidth… However, it only wins out in 4 of the 16 tests, one of which is the graphics test...

Perhaps it is time to update the benchmark with some multi-threaded tests?

April 20, 2011 at 05:37 am - Permalink

Core i7 860 @2.8 GHz; 4 GB Ram; NVIDIA Quadro NVS 290
64bit Win7 with 32 Bit Igor Igor 6.21

32.1 s with the window open vs 27.5 s with Igor minimized - see below

Igor window Open:
**** test on Professional (Build 7600)6.1.7600 using 6.21 and 121 passes; Core i7 860 @ 2.8 GHz 64bit Win7 4 Gb Ram
Create new graph time: 113.80ms, relative speed= 2.63
big data update time: 67.45ms, relative speed= 2.12
curve fit time: 899.39µs, relative speed= 2.10
user curve fit time: 7.12ms, relative speed= 4.98
double complex fft time: 577.78µs, relative speed= 2.55
single complex fft time: 551.85µs, relative speed= 2.15
double real fft time: 285.08µs, relative speed= 1.98
single real fft time: 262.26µs, relative speed= 1.88
5 pass smooth time: 214.49µs, relative speed= 2.53
Sort 8192 points time: 8.73ms, relative speed= 4.38
WaveStats time: 369.70µs, relative speed= 0.72
simple eqn time: 363.76µs, relative speed= 3.84
exp eqn time: 471.47µs, relative speed= 4.26
sqrt eqn time: 436.31µs, relative speed= 4.05
sin eqn time: 432.44µs, relative speed= 2.69
User fit fctn time: 242.62µs, relative speed= 5.33
MatrixOp eqn time: 14.58µs, relative speed= 1.64
**** done ****
total test time= 32.122


Igor Minimized:
**** test on Professional (Build 7600)6.1.7600 using 6.21 and 121 passes; Core i7 860 @ 2.8 GHz 64bit Win7 4 Gb Ram
Create new graph time: 100.13ms, relative speed= 2.99
big data update time: 67.06ms, relative speed= 2.13
curve fit time: 897.49µs, relative speed= 2.10
user curve fit time: 7.13ms, relative speed= 4.97
double complex fft time: 577.56µs, relative speed= 2.55
single complex fft time: 541.11µs, relative speed= 2.20
double real fft time: 285.11µs, relative speed= 1.98
single real fft time: 262.05µs, relative speed= 1.88
5 pass smooth time: 221.16µs, relative speed= 2.45
Sort 8192 points time: 8.70ms, relative speed= 4.39
WaveStats time: 379.42µs, relative speed= 0.70
simple eqn time: 363.05µs, relative speed= 3.84
exp eqn time: 471.47µs, relative speed= 4.26
sqrt eqn time: 437.58µs, relative speed= 4.04
sin eqn time: 432.42µs, relative speed= 2.69
User fit fctn time: 242.58µs, relative speed= 5.33
MatrixOp eqn time: 14.60µs, relative speed= 1.64
**** done ****
total test time= 27.4538

April 22, 2011 at 08:21 am - Permalink

on Mac 10.67, IgorPro 6.21 and Macbookpro core7 8 Gb Ram

32 sec in front, 23 in background ....

April 27, 2011 at 12:18 am - Permalink

I discovered this morning that globally disabling vsync on my mac leads to a major speedup in the graphical part of the benchmark.

To do this, install the apple developer tools (Xcode) located on your os X install disc. Then, use the Quartz Debug application located here:

/Developer/Applications/Performance Tools/Quartz Debug

Click Window-> Quartz Debug Settings.

Choose Beam Sync -> Disable

The reason for the speedup is that vsync only allows one frame to be drawn per refresh. My laptop defaults to 60 Hz, so the maximum number of frames per second is 60. Indeed, during the benchmark, I was pulling 57-59 fps (checked by at-monitor). After disabling vsync, I was seeing numbers between 320-340 fps instantaneous and 160 fps average.

All other parts of the test were unaffected by this, so if you need to draw graphs very quickly, you might consider disabling vsync and see if it gives any performance improvement.

Create new graph time: 122.34ms, relative speed= 2.45 (Vsync automatic)
Create new graph time: 81.75ms, relative speed= 3.66 (Vsync off); 49% faster

Edit:
Additional gains were realized by disabling 2D acceleration.
Create new graph time: 77.01ms, relative speed= 3.89 (Vsync off, 2D Acceleration off); 58% faster
And surprising marginal gains were again seen by forcing the cpu integrated graphics using gfxCardStatus 2.01.
Create new graph time: 74.85ms, relative speed= 4.00 (Vsync off, 2D Acceleration off, int. gpu forced); 63% faster

June 1, 2011 at 05:08 am - Permalink

12-core Xeon Mac Pro with an 8-core process running simultaneously + one additional single-core process running as well

**** test on Macintosh OS X10.7.2 using 6.22 and 121 passes; 2 x 2.93 GHz 6-Core Xeon Mac Pro, 32 GB
Create new graph time: 126.72ms, relative speed= 2.36
big data update time: 61.70ms, relative speed= 2.32
curve fit time: 751.68µs, relative speed= 2.51
user curve fit time: 7.11ms, relative speed= 4.99
double complex fft time: 491.70µs, relative speed= 2.99
single complex fft time: 499.29µs, relative speed= 2.38
double real fft time: 239.37µs, relative speed= 2.36
single real fft time: 226.45µs, relative speed= 2.17
5 pass smooth time: 290.15µs, relative speed= 1.87
Sort 8192 points time: 12.36ms, relative speed= 3.09
WaveStats time: 139.59µs, relative speed= 1.91
simple eqn time: 366.47µs, relative speed= 3.81
exp eqn time: 547.17µs, relative speed= 3.67
sqrt eqn time: 400.80µs, relative speed= 4.41
sin eqn time: 312.43µs, relative speed= 3.72
User fit fctn time: 240.22µs, relative speed= 5.38
MatrixOp eqn time: 12.96µs, relative speed= 1.84
**** done ****
total test time= 29.9256

With Igor hidden (which is cheating of course)

**** test on Macintosh OS X10.7.2 using 6.22 and 121 passes; 2 x 2.93 GHz 6-Core Xeon Mac Pro, 32 GB
Create new graph time: 73.08ms, relative speed= 4.10
big data update time: 61.36ms, relative speed= 2.33
curve fit time: 746.97µs, relative speed= 2.53
user curve fit time: 7.27ms, relative speed= 4.88
double complex fft time: 492.37µs, relative speed= 2.99
single complex fft time: 498.94µs, relative speed= 2.38
double real fft time: 239.61µs, relative speed= 2.36
single real fft time: 225.67µs, relative speed= 2.18
5 pass smooth time: 288.36µs, relative speed= 1.88
Sort 8192 points time: 12.35ms, relative speed= 3.10
WaveStats time: 139.36µs, relative speed= 1.91
simple eqn time: 362.95µs, relative speed= 3.85
exp eqn time: 546.15µs, relative speed= 3.68
sqrt eqn time: 403.36µs, relative speed= 4.39
sin eqn time: 312.17µs, relative speed= 3.73
User fit fctn time: 239.06µs, relative speed= 5.41
MatrixOp eqn time: 12.87µs, relative speed= 1.86
**** done ****
total test time= 22.327

November 4, 2011 at 05:15 pm - Permalink

The other computer at my desk...sitting idle except for this test

**** test on Professional Service Pack 2 (Build 3790)5.2.3790 using 6.22 and 121 passes; 2.60 GHz AMD Phenom 9950, 16 GB, Win XP 64
Create new graph time: 122.29ms, relative speed= 2.45
big data update time: 85.79ms, relative speed= 1.67
curve fit time: 1.32ms, relative speed= 1.43
user curve fit time: 9.89ms, relative speed= 3.59
double complex fft time: 1.03ms, relative speed= 1.43
single complex fft time: 865.34µs, relative speed= 1.37
double real fft time: 422.99µs, relative speed= 1.33
single real fft time: 439.24µs, relative speed= 1.12
5 pass smooth time: 341.78µs, relative speed= 1.59
Sort 8192 points time: 13.52ms, relative speed= 2.83
WaveStats time: 185.53µs, relative speed= 1.44
simple eqn time: 612.97µs, relative speed= 2.28
exp eqn time: 941.52µs, relative speed= 2.14
sqrt eqn time: 790.28µs, relative speed= 2.24
sin eqn time: 679.69µs, relative speed= 1.71
User fit fctn time: 371.28µs, relative speed= 3.48
MatrixOp eqn time: 26.01µs, relative speed= 0.92
**** done ****
total test time= 39.499

With Igor minimized this drops to 28.7802 s. Probably because the "Create new graph time" is cut almost in half.

November 4, 2011 at 05:13 pm - Permalink

Other factors...

Any processor with turbo-boost is inherently difficult to benchmark. Turbo boost is a technology that automatically over clocks your processor by steps, depending on the temperature, power consumption and number of active cores. Because of this issue, you may not be running full speed as you think you are. I have found, however, that my simulations finish quicker if I just point a fan at my laptop while it is grinding away.

A nice overview of turbo-boost can be found on wikipedia. I have attached my latest benchmark below, with integrated graphics forced, disabled 2D acceleration and beam sync via quartz debug.

**** test on Macintosh OS X10.7.2 using 6.22 and 121 passes; 2.2 GHz Quad core MBP 8,2, 8 GB ram
Create new graph time: 85.48ms, relative speed= 3.50
big data update time: 54.32ms, relative speed= 2.64
curve fit time: 650.88µs, relative speed= 2.90
user curve fit time: 7.35ms, relative speed= 4.82
double complex fft time: 423.66µs, relative speed= 3.48
single complex fft time: 407.57µs, relative speed= 2.91
double real fft time: 207.15µs, relative speed= 2.72
single real fft time: 199.60µs, relative speed= 2.47
5 pass smooth time: 260.13µs, relative speed= 2.09
Sort 8192 points time: 11.27ms, relative speed= 3.39
WaveStats time: 138.01µs, relative speed= 1.93
simple eqn time: 269.35µs, relative speed= 5.18
exp eqn time: 402.24µs, relative speed= 5.00
sqrt eqn time: 299.86µs, relative speed= 5.90
sin eqn time: 283.29µs, relative speed= 4.11
User fit fctn time: 205.88µs, relative speed= 6.28
MatrixOp eqn time: 11.72µs, relative speed= 2.04
**** done ****
total test time= 23.1571

November 7, 2011 at 08:16 am - Permalink

Some numbers of my new machine. As the previois poster noted CPUs with Turbo Boost are difficult to benchmark properly. All benchmarks were acquired watching Igor drawing.

Igor (64-bit)
**** test on Win7 Professional Service Pack 1 (Build 7601)6.1.7601 using 6.23 and 121 passes; i7-3930K 6 cores Win7 x64 Igor x64, 32GB RAM Create new graph time: 64.89ms, relative speed= 4.61 big data update time: 39.63ms, relative speed= 3.61 curve fit time: 506.93µs, relative speed= 3.72 user curve fit time: 3.73ms, relative speed= 9.52 double complex fft time: 372.70µs, relative speed= 3.95 single complex fft time: 348.10µs, relative speed= 3.41 double real fft time: 179.75µs, relative speed= 3.14 single real fft time: 172.00µs, relative speed= 2.86 5 pass smooth time: 163.65µs, relative speed= 3.32 Sort 8192 points time: 7.39ms, relative speed= 5.17 WaveStats time: 81.90µs, relative speed= 3.26 simple eqn time: 171.18µs, relative speed= 8.15 exp eqn time: 278.05µs, relative speed= 7.23 sqrt eqn time: 207.41µs, relative speed= 8.53 sin eqn time: 212.89µs, relative speed= 5.46 User fit fctn time: 136.67µs, relative speed= 9.46 MatrixOp eqn time: 14.07µs, relative speed= 1.70 **** done **** total test time= 18.1146

Igor (32-bit)
**** test on Professional Service Pack 1 (Build 7601)6.1.7601 using 6.22 and 121 passes; i7-3930K 6 cores Win7 x64 Igor x32, 32GB RAM Create new graph time: 75.43ms, relative speed= 3.97 big data update time: 47.65ms, relative speed= 3.00 curve fit time: 604.79µs, relative speed= 3.12 user curve fit time: 4.85ms, relative speed= 7.31 double complex fft time: 464.54µs, relative speed= 3.17 single complex fft time: 342.43µs, relative speed= 3.47 double real fft time: 300.30µs, relative speed= 1.88 single real fft time: 181.83µs, relative speed= 2.71 5 pass smooth time: 189.11µs, relative speed= 2.87 Sort 8192 points time: 7.62ms, relative speed= 5.02 WaveStats time: 94.29µs, relative speed= 2.83 simple eqn time: 169.04µs, relative speed= 8.26 exp eqn time: 275.59µs, relative speed= 7.29 sqrt eqn time: 211.72µs, relative speed= 8.35 sin eqn time: 224.09µs, relative speed= 5.19 User fit fctn time: 148.07µs, relative speed= 8.73 MatrixOp eqn time: 13.62µs, relative speed= 1.75 **** done **** total test time= 21.2596

Igor 64bit with data from Igor 32bit as baseline
**** test on Professional Service Pack 1 (Build 7601)6.1.7601 using 6.23 and 121 passes; Create new graph time: 65.97ms, relative speed= 1.14 big data update time: 39.63ms, relative speed= 1.20 curve fit time: 504.63µs, relative speed= 1.19 user curve fit time: 3.71ms, relative speed= 1.29 double complex fft time: 373.09µs, relative speed= 1.27 single complex fft time: 348.21µs, relative speed= 1.00 double real fft time: 180.32µs, relative speed= 1.63 single real fft time: 174.40µs, relative speed= 1.02 5 pass smooth time: 166.75µs, relative speed= 1.14 Sort 8192 points time: 7.39ms, relative speed= 1.03 WaveStats time: 85.07µs, relative speed= 1.11 simple eqn time: 156.02µs, relative speed= 1.08 exp eqn time: 267.04µs, relative speed= 1.05 sqrt eqn time: 187.18µs, relative speed= 1.13 sin eqn time: 207.08µs, relative speed= 1.08 User fit fctn time: 131.58µs, relative speed= 1.14 MatrixOp eqn time: 14.07µs, relative speed= 0.89 **** done **** total test time= 17.7826

February 28, 2012 at 01:33 pm - Permalink

I got a new core i7 ivy bridge machine...

**** test on WIN 7 using 6.30 and 121 passes; Core i7 3770k
Create new graph time: 37.44ms, relative speed= 8.00
big data update time: 35.44ms, relative speed= 4.04
curve fit time: 800.36µs, relative speed= 2.36
user curve fit time: 3.60ms, relative speed= 9.84
double complex fft time: 334.91µs, relative speed= 4.40
single complex fft time: 304.25µs, relative speed= 3.90
double real fft time: 158.36µs, relative speed= 3.56
single real fft time: 150.02µs, relative speed= 3.28
5 pass smooth time: 150.45µs, relative speed= 3.61
Sort 8192 points time: 6.84ms, relative speed= 5.59
WaveStats time: 80.29µs, relative speed= 3.32
simple eqn time: 139.34µs, relative speed= 10.02
exp eqn time: 229.45µs, relative speed= 8.76
sqrt eqn time: 177.05µs, relative speed= 9.99
sin eqn time: 176.52µs, relative speed= 6.59
User fit fctn time: 111.33µs, relative speed= 11.61
MatrixOp eqn time: 12.62µs, relative speed= 1.89
**** done ****
total test time= 12.759

July 20, 2012 at 03:17 pm - Permalink

BMangum wrote:
**** test on WIN 7 using 6.30 and 121 passes; Core i7 3770k
curve fit time: 800.36µs, relative speed= 2.36

If you get the latest build you will get the fix for this slow-down that I introduced when I fixed a bug in curve fit threading.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

July 20, 2012 at 04:01 pm - Permalink

With Igor Pro in front and command window visible ...

**** test on Macintosh OS X10.9.2 using 6.34 and 121 passes
**** MacBook Pro Late 2013 2.6 GHz I7 16GB RAM OS 10.9.2
Create new graph time: 118.84ms, relative speed= 2.52
big data update time: 42.63ms, relative speed= 3.36
curve fit time: 518.17µs, relative speed= 3.64
user curve fit time: 4.39ms, relative speed= 8.08
double complex fft time: 322.54µs, relative speed= 4.57
single complex fft time: 298.06µs, relative speed= 3.99
double real fft time: 150.95µs, relative speed= 3.74
single real fft time: 142.81µs, relative speed= 3.45
5 pass smooth time: 207.85µs, relative speed= 2.61
Sort 8192 points time: 8.88ms, relative speed= 4.30
WaveStats time: 97.10µs, relative speed= 2.75
simple eqn time: 174.01µs, relative speed= 8.02
exp eqn time: 280.15µs, relative speed= 7.18
sqrt eqn time: 192.89µs, relative speed= 9.17
sin eqn time: 220.06µs, relative speed= 5.28
User fit fctn time: 140.25µs, relative speed= 9.22
MatrixOp eqn time: 10.79µs, relative speed= 2.21
**** done ****
total test time= 24.2226

With command window closed and Igor Pro put in background as quickly as possible ...

**** test on Macintosh OS X10.9.2 using 6.34 and 121 passes
**** MacBook Pro Late 2013 2.6 GHz I7 16GB RAM OS 10.9.2
Create new graph time: 56.47ms, relative speed= 5.30
big data update time: 41.74ms, relative speed= 3.43
curve fit time: 518.24µs, relative speed= 3.64
user curve fit time: 4.40ms, relative speed= 8.06
double complex fft time: 328.25µs, relative speed= 4.49
single complex fft time: 299.51µs, relative speed= 3.97
double real fft time: 154.07µs, relative speed= 3.66
single real fft time: 143.35µs, relative speed= 3.43
5 pass smooth time: 207.10µs, relative speed= 2.62
Sort 8192 points time: 8.83ms, relative speed= 4.33
WaveStats time: 96.47µs, relative speed= 2.77
simple eqn time: 174.03µs, relative speed= 8.02
exp eqn time: 280.48µs, relative speed= 7.17
sqrt eqn time: 192.81µs, relative speed= 9.17
sin eqn time: 219.49µs, relative speed= 5.30
User fit fctn time: 138.19µs, relative speed= 9.35
MatrixOp eqn time: 9.66µs, relative speed= 2.47
**** done ****
total test time= 16.2594

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 10, 2014 at 03:09 pm - Permalink

@jjweimer: Can you repost with 121 iterations instead of 21? This would greatly simplify the comparison :)

April 10, 2014 at 09:53 am - Permalink

thomas_braun wrote:
@jjweimer: Can you repost with 121 iterations instead of 21? This would greatly simplify the comparison :)


Did it!

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 10, 2014 at 03:18 pm - Permalink

johnweeks wrote:
I presume that you posted this to show the contrast in Create New Graph time.


Yes.

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAHuntsville

April 11, 2014 at 11:06 am - Permalink

**** test on Macintosh OS X10.9.3 using 6.34 and 121 passes; late 2012 mac mini Quad Core 2.6 GHz core i7, 16GB ram, SSD, intel 4000 gfx
Create new graph time: 126.15ms, relative speed= 2.37
big data update time: 49.75ms, relative speed= 2.88
curve fit time: 577.79µs, relative speed= 3.26
user curve fit time: 5.43ms, relative speed= 6.54
double complex fft time: 352.52µs, relative speed= 4.18
single complex fft time: 336.69µs, relative speed= 3.53
double real fft time: 181.27µs, relative speed= 3.11
single real fft time: 174.31µs, relative speed= 2.82
5 pass smooth time: 228.26µs, relative speed= 2.38
Sort 8192 points time: 9.92ms, relative speed= 3.86
WaveStats time: 118.96µs, relative speed= 2.24
simple eqn time: 203.24µs, relative speed= 6.87
exp eqn time: 318.22µs, relative speed= 6.32
sqrt eqn time: 233.84µs, relative speed= 7.56
sin eqn time: 240.80µs, relative speed= 4.83
User fit fctn time: 162.32µs, relative speed= 7.96
MatrixOp eqn time: 10.57µs, relative speed= 2.26
**** done ****
total test time= 26.9465

June 11, 2014 at 07:37 pm - Permalink

Just updating a new data point with a current generation CPU

32 bit Igor
**** test on Windows 8 Professional (Build 9200)6.2.9200 using 6.34 and 121 passes; i7 4970k @ 4.9Ghz
Create new graph time: 52.44ms, relative speed= 5.71
big data update time: 32.97ms, relative speed= 4.34
curve fit time: 419.00µs, relative speed= 4.50
user curve fit time: 3.15ms, relative speed= 11.25
double complex fft time: 345.18µs, relative speed= 4.27
single complex fft time: 240.33µs, relative speed= 4.94
double real fft time: 164.28µs, relative speed= 3.44
single real fft time: 121.42µs, relative speed= 4.05
5 pass smooth time: 101.68µs, relative speed= 5.34
Sort 8192 points time: 4.82ms, relative speed= 7.93
WaveStats time: 60.87µs, relative speed= 4.38
simple eqn time: 118.00µs, relative speed= 11.83
exp eqn time: 191.99µs, relative speed= 10.47
sqrt eqn time: 144.75µs, relative speed= 12.22
sin eqn time: 157.84µs, relative speed= 7.37
User fit fctn time: 98.34µs, relative speed= 13.14
MatrixOp eqn time: 10.86µs, relative speed= 2.20
**** done ****
total test time= 13.6788

64-bit Igor
**** test on Windows 8 Professional (Build 9200)6.2.9200 using 6.34 and 121 passes; i7 4970k @ 4.9Ghz
Create new graph time: 47.80ms, relative speed= 6.26
big data update time: 27.30ms, relative speed= 5.25
curve fit time: 1.93ms, relative speed= 0.98
user curve fit time: 2.61ms, relative speed= 13.60
double complex fft time: 277.13µs, relative speed= 5.31
single complex fft time: 243.02µs, relative speed= 4.89
double real fft time: 131.44µs, relative speed= 4.29
single real fft time: 124.43µs, relative speed= 3.95
5 pass smooth time: 113.22µs, relative speed= 4.79
Sort 8192 points time: 4.47ms, relative speed= 8.55
WaveStats time: 49.68µs, relative speed= 5.37
simple eqn time: 107.31µs, relative speed= 13.01
exp eqn time: 213.34µs, relative speed= 9.42
sqrt eqn time: 125.92µs, relative speed= 14.05
sin eqn time: 134.37µs, relative speed= 8.66
User fit fctn time: 91.95µs, relative speed= 14.06
MatrixOp eqn time: 12.12µs, relative speed= 1.97
**** done ****
total test time= 12.3603

Even though the 64-bit version outperformed the 32-bit version overall, it is curious that it falls flat on the curve fit test. Does anyone have an ideal why this might be?

September 10, 2014 at 10:00 pm - Permalink

mtaylor wrote:
Even though the 64-bit version outperformed the 32-bit version overall, it is curious that it falls flat on the curve fit test. Does anyone have an ideal why this might be?

That is odd.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

September 11, 2014 at 09:08 am - Permalink


hmm, the comparison between IP6 and IP7 surprised me a bit!

**** test on Macintosh OS X10.11.6 using 6.37 and 121 passes; iMac (Retina 5K, 27-inch, Late 2015)
Create new graph time: 164.28ms, relative speed= 1.82
big data update time: 40.53ms, relative speed= 3.53
curve fit time: 563.94µs, relative speed= 3.34
user curve fit time: 4.48ms, relative speed= 7.92
double complex fft time: 315.65µs, relative speed= 4.67
single complex fft time: 278.30µs, relative speed= 4.27
double real fft time: 148.55µs, relative speed= 3.80
single real fft time: 133.43µs, relative speed= 3.69
5 pass smooth time: 201.60µs, relative speed= 2.69
Sort 8192 points time: 8.40ms, relative speed= 4.55
WaveStats time: 105.02µs, relative speed= 2.54
simple eqn time: 173.16µs, relative speed= 8.06
exp eqn time: 306.49µs, relative speed= 6.56
sqrt eqn time: 193.90µs, relative speed= 9.12
sin eqn time: 230.50µs, relative speed= 5.05
User fit fctn time: 132.54µs, relative speed= 9.75
MatrixOp eqn time: 19.50µs, relative speed= 1.23
**** done ****
total test time= 30.6379


**** test on Macintosh OS X10.11.6 using 7.00 and 121 passes; iMac (Retina 5K, 27-inch, Late 2015)
Create new graph time: 223.85ms, relative speed= 1.34
big data update time: 155.12ms, relative speed= 0.92
curve fit time: 221.20µs, relative speed= 8.53
user curve fit time: 3.59ms, relative speed= 9.89
double complex fft time: 359.19µs, relative speed= 4.10
single complex fft time: 337.72µs, relative speed= 3.52
double real fft time: 179.87µs, relative speed= 3.14
single real fft time: 165.10µs, relative speed= 2.98
5 pass smooth time: 117.49µs, relative speed= 4.62
Sort 8192 points time: 6.37ms, relative speed= 6.01
WaveStats time: 36.86µs, relative speed= 7.24
simple eqn time: 194.48µs, relative speed= 7.18
exp eqn time: 280.49µs, relative speed= 7.17
sqrt eqn time: 219.87µs, relative speed= 8.04
sin eqn time: 221.08µs, relative speed= 5.26
User fit fctn time: 119.42µs, relative speed= 10.82
MatrixOp eqn time: 21.62µs, relative speed= 1.11
**** done ****
total test time= 55.7927

August 25, 2016 at 01:16 am - Permalink

ChrLie wrote:
hmm, the comparison between IP6 and IP7 surprised me a bit!
...


The biggest slow downs seem to be in the graphics operations. Perhaps not surprising since IP7 supports Retina?

Modest slow downs in some operations are interesting.

Big speed ups in MatrixOP and other functions are good news.

Thanks.

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAH

August 25, 2016 at 07:14 am - Permalink

jjweimer wrote:

The biggest slow downs seem to be in the graphics operations. Perhaps not surprising since IP7 supports Retina?


You're right! This is what Wavemetrics says:

"Macintosh Retina and Windows high-resolution displays bring a more pleasing visual experience but also present performance challenges. Previously, one point lines were one pixel wide. Now, on high-resolution displays, such lines are two pixels wide and, depending on the operating system, graphics technology and the actual data, drawing can be thousands of times slower."

August 25, 2016 at 08:32 am - Permalink


**** test on Windows 7 Professional6.1.7601 using 7.01 and 121 passes; Dual 2 GHz PowerPC G5 Mac Pro
Create new graph time: 216.92ms, relative speed= 1.38
big data update time: 118.53ms, relative speed= 1.21
curve fit time: 283.17µs, relative speed= 6.66
user curve fit time: 4.66ms, relative speed= 7.61
double complex fft time: 326.21µs, relative speed= 4.51
single complex fft time: 296.96µs, relative speed= 4.00
double real fft time: 157.68µs, relative speed= 3.58
single real fft time: 142.47µs, relative speed= 3.45
5 pass smooth time: 117.64µs, relative speed= 4.61
Sort 8192 points time: 9.04ms, relative speed= 4.23
WaveStats time: 103.43µs, relative speed= 2.58
simple eqn time: 206.61µs, relative speed= 6.76
exp eqn time: 293.18µs, relative speed= 6.86
sqrt eqn time: 245.06µs, relative speed= 7.22
sin eqn time: 245.11µs, relative speed= 4.74
User fit fctn time: 188.31µs, relative speed= 6.86
MatrixOp eqn time: 28.77µs, relative speed= 0.83
**** done ****
total test time= 47.7564


And now with testing IP6 using the IP7 measurement as baseline.


**** test on Windows 7 Professional Service Pack 1 (Build 7601)6.1.7601 using 6.38 and 121 passes; Dual 2 GHz PowerPC G5 Mac Pro
Create new graph time: 102.42ms, relative speed= 2.16
big data update time: 46.29ms, relative speed= 2.56
curve fit time: 619.97µs, relative speed= 0.47
user curve fit time: 5.07ms, relative speed= 0.92
double complex fft time: 470.59µs, relative speed= 0.68
single complex fft time: 339.95µs, relative speed= 0.88
double real fft time: 292.47µs, relative speed= 0.54
single real fft time: 169.93µs, relative speed= 0.85
5 pass smooth time: 173.53µs, relative speed= 0.73
Sort 8192 points time: 7.17ms, relative speed= 1.27
WaveStats time: 100.26µs, relative speed= 1.17
simple eqn time: 169.21µs, relative speed= 1.22
exp eqn time: 283.68µs, relative speed= 1.04
sqrt eqn time: 215.74µs, relative speed= 1.13
sin eqn time: 227.13µs, relative speed= 1.08
User fit fctn time: 168.51µs, relative speed= 1.10
MatrixOp eqn time: 26.51µs, relative speed= 1.10
**** done ****
total test time= 26.6648


The machine is the same as in http://www.igorexchange.com/node/1489#comment-4988.

November 28, 2016 at 11:50 am - Permalink