Fallout 4 Optimization and Performance Systems Explained

[Giggilyomeromicon]

Alright, I've been trying to figure out what the hell is going on with one specific area not culling correctly. It seemed like just one cell is causing problems, but after looking some more I'm getting more confused. There might be something really bizarre going on with the terrain LOD.

 

http://i.imgur.com/vvM2sLh.jpg

 

http://i.imgur.com/fWtFwCy.jpg

 

For some reason while standing in certain positions on the map the game starts absolutely flooding draw calls and I really have no idea why. There's nothing visible in either of these pictures besides the LOD and low detail terrain. When I move to the same places above ground there are actually much fewer draw calls.

 

http://i.imgur.com/PLIowMl.jpg

 

There are also these seams underground that, for whatever reason, have shadows being drawn on them. Here's a video of the lighting on these bugging out:

 

 

I am very, very confused. However, the above ground FPS problems seem to happen near and while looking in the direction of these cell boundary seams. If I stand in the right spots and look at the ground I can wrack up 10,000 draw calls, and if I look up they drop to just a few thousand.

[zilav]

Check those terrain LOD meshes in NifScope.

[SMB92]

Where is that exactly? I cant make it out on my phone lol

[VIitS]

The sheer amount of laziness and lack of optimization in Vault 88 is astounding.

[Giggilyomeromicon]

Check those terrain LOD meshes in NifScope.

 

Looks like they're there in the mesh.

 

http://i.imgur.com/rMWtrCK.jpg

 

Where is that exactly? I cant make it out on my phone lol

 

It's the area between Faneuil Hall and the airport.

 

e: Okay, after substantially more messing around I've figured out that it's the water material and settings being used by the mod Natural Atmospheric Commonwealth version 9b. Running the game without that mod still results in some FPS loss, but it's an 8k draw call improvement.

Edited May 17, 2017 by Giggilyomeromicon

[VIitS]

Found something interesting when trying to get Scrap Everything to work right for Vault 88 (seriously, f*#@ that cell): The XPRI record is not touched when you generate precombineds, only when you generate the previs.

[SMB92]

Found something interesting when trying to get Scrap Everything to work right for Vault 88 (seriously, f*** that cell): The XPRI record is not touched when you generate precombineds, only when you generate the previs.

Interesting. Can't say I think much of that DLC. Spent about 15 minutes in there, haven't been back since. Talk about "Improvised" lol.

 

Gave up on this business for now, busy building new spawn mod :)

[Trosski]

I know some of you are aware of my Tutorial video on this subject, but I'll add a link here just in case some aren't aware of it.

 

I have found that when creating compatibility patches sometimes the process doesn't work correctly. The times you will find this to be true is when two mods add items to the same spot in a cell. For instance My Cleaner Settlements mod changes the stairs up to one of the exits in Hangman's alley by adding in some shack floors with supports, and the mod Hangman's Alley Interior Apartments removes the rubble piles, and adds some brick wall components for support. When you make a patch in this case you will find that instead of one mod winning over the other(even if you delete one of the objects in the patch) both mods add the objects, and they will clip into each other. At this time, the only work around I have found is changing one of the mods directly to pick a winner in this way, and then making a patch.

 

I have also found that sometimes patches aren't necessary. If you have a comprehensive mod like my cleaner settlements, and you want to use another mod like Starlight is bright again for one of the settlements, then just loading the second mod later will overwrite the first mod, and not cause any conflicts. I know that may seem like a duh kind of statement, but I get so into ensuring compatibility with mods I use, that i can get a bit trigger happy with the precombine/Previs system, and it can end up causing more problems than it solves.

 

I wish I had found this discussion earlier. You guys have been having some really in depth discussion on this issue, and I have picked up some extra info I wasn't aware of before.

[SMB92]

I know some of you are aware of my Tutorial video on this subject, but I'll add a link here just in case some aren't aware of it.

Â

I have found that when creating compatibility patches sometimes the process doesn't work correctly. The times you will find this to be true is when two mods add items to the same spot in a cell. For instance My Cleaner Settlements mod changes the stairs up to one of the exits in Hangman's alley by adding in some shack floors with supports, and the mod Hangman's Alley Interior Apartments removes the rubble piles, and adds some brick wall components for support. When you make a patch in this case you will find that instead of one mod winning over the other(even if you delete one of the objects in the patch) both mods add the objects, and they will clip into each other. At this time, the only work around I have found is changing one of the mods directly to pick a winner in this way, and then making a patch.

Â

I wish I had found this discussion earlier. You guys have been having some really in depth discussion on this issue, and I have picked up some extra info I wasn't aware of before.Â

You can load both mods as ESMs into CK and generate a new precoms esp with both mods (after removing all precoms data from both first and doing any hand edits in the render).

 

Thats how I generate mine.

[Trosski]

 

You can load both mods as ESMs into CK and generate a new precoms esp with both mods (after removing all precoms data from both first and doing any hand edits in the render).

 

Thats how I generate mine.

 

That's good to know. I haven't messed around much with making .esps into masters, because the few times I tried it the CK didn't want to play nice with the ESM i created.