Stata/MP 18 slower than Stata/MP 17? Testing Stata speed (version 17 vs 18)

[Abdullah Algarni]

Hi everyone,
I just upgraded to Stata/MP 18, and I present here the comparison in performance between Stata/MP 17 vs. 18, 4 core license.

My computer:

• Apple silicon M1 (Total Number of Cores: 8 [4 performance {3220 MHz} and 4 efficiency {2064 MHz}])
• Memory (RAM: 16 GB; and 1 TB hard disk).

Code:

• I used the code from this post on Statalist.

Conclusion:

• Stata/MP 18 seems slower than Stata/MP 17, Am I right?

Using one core

set processors 1
The maximum number of processors or cores being used is changed from 4 to 1.  It can be set to any number between 1 and 4

• Stata/MP 17

. tabstat replace regress predict correl bootstrap mvtest xtile expand_drop arfima eigenv , stats(median mean sd min max) columns(s)


    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .012     .0124  .0005164      .012      .013
     regress |     .0705     .0704  .0006992      .069      .071
     predict |      .021     .0214   .001075       .02      .024
      correl |      .056     .0577  .0064472      .055      .076
   bootstrap |    6.5075     6.513  .0452769      6.46     6.598
      mvtest |      .187      .186  .0048305      .173      .191
       xtile |      .433     .4309  .0153076      .391      .445
 expand_drop |     .0415     .0421   .002331      .039      .046
      arfima |  104.1265  305.3392  636.4154   103.715  2116.609
      eigenv |      .214     .2142  .0013166      .211      .216
----------------------------------------------------------------

• Stata/MP 18

. tabstat replace regress predict correl bootstrap mvtest xtile expand_drop arfima eigenv , stats(median mean sd min max) columns(s)

    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .014     .0143  .0014181      .013      .018
     regress |      .086     .0881  .0049542      .085      .101
     predict |      .024     .0241  .0007379      .023      .025
      correl |      .066     .0661  .0007379      .065      .067
   bootstrap |     7.286    7.3156  .1050039     7.218     7.594
      mvtest |     .1855     .1819  .0146322      .165      .197
       xtile |     .4445     .4405   .034394      .399      .481
 expand_drop |      .042     .0419  .0011972      .041      .045
      arfima |  137.8045  142.1306  16.53265   132.644    188.13
      eigenv |      .214     .2142  .0006325      .213      .215
----------------------------------------------------------------

Using four cores

set processors 4
The maximum number of processors or cores being used is changed from 1 to 4.  It can be set to any number between 1 and 4

• Stata/MP 17

. tabstat replace regress predict correl bootstrap mvtest xtile expand_drop arfima eigenv , stats(median mean sd min max) columns(s)

    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .003     .0033  .0013375      .002      .007
     regress |      .019     .0192  .0009189      .018      .021
     predict |      .006      .006  .0008165      .005      .008
      correl |      .011     .0119  .0021833      .011      .018
   bootstrap |      3.97    3.9837  .0408223     3.941     4.064
      mvtest |      .065     .0659  .0015951      .064      .069
       xtile |      .321     .3209  .0088122      .299      .333
 expand_drop |      .041     .0408  .0006325       .04      .042
      arfima |    86.532   86.7834  .9220214    85.671    88.297
      eigenv |      .212     .2126  .0017764      .211      .217
----------------------------------------------------------------

• Stata/MP 18

. tabstat replace regress predict correl bootstrap mvtest xtile expand_drop arfima eigenv , stats(median mean sd min max) columns(s)

    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .003      .003         0      .003      .003
     regress |      .024     .0241  .0003162      .024      .025
     predict |      .007     .0073  .0006749      .007      .009
      correl |      .015     .0153  .0013375      .014      .019
   bootstrap |     4.283     4.346  .1862484     4.225     4.854
      mvtest |     .0685     .0645  .0070435      .055      .071
       xtile |     .3275     .3214  .0185784      .294      .342
 expand_drop |      .041      .041  .0010541      .039      .043
      arfima |   118.605  118.6945  .4226059   118.274   119.682
      eigenv |      .213     .2143   .003401      .212      .221
----------------------------------------------------------------

[Jared Greathouse]

How fast is it exactly? Perhaps set a timer for each to compare?

[Leonardo Guizzetti]

How fast is it exactly? Perhaps set a timer for each to compare?

Those tables present a summary of timings for various commands, if I have understood correctly.

[Abdullah Algarni]

Those tables present a summary of timings for various commands, if I have understood correctly.

You are right Leonardo Guizzetti , they present a summary of timing for various commands.

Frustrating 😞

[Abdullah Algarni]

How fast is it exactly? Perhaps set a timer for each to compare?

Hi Jared Greathouse

A detailed discussion of the code is presented here

[Chinh Nguyen (StataCorp)]

Hi everyone,
I just upgraded to Stata/MP 18, and I present here the comparison in performance between Stata/MP 17 vs. 18, 4 core license.

• I used the code from this post on Statalist.

Conclusion:

• Stata/MP 18 seems slower than Stata/MP 17, Am I right?

There is something off with your setup because I'm not seeing that at all with a similar setup of an M1 Mac with 16 GB of memory as well as an M2 Pro with 64 GB of memory.

On my M1/16 GB that's stock with no third party software installed:

Stata/MP 17/1 processor:
    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .013     .0126  .0005164      .012      .013
     regress |      .069     .0694  .0005164      .069       .07
     predict |      .021     .0211  .0005676       .02      .022
      correl |      .055     .0558  .0025298      .055      .063
   bootstrap |    6.6755    6.6674  .0524471     6.577     6.739
      mvtest |      .185     .1836  .0064498      .167      .193
       xtile |     .4355     .4332  .0158941      .393      .458
 expand_drop |      .041     .0409  .0007379       .04      .042
      arfima |    52.879   52.8829   .038763    52.827    52.944
      eigenv |      .208     .2084  .0005164      .208      .209
----------------------------------------------------------------

Stata/MP 18/1 processor:
    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .013     .0134  .0006992      .013      .015
     regress |      .083     .0832  .0004216      .083      .084
     predict |      .023     .0233   .000483      .023      .024
      correl |      .065     .0654  .0020656      .064      .071
   bootstrap |     7.134    7.1382  .0103901     7.129     7.163
      mvtest |      .187     .1855  .0060415      .169      .191
       xtile |      .447     .4427  .0134002      .405      .451
 expand_drop |     .0415     .0414   .001075      .039      .043
      arfima |     53.22   53.2544  .2159849    52.986    53.707
      eigenv |      .209     .2089  .0005676      .208       .21
----------------------------------------------------------------

Stata/MP 17/4 processors:
    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .003     .0026  .0005164      .002      .003
     regress |      .019     .0186  .0005164      .018      .019
     predict |      .006     .0062  .0004216      .006      .007
      correl |      .011     .0118  .0015492      .011      .016
   bootstrap |    3.9775    3.9779  .0149105     3.957     4.013
      mvtest |     .0665     .0656   .003534      .056      .069
       xtile |     .3255     .3222  .0143201      .282      .331
 expand_drop |      .042      .042   .001633       .04      .046
      arfima |    47.628   47.6295  .0946932    47.508    47.818
      eigenv |      .208     .2084   .001075      .207      .211
----------------------------------------------------------------

Stata/MP 18/4 processors:
    Variable |       p50      Mean        SD       Min       Max
-------------+--------------------------------------------------
     replace |      .003     .0032  .0004216      .003      .004
     regress |      .023     .0227   .000483      .022      .023
     predict |      .007     .0067  .0006749      .006      .008
      correl |      .015     .0152  .0017512      .014       .02
   bootstrap |    4.1655    4.1643  .0148926     4.147       4.2
      mvtest |     .0695     .0685  .0038079      .058      .071
       xtile |      .333      .329  .0156702      .285       .34
 expand_drop |      .042     .0425  .0019579      .041      .048
      arfima |    47.143   47.1331  .1053238    47.004    47.365
      eigenv |     .2095     .2096  .0009661      .208      .211
----------------------------------------------------------------

I'd create a new user account and login to that account and rerun the tests. You might have something running in the background that's affecting the timings.