Thanks Johnwhelan for the explanation enlarge.
Regards
John
This is a simplified version of what happens.
The impact that an asset has depends on a number of things and on the hardware. In a built up area the number of different assets has the most impact due to the 300 mesh and 200 texture poly equivalent overhead. Within each screen shot you have a budget that the machine will support at a particular frame rate.
The size of the texture files has an impact so big uniform colour texture files are bad news. Generally speaking some content creators are better at producing good performing content than others. So if you find one who creates content you like stay with them. I’ve seen some scenery assets that have 30 texture files the poly count is a mere 300 polys but the impact of 30 texture files is about the same as 6,000 polys. Many of these were created in GMAX by users who didn’t understand how to UV map.
The mention of Sketchup should ring warning bells, it is technically possible to create low poly content in Sketchup but it is even more complex to do than using Blender and most Sketchup content creators don’t know how to do it.
Normal mapping can be used to emulate polys, there is a very good Australian Hotel that uses Normal mapping very effectively.
If you repeat an asset it has a much lower impact the second time. In TS12 splines written for earlier versions can kill performance. There isn’t really a simple way to identify the impact of an asset, but in both Surveyor and driver in options general show performance is a way to identify the worse performing asset. I suggest you click show kuid number as well.
Speedtrees use a different rendering system, dig in the forum and there are one or two people who know them fairly well and have created some reasonable ones. Don’t use more than five types.
Trainz is a rich environment that means there is room to make mistakes especially in performance. If you have a tame Blender content creator its quite easy to merge a number of scenery objects into one big asset and avoid the individual overheads.
You can of course just throw hardware at the problem, a faster machine has a bigger budget per scene but it’s possible to build a layout that will bring any machine to its knees.
Have fun
Cheerio John