Sketch pattern performance

[Helmi74]

Hey guys,

i'm trying to model a vacuum mounting table for my cnc router. I thought going with patterns for the whole might be a good idea but i nearly have no chance to work with patterns as their performance seems to be somehow unusable. The strange thing is that it doesn't even use my CPU. CPU usage stays below 30% while i can't do anything with fusion as soon as the pattern contains above 500 instances (i need around 300). Sometimes CPU peaks to around 100% but still only uses one cpu core. This seems to happen when i move the view. Deactivating single instances with the checkboxes also takes around 10 seconds each but doesn't use up any CPU.

Anyone experiencing the same issues? Any chance to work around this?

To give you a better impression of what it is about - here's a picture of two of the vacuum tables together:

I've already placed the bigger (mounting) holes. The small holes are 6mm wide and 10mm apart from each other. 

[Helmi74]

Hey guys,

i'm trying to model a vacuum mounting table for my cnc router. I thought going with patterns for the whole might be a good idea but i nearly have no chance to work with patterns as their performance seems to be somehow unusable. The strange thing is that it doesn't even use my CPU. CPU usage stays below 30% while i can't do anything with fusion as soon as the pattern contains above 500 instances (i need around 300). Sometimes CPU peaks to around 100% but still only uses one cpu core. This seems to happen when i move the view. Deactivating single instances with the checkboxes also takes around 10 seconds each but doesn't use up any CPU.

Anyone experiencing the same issues? Any chance to work around this?

To give you a better impression of what it is about - here's a picture of two of the vacuum tables together:

I've already placed the bigger (mounting) holes. The small holes are 6mm wide and 10mm apart from each other. 

[HughesTooling]

Rather than pattern in the sketch just make one hole and use pattern from the create menu, patterning the surfaces works better then patterning in sketches if you have a lot of profiles.

Mark

[HughesTooling]

Rather than pattern in the sketch just make one hole and use pattern from the create menu, patterning the surfaces works better then patterning in sketches if you have a lot of profiles.

Mark

[Helmi74]

It's definitely much better to work with after the pattern is done that way.

Unfortunately it's still a real pain to tick single instances on/off while still editing the pattern. I already tried disabling nearly everything visually but it seems to be something else not related to the GPU. Maybe this is something devs should have a look at as it seems this should be possible in a more performant way.

[Helmi74]

It's definitely much better to work with after the pattern is done that way.

Unfortunately it's still a real pain to tick single instances on/off while still editing the pattern. I already tried disabling nearly everything visually but it seems to be something else not related to the GPU. Maybe this is something devs should have a look at as it seems this should be possible in a more performant way.

[HughesTooling]

It is a known issue with sketches, it has come up quite a few time with imported DXF and SVG files. 

Are you working on just one quarter at a time then using mirror to make the other 3?

Mark

[HughesTooling]

It is a known issue with sketches, it has come up quite a few time with imported DXF and SVG files. 

Are you working on just one quarter at a time then using mirror to make the other 3?

Mark

[Helmi74]

That's my workaround and just to be sure i'm not working on the sketches anymore but with surfaces now.

[Helmi74]

That's my workaround and just to be sure i'm not working on the sketches anymore but with surfaces now.

[Helmi74 · Accepted solution]

I finally got it done but it was a relatively hard way. As things aren't perfectly symetric here i had to manually delete single surfaces after each mirroring and add some new before mirroring again.

The performance to work with the final result is quite okay but if you try to edit a mirror or pattern feature with that amount of instances it's a real pain. This is something that should be worked on.

[Helmi74 · Accepted solution]

I finally got it done but it was a relatively hard way. As things aren't perfectly symetric here i had to manually delete single surfaces after each mirroring and add some new before mirroring again.

The performance to work with the final result is quite okay but if you try to edit a mirror or pattern feature with that amount of instances it's a real pain. This is something that should be worked on.

[HughesTooling]

I think what slows it down is when there's the opportunity for user input, so for example even with a sketch closed Fusion is looking for interactions with the profiles. And when you are making a pattern with surfaces it's scanning for interactions, if you just have a model with a lot of surfaces the display is quite fast so I don't think it's just a display problem.

Mark

[HughesTooling]

I think what slows it down is when there's the opportunity for user input, so for example even with a sketch closed Fusion is looking for interactions with the profiles. And when you are making a pattern with surfaces it's scanning for interactions, if you just have a model with a lot of surfaces the display is quite fast so I don't think it's just a display problem.

Mark

[Helmi74]

oh maybe i just wasn't clear enough. I definitely don't think it's a display problem after how it behaves after switching of different display features.

Maybe there's just too much live calculation going on inside pattern and mirror feature. This should be reduced especially on a higher amount of instances to make working with that more handy again.

[Helmi74]

oh maybe i just wasn't clear enough. I definitely don't think it's a display problem after how it behaves after switching of different display features.

Maybe there's just too much live calculation going on inside pattern and mirror feature. This should be reduced especially on a higher amount of instances to make working with that more handy again.

[Helmi74]

after some time i'm sure this is a general path-count performance problem in sketches.

Currently i'm already waiting around 10 minutes for an already selected array of around 2500 circles to be deleted. I mean sure this is a lot but shouldn't this be doable?

EDIT: and then it crashed a few minutes later...

[Helmi74]

after some time i'm sure this is a general path-count performance problem in sketches.

Currently i'm already waiting around 10 minutes for an already selected array of around 2500 circles to be deleted. I mean sure this is a lot but shouldn't this be doable?

EDIT: and then it crashed a few minutes later...

[promm]

@Helmi74,

Patterning, especially when you get into the 2500 range is something that CAD has struggled with for a long time.  Mainly calculations affect the performance, but there are other factors such as visuals and file size when inserting into other models that will cause models to slow down.  We have recently done some work on how our recomputed pattern optimization works, but that does not affect the situation you are seeing.   There are feature enhancements in our backlog that will improve the initial performance, however it is not prioritized very high.  My recommendation for the best performance is start in direct modeling creating your main shape.  In the example below I made a 2'x1' rectangle.  Next create your cut feature, I made a circle.  Then pattern the cut and do not worry about which instances to remove until the next step.  After the patter is complete select the instances that you want to remove and press delete.

http://autode.sk/1I5mo4A

Cheers,

Mike Prom

[promm]

@Helmi74,

Patterning, especially when you get into the 2500 range is something that CAD has struggled with for a long time.  Mainly calculations affect the performance, but there are other factors such as visuals and file size when inserting into other models that will cause models to slow down.  We have recently done some work on how our recomputed pattern optimization works, but that does not affect the situation you are seeing.   There are feature enhancements in our backlog that will improve the initial performance, however it is not prioritized very high.  My recommendation for the best performance is start in direct modeling creating your main shape.  In the example below I made a 2'x1' rectangle.  Next create your cut feature, I made a circle.  Then pattern the cut and do not worry about which instances to remove until the next step.  After the patter is complete select the instances that you want to remove and press delete.

http://autode.sk/1I5mo4A

Cheers,

Mike Prom

[Helmi74]

Thanks Mike.

I already got that solved using the method you are suggesting. Unfortunately in Sketch Mode it's not only the pattern that causes enormous performance issues but that path count itself does that too. I need to project these hole pattern onto sketches every now and then to create a gauge based on that vacuum table. After projection i'm deleting around 50-70% of the projected geometry but that is really a hard job due to the bad performance.

[Helmi74]

Thanks Mike.

I already got that solved using the method you are suggesting. Unfortunately in Sketch Mode it's not only the pattern that causes enormous performance issues but that path count itself does that too. I need to project these hole pattern onto sketches every now and then to create a gauge based on that vacuum table. After projection i'm deleting around 50-70% of the projected geometry but that is really a hard job due to the bad performance.

[daniel_lyall]

I found when I did my vacuum table it was bad I increased from having 4 cells of holes to 6 cell all mirrored from one cell of 500 holes each 500 holes took at lest 3 min to compute on the first then each mirror the time it took to compute was about double, I went to 6 cell from 4 as it just crashed fusion with 4 cells of 750 holes.

I know it is not as big as your one Frank but still same stuff diffrent size.

[daniel_lyall]

I found when I did my vacuum table it was bad I increased from having 4 cells of holes to 6 cell all mirrored from one cell of 500 holes each 500 holes took at lest 3 min to compute on the first then each mirror the time it took to compute was about double, I went to 6 cell from 4 as it just crashed fusion with 4 cells of 750 holes.

I know it is not as big as your one Frank but still same stuff diffrent size.