Hard data on refile effects for NX9

[Fooooore]

I've heard that refile_part has benefits and is important. However, I've run batch-automated loading tests on 1000's of parts randomly sampled from an active Teamcenter database. Many of the parts go back as far as UG v16 vintage and I've seen no significant performance benefit for load times. There was also no issue loading the oldest parts which is consistent with about almost 20 years of UG/NX use. On occasion in the past there was a part that wouldn't open but that was rare.

Does anyone have any hard data suggesting a real impact or is this a myth. If disputed, I'd love to see any evidence anyone can generate for open performance. Just show load time for any pre-NX9 assembly being opened in NX9.0 before and after refile. The NX syslog has a nice printout of the load time (cpu and real time) that can be pulled for quantitative comparison.

I do recognize for parts frequently reused, like standard part families, having the modified flag on a read-only component is a repeated user nuisance when saving.

 

[JohnRBaker]

The biggest thing recently (starting with NX 7.5) has been the idea of using Refile to populate your part models with Lightweight Representations so as to improve the load and display speed of large assemblies, and since NX 8.5, to provide the properly updated data so as to support Smart Lightweight Drawing views, again to improve performance.

John R. Baker, P.E.
Product 'Evangelist'
Product Engineering Software
Siemens PLM Software Inc.
Digital Factory
Cypress, CA
Siemens PLM:
UG/NX Museum:

To an Engineer, the glass is twice as big as it needs to be.

 

[Fooooore]

Thanks John...
Seems like the recommendation of refile has been around a longer than this enhancement, but I surmise lightweight data would have an impact depending on the workflow. Do you have any pre NX9 assembly or parts that you can quantitatively quote differences in the load time data for NX9 due just to refiling the part? I'm really interested in the load time difference without using the lightweight option but whatever you can show would help.
Ruben

 

[markrief]

Historicaly, the recomendation to refile will save you time because it avoids part conversion when you open a part in a new release. When you open an old part, the system always updates the part database to the current version, so the older the part, the longer it takes to open.

In particular for NX9, I have noticed the most improvement in display times, due to the representations the system saves, AFTER the geometry has been displayed and the part is filed in NX9.

I can't get teh spacing correct to show a "table" in the browser, but you should get the ideaa here. In this example I am display two large contoured solid bodies. Also notice that even though the representations are saved, the part file size was reduced from 244 MB to 210 MB.

Version , Seconds to display 1st time, Seconds to display 2nd time
NX 8 , 18 , 15
NX 8.5 , 11 , 8
NX 9 first time opening , 32 , <1
NX 9 after display/save , 7 , <1

Of course your milieage may vary, but I find once you have done the initial display and save, things get much faster :*)

Mark Rief
NX CAM Customer Success
Siemens PLM Software

 

[Fooooore]

Thanks Mark,
We are getting warmer on data...but I'd guess the effect you're seeing from the first time opening and second time opening are related to NX loading libraries and initializing more than anything else. As I'm sure you know, NX is broken into hundreds of dlls which load on demand depending on the content of the partfile (e.g. CAM libraries won't load untill there is CAM content in the .prt, etc,etc). Network installs of NX can be really slow until the libraries are loaded in RAM. The other variable in play are processes eating CPU or sharing network bandwidth.
In the load performance data I've generated, I use a .vb journal to open all the files in random order, 4 times for each file. I rerun the .vb journal a second time in a different random order. The min value shown in the "Loaded and displayed XXX.prt" line of the NX syslog is used for comparison.

Here's the test for whether refile has impact or not on load performance:
-Launch NX9 on a machine with nothing else running. A local install of NX is best.
-Open an old .prt file and close all parts (without saving) 4 times
-Search for "Loaded and displayed" in syslog, record min time as X
-Save old .prt in new version ( a purer test of refile would be to run refile on the assy, but this is close enough)
-Open updated .prt then close all parts in new version 4 times
-Search for "Loaded and displayed" in syslog, record min time as Y

Y minus X is an estimate of the impact of refiling in NX9