I’d like to know the math behind why
Shows a tile size of 1017x1017
While
Doesn’t include this size but instead has 1009x1009 as the closest matching.
I have attempted using both sizes as tiles.
Performance is rather poor on both.
In both cases the default levels are made up of a movable directional light, a skylight, and the sky-sphere.
There are only 9 tiles for testing purposes - no PP, so the default settings for epic are in use.
When the project is launched however, the performance at 4k res is far beyond that of a single map with a 1009 tile.
I’m getting around 120fps on the single map, but around or less then 70fps on both the world composition maps.
With the 1009 tiles performing slightly better apparently.
Could someone - aside from explaining why the docs have 1017px vs the recommendation on the maps is 1009px - suggest a tile size that may perform better overall at 4k?
Testing 505x505 tiles did not change the FPS output.
----- Steps to test -----
step 1:
Create a base level.
Step 2:
Add a skylight, a directional light, player start, and BP_Skysphere
step 3:
Configure directional light (make movable). Configure BP_Skysphere (select skylight).
Step 4:
Change game mode - Create BP if needed. use variation of 3rd person sample.
Additionally, Enable World Composition.
Step 5:
Save
Step 6:
Import these tiles (1017x1017). Use 2x2 settings.
[1017x1017 black tiles set][1]
Step 7:
move tile 0,0 to the be at the center of the player - to do this, load the tile and move it in the world composition map viewer. It should snap. save the level, unload it.
step 8:
Move the Player Start z to -25507.800781 within the persistent level. save.
Step 9:
Disable Real Time in editor, make sure your project doesn’t use frame smoothing.
Step 10:
Launch the game.
Step 11:
` stat startfile
let it run a few secs, jump around.
` stat stopfile
` exit
We succesfully created the first log.
Save it out.
It will be in your project directory
/Saved/StagedBuilds/WindwsNoEditor/“projectname”/saved/Profiling/UnrealStats/“nameofmap”.“WindowsNoEditor”.Date"/“filename”.ue4stats
You need to move it to a safe location, as I noticed the engie overwriting it.
Step 11:
Open the level called tile_00_X0_Y0 within the sub folder that UE4 automatically generated when importing tiles.
Step 12: (same as step 2)
Add a skylight, a directional light, player start, and BP_Skysphere
Step 13:
Move all components to 50800.0x, 50800.0y. 0z. (this is the relative 0,0,0 of the level).
Take care to use -25507.800781 as playerstart’s Z.
step 14:
Configure directional light (make movable). Configure BP_Skysphere (select skylight).
Step 15:
Change game mode
Step 16:
Save and Launch.
step 17:
Minimize Editor
Step 18:
` stat startfile
let it run a few secs, jump around.
` stat stopfile
` exit
step 19:
Compare the 2 results: Window > Developer Tools > Session Front End.
Step 20:
Create a folder with both stats.
Step 21:
Load the folder in session front end.
Step 22:
Have fun analyzing data.
The easiest thing to look at is the FPS Chart resutls.
Tiled Map FPS:
vs
Single Map FPS:
I’ll admit, the sample is small and the results here are more then acceptable here.
FPS is hardly anything of substance to check within all of this data, However, you can notice that the MIN FPS is better by 10+ frames in the Single Map vs World Composition.
Whereas the MAX fps is even more drastically offset by 188 vs 150 within the 2 data sets.
OFC, you can also notice some sort of jump at the beginning of the tiled map vs none at the beginning of the single map, which may have thrown off my results as well.
But the point I’m trying to make, and what I’m asking is, Why is there such a difference in MAX fps between having world composition On or Off with only the same level being loaded?