Thin instances from a merge of meshes VS single meshes thin instances?

[mawa]

Dears,
I have a question for you in terms of handling and performance with regards to thin instances.
See below one of my cases with this restaurant area. I have a chair mesh and a table mesh.
In the end, there are currently 14 tables and 56 chairs in this restaurant area.
Each group/prefab formed by a table and 4 chairs is ‘copied’ and positioned/translated only. Currently, during the sketching phase, I was just instanciating the 4 chairs, parenting them to table to form a group and next creating 14 instances from this hierarchy.
Now I thought I would want these static objects to become thin instances.

Question is: Is it better to create a merge of the group (table + 4 (cloned) chairs) and create thin instances from this object group using translation only - - OR - - is it better (in terms of performance) to keep the chair and table meshes as the source and form new groups of thin instances (with Matrix.Compose I believe, is it)?
I made a try creating thin instances from a merge with multimat. Didn’t notice anything in particular and would likely be easier to handle (in case I want to change this layout, add or remove groups).
But what do you think the impact in terms of performance is?

Babylon.js Playground

[Evgeni_Popov]

If you don’t plan to move the chairs and the table independently, then it’s better to have a single mesh which is comprised of the table+4 chairs (T4C).

If you plan to move in the restaurant and have some tables/chairs hidden from view because they are behind you, then it’s probably better to have separate groups of T4C and create instances from a source T4C. If you don’t plan to move or most of the tables/chairs will always be visible, then creating thin instances of the source T4C is the best way.

[mawa]

Thank you, This was my plan A so I’m gonna take this answer without any further questions