9
$\begingroup$

The built in benchmark test on my stationary workstation shows an improvement from 2.82 to 5.99, all other hardware and operating systems settings the same.

This looks very good. My question is: What are the key reasons/changes that has been done to the product? I can't see the same improvement on my laptop.

enter image description here

$\endgroup$
8
  • $\begingroup$ Is it right that there are different computer name and OS? $\endgroup$ May 22 at 16:00
  • $\begingroup$ No, it is the same computer. But in the benchmark today, the program decided to name otherwise - seems to be related to licensed processors (it is an 8 core processor i7-7820X) $\endgroup$
    – FredrikD
    May 22 at 16:19
  • $\begingroup$ ..and Windows10 $\endgroup$
    – FredrikD
    May 22 at 16:52
  • 1
    $\begingroup$ Just for curiosity, I did a test on my MacBook Pro: for total score, v12.2 gives 2.78 while v12.3 gives 3.30. Besides, I also feel the operations related to Dataset are generally more performant on v12.3. Exciting! $\endgroup$
    – sunt05
    May 22 at 19:58
  • 1
    $\begingroup$ The "4-node homogenous cluster" makes me suspicious of the benchmark - see this post (mathematica.stackexchange.com/a/235930/65858) and the comments below. It seems to indicate it is running on parallel kernels which artificially inflates the result. $\endgroup$
    – ala10
    May 23 at 15:36
5
$\begingroup$

In Jon McLoone's 12.3 overview video, https://www.twitch.tv/videos/1026952947?t=00h09m04s, he mentions that they used the new compiler machinery (FunctionCompile, etc.) on several built-in functions (he mentions e.g., EuclideanDistance being 1.5x faster, CoordinateBounds 3x faster, etc).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.