Prime10 is 30 to 50% slower than P9 - Unacceptable

[DS_10314450]

Have dual loads of P9 and P10.  My application is CPU intensive.  Recent testing showed a big difference so I did a simple sum test.  The results are not good in my opinion.  Can this be fixed with a patch?  Bloatware at its finest - added functionality with engine overhead. 

[rgunwaldsen]

My old Microsoft Surface Pro 3 isn't nearly as fast as your computer, but I can confirm your observation that Prime 10 is a dog.

Roger

[Cornel]

Did you tried to enable multithreading from calculation option?

[Werner_E]

Multithreading must be turned off for the timing function to work correctly.

[rgunwaldsen]

Thanks for the information -- but Multithreading is off in both cases.

[Werner_E]

---

@rgunwaldsen wrote:

Thanks for the information -- but Multithreading is off in both cases.

---

I know, I checked it when I downloaded your files 🙂

I also can confirm that P10 time usually is longer than P9, which indeed is disappointing.

I also noticed that the difference in calc time is much less if the function used is more 'complicated' and so more time from the overall calc time is spent for the external function calls.
So I guess that its especially the program loop which is (for reasons unknown to me) implemented in a less desirable way in the new version.

But I second Martin's suggestion that @DS_10314450  should report the problem to PTC support directly and then hope for improvement in a future version.

The timing function generally is very unreliable - in some (very) rare cases when I let the sheet recalculate multiple times the P10 sheet showed about the same time as the P9.

I can't find the thread but I remember that some time ago a PTC representative claimed that improving speed would be highest priority for  the R&D team. Can't remember which version that speed improvement was scheduled for.

EDIT: Guess I found the thread: Re: Mathcad Prime 9.0 Slow Performance - PTC Community
On the other hand, I also remember PTC's promise that P3 would offer the full functionality and ease of use of real Mathcad (MC15) and when I now compare P10 with MC15 ... 😞

[DJNewman]

What I said then hasn't changed and performance updates remains a major focus in Prime 11.

Please continue to submit cases to PTC (preferably through support cases because it's internally documented better that way) of examples of unexpectedly slow performance in Mathcad Prime 10. I've already provided some such examples to the dev team of things in this thread and other recent things (like an example ttokoro pointed out on his YouTube channel).

[DS_10314450]

Can you comment on where the bottle neck may have formed?

Could it be corrected?

[DJNewman]

I can't comment about that since I don't know the answer.

(If I did know the answer, I probably still couldn't comment on it unless it was to enforce a best practice for Mathcad users, but in this case it's probably a product-side problem that only R&D can deal with.)

[Cornel]

@DJNewman 

I don't know what to say, but I also see in my company that they have intention to introduce wolfram mathematica, this in addition to matlab. I'm afraid that I will be the only one who will use Mathcad prime or that there will be a few users of Mathcad prime and the company will decide to stop paying the Mathcad prime license as will make no sense anymore to pay for few users and tell us to migrate to other math software that the company will provide to us.

What do you think? Will it still be worth it to remain users of Mathcad prime for the future?

On the other hand, if PTC goes worse and worse with the development of Mathcad prime at some point in time it will be so bad for users that they will no longer tolerate the poor development of the software because they will only be able to use it in a few situations that will make no sense to use seriously Mathcad prime in industry and for real work because they will not be able to solve real work more complex problems.

Pobably that also the other competitors of Mathcad have also given up comparing their software with Mathcad, considering that Mathcad development itself is behind them more and more as days go, Mathcad being poorly developed for Mathcad's competitors to take seriously into account Mathcad as a serious competitor for them, and probably it makes no sense for them to further compare their software with Mathcad (probably that they are ashamed to still compare) when as well as the people who own and develop Mathcad/PTC management put their shoulders to the ruin of Mathcad and thus benefiting and helping other math software manufacturers.

Probably that Mathcad will disappear from the market in this way, or be used by a few college or a few hobbyists or retired people, all people uses Mathcad more for fun, if PTC will continue to go worse and worse with the development of Mathcad Prime and if PTC will not make more serious improvements of Mathcad Prime and to really remain a real and serious competitor for other math software manufacutrers.

Maybe you can take these informations that people are saying on this forum and present to the PTC management.

[DJNewman]

Well, there's no scenario where you're going to get a PTC employee to answer anything besides "Of course it'll be worth it to keep using PTC Mathcad in the future!" Though even if I wasn't paid to say that, I'd say that.

I think it's demonstrably clear that Mathcad development has been getting better in recent years (Prime 7+; especially Prime 9 and 10), not worse, compared to certain previous versions of Prime (like... 5), with regards to how much stuff gets fixed/improved/added in each release in recent years. That's not a coincidence. PTC is more responsive to Mathcad customers' needs than we used to be. We're better resourced internally than before. The doomsday scenario you're portraying is not based on reality. I can say that with 100% certainty. With regards to this thread about the performance issues, it's actively being worked on for Prime 11 per the roadmap.

The paragraph about our competitors no longer comparing themselves to Mathcad is really funny, because a certain Canadian competitor is currently obsessed (and has been for at least a few years now) with comparing themselves to Mathcad and having their software be a Mathcad imitation. Their entire reason for existence is to compare themselves to Mathcad, and they've even quite recently come to PTC Community to try to sell to people here. I'm also flattered they're even copying (ah, maybe I should say "heavily inspired by") PTC Mathcad marketing. 🙂

[Cornel]

I do not know the name of this "certain Canadian competitor" and if it is a leading math software manufacturer, but for sure you should take a look at competitors like Maple or Wolfram Mathematica or Matlab which indeed are leaders in math software and see what improvements they made for they software and what improvements PTC made for Mathcad. But I will not go into such discussion with PTC because for sure PTC sees Mathcad better than users sees. I said above things only because I am an user of Mathcad Prime and I like Mathcad and I would still like to use Mathcad (as for example I do not like to write code as it is the case with other math software), but for sure there are problems with Mathcad as many users founded and have complaints about slow performance of Mathcad, lack of some functionalities, and so on.

And also maybe R&D team developers of PTC Mathcad Prime should take more look in this community and better to come and be involved with better support and answers/workaround/solution here with us for Mathcad users problems that they encounter. 

[DS_10314450]

Hi Cornel,

Yes, I used the same settings on both.  Multithreading enabled.

[MartinHanak]

Hi,

please open Case at PTC Support and report the problem directly to PTC.

[LucMeekes]

OK, Let's see how Prime progressed over time. Since I'm limited to Express, I had to use a different speed test utility.

Also know that the time() function was a premium function in Prime 1 and 2...

Prime 3.1:

Prime 4:

Prime 5:

Prime 6:

Prime 7:

Prime 8:

Prime 9:

Prime 10:

Notice that there is quite some variation in the speed test execution times.

I ran all tests on the same machine, with just the Prime-version-under-test running during the test. They all loaded the same file (attached, Prime 3.1) and after loading I pressed the 'Calculate' button to have the sheet fully recalculated. Then I took the result.

I'd say Prime 10 is the slowest, Prime 8 comes next.

Success!
Luc

[Werner_E]

I couldn't resist and I used your sheet (with 30 runs) to test the five versions of Prime installed on my machine.
Nothing running besides the appropriate prime version (and Win10 of course), multithreading turned off).

I exported the results via WRITEPRN and made the comparison plot in Mathcad:

'Winner' is again P10 😞

[JeffH1]

My results are slightly different.   I modified your test a bit to evaluate and compare both the trig function as well as the trivial function that returns 1.  I also calculated both scenarios manually with in a program without passing them in as parameter functions.

I see some significantly slower times if the summation functions are passed in vs. not.  However, I see a speedup in Prime 10 when not passed.  We're measuring 100's of milliseconds here so results fluctuate with each recalculation with whatever else is going on the machine.  After many recalcs, the above screenshot is typical for my machine with nothing but Windows running.  While I do see some increase in calc time when passing the functions in, it is not nearly the 300% increase shown in the original post above.  It's possible this is machine/chip dependent, so I posted my machine specs as well.

[DS_10314450]

Update:

• thank you all for confirming
• Mr. Newman, thank you for your focus
• Final comments
  • I am building a computational FEA analysis tool.  It was originally started in Prime8 just for the prototyping aspect.  Ideally, it gets built out as standalone C++ exe with full opengl support.  At the moment, the entire function suite in in Prime.  I am looking to convert a few heavy use functions to DLLs that can then be mounted in a full standalone executable.  However, this is a bit off.  I would like to advance to P10 but this will prevent updates from running in P9.  This is the only way to offset the new bloatware burden placed by P10. I assume, without substantial rework, P11 will only get worse.  This is indicating a haste in program decisioning.
  • Is there any possibility of taking the dynamic build that is created in Prime, save, and run as a DLL?

Regards