A well made model will include several Resolution LODs that the game engine will switch in between.
How the engine determines which of the available LODs to use is a complex issue that involves several factors:
- View distance (how far the player is from the object he is looking at)
- Visible objects (how many objects are currently displayed on the screen)
- Video settings (visual quality, terrain detail, shadows, etc.)
- CPU load (how much processing power is left to handle visible objects)
- Object size (how large an object is)
To demonstrate the effect of these different parameters a series of tests have been made, and are documented below.
Disclaimer: Keep in mind that the numbers listed are only useful to demonstrate the interdependency of the different factors. They are not to be taken as absolutes, since they will vary from situation to situation. (i.e. if you reproduce this test on your system you will most likely receive different absolute numbers, since your hardware/software configuration will be different from the test setup.)
The results should give you some clues as to what kind of LODs make sense, and how they are chosen, but don't interpret them as hard numbers.
For A3: Note that the numbers of the resolution LOD's do not have any impact on range, they only determine the order
For Proxied Objects in A3: Proxies within models change their resolution LOD independent of the main object. If you have a MG as a proxy in a tank model, the MG might switch down resolution while the tank is still in first resolution. (It hasn't been tested if this is the case for previous games)
Test Results 编辑
The official BIS 5tTruck uses the following resolution LODs:
If you test this on an empty map (desert), with just one truck visible, you will get the following switching distances:
(Which means that LOD 0.75 will be used from 0-175 meters, LOD 1.50 from 175-310 meters, and so on.)
View distance isn't the only variable that comes into play though.
If, instead of one truck, you look at ten instead, you will get the following results:
As you can see, the switching distances are significantly smaller.
Instead of switching to LOD 1.50 at a distance of 175 meters, with 10 trucks visible LOD 1.50 becomes active at a distance of 40 meters already!
If you tighten the whole range a bit (instead of going from 0.75 to 9.00 you go from 0.00 to 5.00), you will see that the lowest resolution LOD (5.0) becomes active at 190 meters already. (Whereas in the original setting the car wasn't even visible anymore at the highest LOD, as it was more than 1000 meters away.)
Game settings also come into play, when it comes to LOD selection, as you can see from the following example.
The first series was done with video settings set to 'High Detail', and the second one with 'Low Detail'. Since in the 'Low Detail' setting fewer other objects/polygons are visible, model LODs are switched later:
To demonstrate a very extreme setting, you will see that with LOD settings of 100 and more, the second LOD never even comes into play with a single vehicle, and even with 10 vehicles it only switches from the highest LOD to the second stage one at a distance of 520 meters: