Memory Leaks in 2015?

[LT.Rusty]

So this is something I've noticed on my office computer but haven't seen on my desktop or laptop at home.

Inventor doesn't seem to release memory when it's finished with thinking about something.  I added 16 GB today - total of 20 now, up from 8 - and after closing all my files out, Inventor was still sitting there at 0% CPU usage, but sucking up 18.1 GB of RAM.  When I had only 8 GB, I thought I was just running out of RAM, but now that I'm up to 20 and it's still behaving the same way ... The assembly I had open previously was only 30 instances of 18 parts, none of them complicated in any way.  (Most complex part was a 2 x 4 x 1/4 channel from content center with a split and an extrude done to it.)  Additionally, when I go to close out Inventor, even though the program disappears from the screen, it's still showing as a process with ~18 GB of RAM used in the task manager - I have to end the process tree (not just the process) to get it to actually go away.

I've turned off Quick File Open, and I'm sort of at a loss right now as to other things to try.  

System is an HP Z220 SFF workstation.  Xeon E3-1240v2 running at 3.4 gHz.  20 GB of RAM in dual channel (2x2GB, 2x8GB).  ATI FirePro v3900.  Win7 Enterprise, up to date.  Inventor is fully updated.

Anyone have any ideas?

[ChrisMitchell01]

I'm not aware of any general memory leaks in the latest SP/update for 2015. I suspect that if you waited long enough after closing Inventor the RAM would eventually be released.

Does this happen on all datasets or just that particular assembly ? If you don't see it on your other computers, (presumably with the exact same assembly & Project configuration), then what else is different ? Graphics driver perhaps ? What Inventor graphics setting are you running?

If it is dataset specifc, (Ie you see the same problem on other computer for that assemblY). is that data something you can share with Product Support for further investigation ? A full PacknGo including library files & material/appearance libraries, Design Data etc would be needed.

Thanks

Chris

[LT.Rusty]

---

@ChrisMitchell01 wrote:

I'm not aware of any general memory leaks in the latest SP/update for 2015. I suspect that if you waited long enough after closing Inventor the RAM would eventually be released.

Does this happen on all datasets or just that particular assembly ? If you don't see it on your other computers, (presumably with the exact same assembly & Project configuration), then what else is different ? Graphics driver perhaps ? What Inventor graphics setting are you running?

If it is dataset specifc, (Ie you see the same problem on other computer for that assemblY). is that data something you can share with Product Support for further investigation ? A full PacknGo including library files & material/appearance libraries, Design Data etc would be needed.

Thanks

Chris

---

I've waited 5-10 minutes after closing, and the memory isn't being released.

The issue seems to be only with this assembly, although I haven't done a lot of FEA with this particular workstation / Inventor installation, so it may crop up in other assemblies as well.  (I'm working for a new company now, this is the first major project where I've been required to do a lot of FEA stuff.)

I tried it on my laptop (specs as per below in my sig) and got almost exactly the same results, although it did (eventually) release the memory when I closed Inventor.  Took a good little while though - almost 2 minutes.

I am running quality for the graphics setting, but the problem hasn't ever manifested itself as slowness or delay in the graphics.

I'll PM you a link to the data set and the steps to replicate the issue.

[ChrisMitchell01]

OK Thanks. It wasn't clear that this was FEA related initially. I'll see what I can find out from your data.

-Chris

[LT.Rusty]

I'm honestly not sure that it IS FEA related ... it's just that FEA stuff tends to be the fastest way to ramp up memory usage fast.  When you open the files you don't need to actually run any FEA, just play with the visibility and display the results as described.  It appears to be doing the same thing even if I don't use the FEA environment, just at much lower levels of memory usage. 

[ChrisMitchell01]

Yes, there's definitely a problem in this dataset related to switching part visibility when you have FEA results displayed. I just ate up 32GB RAM easily enough. I'll ask Dev to take a look.

In the meantime it seems that if you set the desired visibility before entering the FEA environment it doesn't use anywhere near as much RAM.

Thanks

Chris

[LT.Rusty]

Thanks!

[dgorsman]

"Memory leaks" in the classic sense are exceedingly rare these days, as the tools used to develop the software handle allocation/deallocation automagically.  One thing that is *very* prevalent in modern software is to reduce times by loading as much as possible into memory, thanks to dirt cheap RAM costs.  There could conceivably be a problem with forward-only caching that only loads data and never unloads anything or checks for free memory, but thats not strictly a memory leak.  Checking with a developer to see if there is a memory leak by name may lead them down the wrong diagnostic path.