Hello all! When importing jt and catia files it seems like all the surfaces of the parts are turned into separate actors. As the car model I want to import now is 9000 parts big this is a bit of hazzle. One 0.5 mb file i imported was turned into 132 different actors (a controller chip, so not a big part). One part is 300 mb big and i have not been able to import it as my memory runs out (32gb) at 5 cm chord length. I can understand why it happens but is there any way i can make it not happen?
CAD plugin prior to 4.20 was importing patches contained in CAD parts as single actors. It has been changed in 4.20, e.g. for Catia files it would be a static mesh per body in a CATPart. If you are still using 4.19, you can batch merge your imported actors using python.
There is one issue with the management of units in 4.19: if the unit of your CAD model is in mm your 5 cm chord length could be actually converted to 5 mm when tessellating. If your model CAD unit is mm then try using 50 cm in your chord length.
I would recommend you to try again on 4.20.
Also, if you are doing CAD preparation on a regular basis you might want to invest in a “preparation pc” configuration with more than 32G ram.
Thank you for your reply Flavien, Its Catia and teamcenter so its definetly in mm. I will test with a larger chord length. It is the 4.20 preview 3 im using now for my imports, and it still created separate surfaces, with JT. When i tested another part that was in catproduct, it worked very well though! Could JT files be the problem?
Regarding PC if you would have the time to answer, it is also one of the areas of question I should answer for this project. What woyuld you recommend? My colleagues following a classic Vred workflow for visualization have intel xeon, 256gb ram, dual p6000 in sli. Management wont buy hardware for “just a game engine” if they dont know and understand that such hardware will hurry the pipeline, will it do that? Can i find information about specifically unreal studio recommendations? Will more ram mean linearilly better import times? Thank you.
For import of complex assemblies, you will benefit from faster CPUs right now and lots of RAM (I would have thought 64Gb would be plenty, but I don’t know the complexity you’re dealing with). Cores will matter if you’re using Lightmass in your interiors. Lightmass speed is almost linear with CPU cores, probably 1.7 or something. Similar to rendering linearity.
Don’t forget, in 4.20 we have two-pass CAD import, so the first pass just loads the assembly tree, allowing you to make decisions, merge or delete things BEFORE you actually load it. This will save you lots of time and require less RAM.
Thank you, I actually thought the crash submit was the bug report. I will use the other form. I tried a JT file now in preview 3 with chord 40, and it worked so it definetly had something to do with the chord length. So thank you! Next step the whole car! I will try with a released model so i can maybe send you something.
Where can i read about the two-pass, this sounds exactly what I would need! Its not implemented yet right? Cause I dont have that possibility today I dont think. Thank you for the hardware tips! Will pass on. And thank you again for a great tool! I know several departments within CEVT are experimenting with it! Have a great day.
Using a higher chord error will reduce the number of generated triangles and thus the memory used for generating the assets. However with a high chord error you ll lose precision on the resulting meshes. So you might still need to invest in RAM so you can perform tessellation at the expected level.
RAM and CPU will guarantee a smooth experience in your editor when you are working on the scene and rendering your application. However, as you are preparing CAD data (whole car) these requirements are higher and closer to what computer running CAD software are using. On the GPU side: UE4 does not support multi-GPU (SLI), so aim for a high-end single graphic card.
If needed you can split your work on two computers: one for preparation (strong CPU and a lot of RAM) and one for managing your scene and running the experience (a high-end game pc).
Regarding the meshes generated when importing JT: how are the objects grouped in the JT file? Is there a part per body?
Thank you very much all! This is very good information! Very helpfull! Yes it definetly seems like a whole car is too much for the computer I am currently on and another CAD desktop is needed. But no SLI then, but the high end gpus as p6000 and alike should work, as long as its just one? Regarding the jt parts, they where created as several static mesh actors grouped under a scene component. The chip-set i mentioned was one small 56kb part created as one main scene component, and several child scene components with the surfaces as child actors to these. Am i making sense writing like that?
Now i understood you meant the strucutre of the jt file in catia, I cant reach catia unfortuneatly. I just recieve the exported models from teamcenter.