[Idea] The Benchmark of FreeCAD to make it more popular

[NoUsername_]

Hello, I'm new to the forum, so forgive me in advance if I miss something that is already implemented.

Today websites that publish reviews of CPU and GPU tend to not only benchmark pc games, but also a great amount of other programs for ML, image processing, 3d graphics. It has become became "a must" to measure the performance of Blender if you test CPU or GPU. Websites now test GIMP, code compilation, web-performance etc. (https://www.anandtech.com/show/16214/am ... -tested/12).

What do you think about the idea to create benchmark to calculate multiple operations on objects in FreeCAD? It could help to promote the software almost effortlessly.

[drmacro]

Is there a goal to promote the software?

To what end?

Why would I care if it takes 10 seconds to do on SoftwareXXX and 10.1 seconds on FreeCAD?
(If it took 10 seconds on one and 10 minutes on another, maybe.)

Blender is a render and animation tool. I.e. it has to render thousands upon thousands of frames for an animation sequence.
Bench marking it, makes sense.

IMO

[Kunda1]

To the end that it could be interesting and provide more data points on what could be improved (coparing boot time or execution time of different operations). This would also be helpful with different workbenches that are tackling different industries
BIM (Archtecture)
Trails (Earthworks and Civil Transportation)
Render WB
FEM (Finite Element Analysis
etc....

[watsug]

Why would I care if it takes 10 seconds to do on SoftwareXXX and 10.1 seconds on FreeCAD?

The opposite. Using FreeCAD as a load to benchmark pc hardware.

I think it sound like a fun idea. The question is what can be used as a load? Maybe a FEM project? That is calculation heavy and gives a cool looking visualisation when it's complete.

[Kunda1]

The opposite. Using FreeCAD as a load to benchmark pc hardware.

Inadvertently, we would be benchmarking the OpenCascade kernel

Edit: I mean...using the OCCT kernel to benchmark hardware

[Zolko]

What do you think about the idea to create benchmark to calculate multiple operations on objects in FreeCAD? It could help to promote the software almost effortlessly.

I think that's a very good idea. Not only to promote FreeCAD (what would be gaining by that ?) but to actually benchmark FreeCAD, and have it compared to other CAD systems.

The opposite. Using FreeCAD as a load to benchmark pc hardware.

Inadvertently, we would be benchmarking the OpenCascade kernel

Yes, you're right, but it's actually usefull: with large assemblies FreeCAD (and/or OCC) tend to become very sluggish, which means that in the near future, right after the topo-naming fix in v0.20, FreeCAD's performance on large systems might become the next hurdle.

Therefore, as benchmark, I'd suggest to measure the loading time of a large STEP assembly (~100Mb), and then the FPS when rotating that assembly. I would be really interested in FreeCAD performance compared to other CAD systems.

[Jee-Bee]

If the goal is make FC more popular than i think a course om for example coursera would be better... Just like inkscape: (https://www.coursera.org/search?query=inkscape)

[jbraun]

And on the low end of the spectrum:
This morning I'm making tool paths with FreeCad using the guts of an old Netbook complete with on-board graphics. Simple stuff but there's no way this old hardware runs Blender... or Windows for that matter.
The ability to run on old junk may not be FreeCad's goal but it's appreciated anyway.

[NoUsername_]

Is there a goal to promote the software?

//
1. I would say to "familiarize" visitors of those website with FreeCad.
2. The amount of people who buy videocards and cpus not just for gaming, but for hobby-professional activities(machine learning, 3d modelling etc.) has raised dramatically in the last 5 years. When people see the benchmark of performance of PyTorch, code compilation in VS 2019, or rendering in blender, they regard those tools as default apps to get things done in the corresponding fields, especially when they see the tests of those tools year after year.
3. Once some event triggered a user to make something in a given field (say he saw a character from a game and wants to recreate or modify it) - he knows what kind of tools are available to solve the task.

Why would I care if it takes 10 seconds to do on SoftwareXXX and 10.1 seconds on FreeCAD?
(If it took 10 seconds on one and 10 minutes on another, maybe.)

The goal isn't to measure the performance per se, but to use performance tests as a way to promote FreeCad for... free.

Blender is a render and animation tool. I.e. it has to render thousands upon thousands of frames for an animation sequence.
Bench marking it, makes sense.

As was mentioned: FEM, viewport performance when you are dealing with complex models.
Here is the benchmarks of the viewport performance of SolidWorks, Catia and other software https://www.gamersnexus.net/guides/3399 ... y-and-more

To the end that it could be interesting and provide more data points on what could be improved (coparing boot time or execution time of different operations). This would also be helpful with different workbenches that are tackling different industries
BIM (Archtecture)
Trails (Earthworks and Civil Transportation)
Render WB
FEM (Finite Element Analysis
etc....

Looks good. It's important to find out how to measure something that would be non-linear, i.e. not like
"min(timeit.repeat(stmt=f"[x**2 for x in range(1000000)]", number=500, repeat=10)".

The opposite. Using FreeCAD as a load to benchmark pc hardware.

I think it sound like a fun idea. The question is what can be used as a load? Maybe a FEM project? That is calculation heavy and gives a cool looking visualization when it's complete.

A pretty good point - to visualize something and not just show the results of computations...The more graphics we have, as in blender(3-4 tests) the more output for benchmarks in a review article.
Maybe the following pipeline:
1. Create a complex model
2. FEM analysis
3. Render the stylized results (layers of 2 + 1)

[kkremitzki]

Even though we would really mostly be testing OpenCASCADE functionality underneath, promoting the use of FreeCAD as an interface for such CAD testing is a good thing. This topic made me think of the Phoronix Test Suite and OpenBenchmarking.org, see for example https://openbenchmarking.org/suites