How fast is your Inventor PC really?

[Raider_71]

Hi guys,

We have had to do some testing on a bunch of Inventor PC's recently to determine which of the PC's needs to be replaced. Obviously we needed to find out which of the PC's are the worst eprformers as there was only budget to replace 50% of the design PC's. So we thought the Darwin theory will come in handy right... 🙂

Anyways I started searching on the net for toppics on how to benchmark an Inventor PC. Then I thought whats the point of using gaming benchmark tools because Inventor is not a game and there are more aspects than just graphics performance when it comes to percieved performance on an Inventor PC right.

So we decided to create our own Inventor benchmark tool which tests various aspects of an Inventor PC to give us an overview of our PC's performance. This then helped us make a decision as to whcih pc's to replace.

We have made the tool available free of charge to anyone interested in checking how their PC stacks up to their peers or friends. 🙂

Please download it here and post your results here as well if you want. Would be interesting to see what beast workstations are out there.

I would like to say thanks to Kirk #karthur1, for helping in testing the app.

Please feel free to send any suggestions our way. There is an email link in the app.

Download and Install

The application will work with Inventor 2014 to 2016 only.

IMPORTANT: After installation there will be an Inventor Bench icon on your desktop that looks like this: 

My resluts:

HP Elitebook 8560w with an SSD upgrade.

[smokes2998]

Just wait for 2018 there are doing a lot of work on Large assembly handling to make inventor perform better.

I think it involves a method of automatically reducing the detail of the assembly when it is been worked on.

[mmaes]

I tested a few different graphics cards today with the 7700K CPU overclocked to 5.1 Ghz.  I also did a couple just for fun.  One with no card installed and I also tested the 6700K with the GTX960 and then again on the 7700K to see the difference when all the hardware was the same except the CPU.

I think these results just further highlight the need for assemblies in the Bench tool.

The cards in the picture are GTX1060 (top), GTX1080 (middle), and TitanX Pascal (bottom).  I was lazy when I took the picture and didn't pull the GTX960 out for the photo.

[Neil_Cross]

---

@mmaes wrote:

I think these results just further highlight the need for assemblies in the Bench tool.

---

That is a sexual godly collection of cards, puts my collection to shame! I think though that these results just further prove that unless the GPU is choked on VRAM, it doesn't matter what GPU someone has.

I love a good GPU as much as the next geek but there has to come a point whereby it becomes undeniably obvious that they really do nothing for this application.  I've done the real world tests on a 1000 part assembly, these synthetics fall in line with my testing, you could put this to bed by running OCAT and benching the 100,000 part assembly through the 1080 and the Titan.

If @Raider_71 was able to include assemblies into the next iteration of the bench tool, unless there's a breakthrough in compression technology I would doubt he'd be able to include anything much bigger than 100-500 parts max in a download else his application would end up GB's in size.  That car dataset is 1100 parts and the ZIP file for it is 1.5GB.

[mmaes]

@Neil_Cross do you have a link to download the OCAT software?  I did a quick search and found it but the site looked a little sketchy.

As for the collection of cards...I am lucky enough to be the person who purchases the components for our workstations.  I chose the TitanX for our VR machine.  The machine that is installed in full time doesn't run Inventor at all.  We are a paper packaging machine company and we model and simulate product lines using actual data from the PLC.  We let the customers come in and "walk" around their plant, pick up and look at product, and see their facility with our equipment in it before the equipment is ever built via virtual reality.

It's pretty cool stuff actually...picture your VR video you did where you are inside the inventor model but instead it is an entire paper mill with product running down conveyor systems and going through the packaging process.

Sorry, I got a little off topic.  I will admit I agree about your comment on VRAM.

[Neil_Cross]

https://github.com/GPUOpen-Tools/OCAT/releases

I've done a video on my channel for it 

It isn't default integrated with CAD so you have to manually launch it into our applications, but it works well.

One of the clients I do Vault consulting for are a food packaging and seasoning machinery manufacturer, they could really do with a similar VR experience to what you have and it's been on my list of things to propose to them.  Problem is I just don't have the time to see a project like that through.  Did you use a third party to build the VR environment or was it done in house? I assume it's done in a game engine like Unreal or Unity?

[mmaes]

I will message you with the software details as this is a new thing for us and our competitors don't use it yet.

[Raider_71]

Hi @Neil_Cross

Yes I think a nice size assembly would be cool but I agree with you, the size of the download will be a problem. I am toying with the idea of programmatically building the parts and inserting them into an assembly for testing. I could build an array of cubes but all with different sizes and then alternate chamfers and fillets on their edges. This firstly will increase the polygon count but also ensure that there are non-planar faces which I am sure renders differently.

Do you know of a way to check polygon/face count in Inventor?

[Mario-Villada]

Hi @Raider_71, I have some experience with programming in inventor. of course not as good as yours. But can I suggest to add the option to your program for the user to be able to select its own assembly for testing purposes? you might need to set some conditions to the selected assembly like it should be saved in the local PC, also check it does not have any conflict with vault.

Just an idea.

[Raider_71]

Hi @Mario-Villada

Yes, I will look at something like that. Thanks for the idea.

[Neil_Cross]

@Mario-Villada That would be a good idea in addition to testing a default assembly, if everybody could test a different assembly the end result wouldnt mean anything in a comparison.

@Raider_71 there is a triangle count in the default Inventor stats enabler but it seems to always just say zero, I've got a contact on the Inventor graphics team so I'll give them a shout and see if they know how to enable that stat.

[Raider_71]

Yes for sure, if the test sequence is altered in any way like when you select your own Assy to test, then the overall IPI score indicator will be disabled or something and only the graphics results in the results pane will be visible and can be used for comparison.

[brotherkennyh]

You could allow users to test on their own assembly and compare the results graphically. Assembly size vs FPS.

If you collate all the test results and the main components in a database, including assembly size this will allow you to build the graph above. You could allow filters for the graph by CPU, GPU, HDD, release etc, allowing users to screen for results more relevant to them.

It would be hard to compare at the start, but as more results come in it would be a useful tool. You would get different results on systems with assemblies of the same size due to the number of polys, but a best fit curve would emerge.

I know I would make a point to test several assemblies on different systems, I imagine others would also. 

The assembly I mentioned earlier is around 3000 parts, but the total package size is only around 50MB. This or something like it might be a good standard to test against.

[Neil_Cross]

Is that 3000 occurrences rather than 3000 physical files?  

[brotherkennyh]

Sorry yes, that is around 3000 occurrences. I don have the assembly to hand at the moment, but I think it was around 30parts.

[swalton]

@Raider_71:

Thanks for posting this tool.

Is there some sort of useful complexity metric for assemblies in Inventor?  Maybe some function of the component count, the triangle count, and the constraint count?  If there is one, it would be a good way to rank benchmark scores across different user assemblies.

I see the problem with including a default "large assembly" with the benchmark tool.  My large assemblies run from 2000 to 30000 occurrences and Pack-n-Gos run form 10s to 100s of MB without drawings, so it would be hard to include a default assembly that size. 

If you do build a large assembly from a script, don't forget the Content Center components as a source of models and complexity.  The test assembly build speed could be part of the benchmark.

How about adding a large assembly test that has a separate download of the test assembly?  Users can run the small assembly benchmark out of the box and only download the large benchmark if interested.

[Raider_71]

Hi yes the idea is to create an assembly with as many unique parts as possible and at the same time try and generate loads of faces to render. I recall reading something about the way Inventor's graphics engine handles similar parts, if a part is inserted into an assy at the same orientation, it uses some direct x feature where only one part has to be rendered. The other parts graphics is handled by this direct x function and the rendering is just copied. If that makes sense...

Yes so that's why I think an assy with as many unique parts as possible would be the way to go. This way we could end up with an assy with a much smaller part count but as taxing on the GPU as a normal assy with way more parts.

The optional assy download is also a great idea. Will keep it in mind.

Cheers

[Neil_Cross]

---

@Raider_71 wrote:

Hi yes the idea is to create an assembly with as many unique parts as possible and at the same time try and generate loads of faces to render. I recall reading something about the way Inventor's graphics engine handles similar parts, if a part is inserted into an assy at the same orientation, it uses some direct x feature where only one part has to be rendered. The other parts graphics is handled by this direct x function and the rendering is just copied. If that makes sense...

---

Yea I read that too, and I've noticed it happening in Inventor which is why I only use 1 large assembly for testing rather than copy & paste it 100 times over, the PC doesn't take on any more load when this is done so it's seemingly pointless and unrealistic to do it.

[brotherkennyh]

I just tried this.

I created a model. A Cube with a bunch of polygons, fillets, shells etc.

I checked my CPU/GPU load while 1 part was in an assembly while panning around.

I patterned the part 100 times, then 3375, then 8000 times. In each case there was no increase in memory footprint or CPU/GPU load etc.

I did create some screenshots showing the difference, but you cant even tell the difference in the performance results so there is little point showing them.

[Raider_71]

Yup as I expected. Look its very clever tech and I am glad it even exists because you will see better performance in assemblies with many instances of the same part.

I am busy with a script which will generate an assembly with many unique parts with complex and unique faces with bumpmap textures etc. I am just worried about the textures which may cause the graphics card's Vram to limit quickly. Worth a try.

[jpms24]

Is Inventor CPU or GPU load?

Will having a big amount of CUDA power will benefit Inventor?

When working on large assemblies is it better to derive models in or create single parts and sub assemblies?

I know is a lot of question but maybe this help to find a sweet point between performance and quality. What kind of limitation has people around this forum encounter?

Thanks all!