I have a question that I'm hoping some of the experts here can help clarify.

I model the same competition parts in Onshape and sometimes compare my results against SOLIDWORKS reference values. I consistently observe very small volume deviations, typically around 0.00025–0.00032 mm³.

Since both systems use the Parasolid kernel, I'm curious whether these differences are primarily caused by tolerance handling, feature implementation, or rebuild algorithms rather than the kernel itself.

One thing I have noticed is that the easier competition levels often round values to the nearest decimal place, which makes matching results straightforward. However, once the topology becomes more complex with lofts, shells, fillets, and more advanced geometry, it becomes nearly impossible in Onshape to match the reference volume down to the final decimal place.

For those of you with experience using both systems:

• Have you observed similar behaviour between Onshape and SOLIDWORKS?
• Is a deviation of around 0.0003 mm³ effectively just numerical noise?
• At what point would you consider a volume difference significant enough to indicate an actual modelling error rather than differences in tolerances or implementation?

I'd appreciate any insight from users who have worked extensively with both platforms or who have experience with Parasolid-based systems.