Survey: Inventor Performance

[Neil_Cross]

Q.  Would you fully support the Autodesk Inventor product team in temporarily stopping work on any new features for Inventor, and instead, dedicate a couple of years and all their resource to working ONLY on making Inventor perform much faster than it does now?

Specifically referring to converting more of the application to being multi-threaded, so it can use more than 1 of the many increasing CPU cores that modern architectures have now.

Please note this is a personal survey and in no way has this been actually discussed with Autodesk, there is nothing to read into here!

[IgorMir]

I believe  - seasoned users of Inventor lost all hopes of that ever happen years ago...

[johnsonshiue]

Hi Guys,

I was not planning to reply to the thread since it was targeting Inventor user, not Inventor team. However, as a team member, I personally feel obligated to reply in order to clarify a few things based on what I know.

The question raised here is totally legitimate. Does Inventor have performance issues? Certainly, yes! There is always room for improvement in every corner and every workflow of the product. But, let's put things in perspective. Performance improvement is an endless pursue. Back in Inventor R5, an assembly with a thousand components can be considered a large assembly. Fast forward to today. If I told our users that such assembly was large now, I would be laughed at. Our customers use Inventor 2019 to design half million-component assembly and create associative drawing views. This kind of  capacity and performance was unthought of 15 years ago. Our internal benchmark shows Inventor 2019 outperforms our competitors in large assembly views and drawings.

One can certainly argue, why it took 15 years to reach this level. I get that argument. But, one cannot deny the dramatic improvement. We can always move faster and we should try to do so at all cost.

Multi-threading and multi-core have been brought up repeatedly as a means to enhance performance. The truth is Inventor is not fully multi-threaded but it is also not fully single-threaded. It is partially multi-threaded (see below page). Again, the ability was not even available 5~7 years ago.

https://knowledge.autodesk.com/support/inventor-products/troubleshooting/caas/sfdcarticles/sfdcartic...

There are workflows sequential in nature like feature compute. Making them multi-threaded is not only technically problematic but also dangerous. Not every workflow and code path can be multi-threaded. Not every one should be done so. I believe there is a lot of opportunity to make Inventor faster at workflow level, i.e. command execution, data handling, and data presenting. The team has been focusing on overall performance improvements for years. I hope you are seeing the progress. There is a lot of work to do. Tuning is always harder than building something new.

Many thanks!

[Neil_Cross]

Hi Johnson - this is research for the class I'm doing at AU next month.  I'll PM you as I don't want to publish some of the content of my AU class prior to it happening.

[Neil_Cross]

Hi @johnsonshiue I responded with a PM to explain some findings.  If you're attending AU, feel free to drop in on my class, I think its full but I'm sure you guys are more than welcome to squeeze in.

Regarding that AKN page and Inventor being partially multi-threaded, that's a bit of a stretch isn't it? 

There's 6 headings there with 13 points.  Some of them arguably sound like duplicates of one another i.e. rendering/raytracing, drawing views don't multi-thread very well at all, I strongly dispute the claim that view related commands are truly multi-threaded, the freeform commands where an acquisition from Brigham University but OK I guess that one is valid, the task scheduler doesn't convert the core single threaded processes into being multi-threaded, the rest... OK fair enough we could argue how much of a benefit it was to multi-thread things like selection and mass prop calcs, but this is a tinkle in the ocean of the wider Inventor tool set.  It's a stretch to label Inventor as being partially multi-threaded based on that list!

But regardless - I didn't make the post to undermine the work your guys have done, I know they work hard and are under severe pressure with unreasonable time constraints and expectations, please don't take this as a disregard of the efforts that have been made over the years.  I just want to understand how people feel about the performance right now, and would they be happy to sacrifice new stuff coming in in favour of a sustained "just do it" effort to tackle the areas that have been neglected because they are, as you say, technically challenging. 

[johnsonshiue]

Hi Neil,

Many thanks for clarifying the intent! I heard about the class. I will be at AU this year. I will try to participate in the class if the schedule allows. The list is actually a simplified form. If we do it on per command basis, it can be quite long. In terms of geometry operations alone, almost all ASM related operations are multi-threaded. This includes Extrude, Revolve, Sweep, Fillet, Chamfer, Shell, and so on. The most important of all is Boolean operation is also multi-threaded. The import process is multi-threaded too.

Like I mentioned before, there are certain operations in Inventor simply cannot be made multi-threaded, unless the entire modeling paradigm is changed. Multi-threading for certain operations may not help boost performance, since there are overheads involved in starting and ending threads.

With all that said, I don't believe Inventor has achieved its full performance potential, regardless of using multi-threading. The team is aware of that and it is actively working on it with or without this survey.

Many thanks!

[pball]

Yes.

I'd also hope outstanding bugs could also be dealt with.

[Rory_M]

No.

We don't make 1000+ part assemblies and therefore don't see major issues with performance.

I wouldn't give up on the chance of getting new features that would give me a better speed boost when compared to pure CPU/GPU performance gains.

Also, one important thing to note with any performance metrics, making software twice as fast doesn't necessarily make the end user twice as fast.

I can spend 8 hours a day working on Inventor but have I been pushing buttons for 8 hours? No. I'm figuring out a design, carrying out hand calcs, trying different shapes & forms and generally spending more time thinking about what I'm doing than actually doing it.

Making the software twice as fast might give me a 10% gain in real terms (unless I'm rendering or carrying out FEA calcs in which case speed increases are pure gain).

Having better software with better commands & features, as opposed to just faster software is for me the ideal scenario. I can see why people modelling up huge assemblies who are being hit by massive compute times for drawings might disagree though.

[Frederick_Law]

---

@Neil_Cross wrote:

Q.  Would you fully support the Autodesk Inventor product team in temporarily stopping work on any new features for Inventor, and instead, dedicate a couple of years and all their resource to working ONLY on making Inventor perform much faster than it does now?

---

Dedicate a team to work on performance and new technology, ie multi-core, mulit-thread, GPU computing, touch screen, mobile, VR etc.

Dedicate a few years will make Inventor lack behind the industry.

I see Inventor as ahead of SW right now.  So keep on innovating and stay in the lead.  Inventor has a performance and technical lead using DirectX compare to OpenGL.  Keep that lead on future graphic engine and hardware.

[Neil_Cross]

Thanks to everyone giving their feedback here.  I'm intentionally not entering into a discussion surrounding the points raised as I'm just looking for yes or no's... I have my own opinions and will save them for the class at AU.  But thanks in advance to anyone who puts in.

---

[philip1009]

I agree, it does come down to use cases, for design and modeling work, better and easier tools are in high demand.  Then there's the other workflows that require users to go through a typical production workflow as fast as possible, for those workflows I've found that implementing automation where possible is the best route.  I do agree that it's annoying that Inventor can't quite keep up with me during manual work, but I don't think they should stop trying to improve and introduce new tools at the same time, there's been more than enough improvements that have made my work easier to say Autodesk doesn't know what they're doing.  I agree they certainly need a lot more QA before releasing, there's been too many glitches and breaks in the past couple of releases that raise a lot of questions to Autodesk about the quality of the software.

[Frederick_Law]

---

@Neil_Cross wrote:

Thanks to everyone giving their feedback here.  I'm intentionally not entering into a discussion surrounding the points raised as I'm just looking for yes or no's... I have my own opinions and will save them for the class at AU.  But thanks in advance to anyone who puts in.

---

---

Then my answer is, No.

[Anonymous]

Sure there must be work spent on speed of operating. But speeding up the engine may not be the best option. If it would be a car then fitting a bigger engine will speed it up, but you also might consider making the body lightweight. Take the topic of drawings. Inventor chooses to represent the body directly. This causes a lot of calculation whenever something changes. Some other CAD chooses to build a representation of the body in standalone geometry. It will warn you when the model has changed but it will not recalculate until you say so. These things are core choices in the software and can't be changed overnight. There are more of these core choices in Inventor that IMHO are not optimal. But chances are big that you might need to rebuild the program from the ground up.

[mcgyvr]

No..

I don't want them to stop developing new features as I'm hoping that one of these coming years something comes out that is actually useful to me..  

[dgorsman]

To the original question: no.  Imagine things like no support for 4k displays, Win 10, new display hardware, 3D PDF.  In a couple of years this would leave Inventor well behind the curve, trying to play catch up and inevitably failing (as the target being aimed for is near-obsolete by the time it gets released).

Performance gains are not necessarily objective either.  Consider that a user may be doing something completely wrong, insisting on using a less-than-efficient workflow, or using outdated hardware or drivers.  All of those will affect performance, and a global technological "improvement" will have a much smaller impact.

[Neil_Cross]

---

@dgorsman wrote:

Performance gains are not necessarily objective either.  Consider that a user may be doing something completely wrong, insisting on using a less-than-efficient workflow, or using outdated hardware or drivers.  All of those will affect performance, and a global technological "improvement" will have a much smaller impact.

---

Appreciated, but a user doing something wrong is way beyond and outside the scope of my AU class.  It's a class on choosing the right workstation for Inventor.  I'm concentrating explicitly on the way Inventor interfaces and utilises the workstations that clients are purchasing.  I can't possibly factor in user training and individual workflows.

Which is why I'm trying not to engage in discussions beyond that, those discussions are 100% valid and very important, just not something I can factor into my class.

[SBix26]

No, for the reasons that @mcgyvr and @Rory_M posted.

---

Sam B
Inventor Pro 2019.1.2 | Windows 7 SP1
LinkedIn

[johnsonshiue]

Hi Alex,

Just to clarify drawing view compute, I am not sure what you meant by the drawing views being represented by 3D body geometry. The view geometry is associated with 3D body geometry. When there is change in body geometry, the view will be updated. On 2018 and 2019, the update process has been optimized so that only the affected area of the view is updated. Not all view geometry is recomputed.

Many thanks!

[Anonymous]

Johnson,

Re-tested this and - sorry - I was on the wrong foot there. Seems you can open a drawing of a part that is gone. Meaning that the 2D geometry must be present in the drawing :-). Which is not the case in all CAD.

Some remarks on performance in drawings however. When both model and drawing are open recomputing should IMHO only occur on demand. Some indication in the view that the model has changed can be sufficient as information. Compared to the computer I had in my freelance career I now have a Formula-1. Still, and compared to then, I find drawings on multiple pages and with multiple views slow. I'm not deep enough into the inner workings of IV but I think that it's mainly about 2 things, being sparing use of updating and efficiency of 'hidden line removal' (processing of models into 2D reps).

[johnsonshiue]

Hi Alex,

No worries! Inventor is such a massive program. I am still learning it everyday. Regarding on-demand update and selected update, it was not available on 2014. If I recall correctly, it started on 2018.1. If possible, you might want to upgrade to 2018.1 or later to see the behavior. Drawing view compute should be a lot better than before. However, there is always room for improvement.

Many thanks!