While attending a Revit user group meeting a month or two ago someone mentioned that they prefer to use generic walls to keep their model size (file size) down. It seems reasonable that a compound wall might carry a bit heavier burden on a model but how much more of a burden? The poor scientist that I am I tried the following.
I saved an "empty" template file with no walls placed which resulted in a file of 2076 kb (used Default.rte, the stock Revit template). I then created 19 walls, 4 exterior and 15 interior walls using Generic 12" and 6" respectively. When I saved this file it was 2216 kb. Then I created another but used compound walls (Exterior - Brick and CMU on Mtl. Stud and Interior 5-1/2" Partition(1 hr)) instead with a resulting file size of 2452 kb for an increase or difference of 236 kb. I used the exterior wall type because it is more than just a wall with extra layers, it has embedded sweeps and reveals which should account for some of the difference as well.
A small number of walls. I thought I'd use more this time. I created another pile of walls for a total of 152. This is the result.
The generic walls file came in at 3184 kb and the compound version came in at 4092 for a difference of 908 kb. Here's a spreadsheet breakdown for you "column and row" types.
It seems to me that this means that modelling walls in bulk results in less cost (in kilobytes) per wall. The more walls the more economical their impact is. Not that more walls doesn't mean more...just not a literal 1:1 file size relationship. It also means that compound walls do have a "heavier" impact on the project file size too.
Since I increased the number of walls by eight (8) times I was curious to see if the kilobyte increase was literally eight (8) times as well. The How Close? column in the spreadsheet shows that the generic walls were nearly so but the compound walls were a bit more economical when the quantity increased, just a little over five (5) times as many kilobytes.
To make it more interesting we could edit the profile of walls, attach walls to roofs, embed curtain walls, disallow joins, add views to sheets using different Detail Level settings, place various hosted elements on/in walls, use the Opening Tools, compare curved versus straight impact...and on and on.
Do I have a point? Well...file size isn't really the only measure of a project's performance nor is it even really reliable to use to predict project performance. I've seen rather large projects that perform quite well apart from waiting for it to open initially. Conversely I've seen "small" projects that perform poorly. My interpretation of my results is that I shouldn't really need to worry about using generic walls only, at least not using just file size as a measure. Much more likely that how the walls interact with other elements in the project will affect performance more than just file size.