Hello Girls, Ladys, Boys and Gentlemen.

I want to create Game Assets with Blender, for the use in a Game in Unity.

It will not be High Poly, but with PBR Textures.

For this example I just made up a 3 minute House to show you the problem.

The First Picture with the yellow Roof is one Object made out of one Block. Without using a second Object.

The Second Picture is a house, completely made out of the objects laying on the "floor" and just put together without

connecting any vertexes. So its a bunch of loose objects.

Should I avoid the second solution or can i go with it whitout big problems? Its just easy as hell and done quickly. The First Version is "Waterproof" but

does this make a difference? I have searched for a "Best practise" in Game Assets, but I only found the typical "Avoid geometry whenever possible" Answer.

What is the BEST PRACTISE for Game Assets made in Blender.

It depends on what we're doing with or using the data. 

I would choose the second option but not stop there. If I'm placing multiple 'like' items in a scene (game side) then I'd be thinking about leaving room on the uv space for maybe interior stuffs...but with the goal in mind to limit how many times I have to change textures in my scene. I argue that an artist can also help the result frame rate. 

The other reason I like the second option is, now we're moving into the modular asset concept. Ton of good to be had here. The first to me is more of a design template approach or fine detail concept exploration before retopology to desired engine spec. 

But then again, if I'm going for ultra high realism...hard to say, still would be concerned with what couldn't be seen in a high poly requirement. Good looking stuffs  

Thank you for your feedback! I am pretty glad you chose version 2 to be honest. I wonder if its a problem if I split the House in section 2 into 3 objects. Roof, Wood and Wall. So i can add the material easier. With 3 UV Maps then and 3 Texture packs then. (normal, albedo, roughness).

Is this uncommon? Do you know if its better to have 1 Object (joined in blender) or is it okay to have 3 objects grouped together

common is relative. 

I'm pretty sure Unity does a fine job of some sort of batching under the hood. I was only considering the pro's to option #2 and admit to a bias toward that usage. 

again, depends on the final usage. I'd consider a full (background, maybe local depending on how big) scene in blender an optimization as long as the textures were also arranged engine friendly. Then detail and dynamic separately. 

For instance, Unity (I believe) will produce one draw call per material in the mesh. So to reduce this, one has to consider how the engine works instead of what is easier because easier may result as slower. The keyword you'll be after is "texture atlas" that would be the 'common' approach but your implementation may be unique to your game needs. That and vertex grouping types of thoughts I make more important than my artist comforts. Draw call count is a good one for the artist to take the lead on especially in an environment like Unity. 

https://docs.unity3d.com/2020.1/Documentation/Manual/DrawCallBatching.html you may find interesting, located in the unity manual (graphics/optimizing graphics performance) that speaks to what I'm attempting to describe.