Jump to content

Recommended Posts

Posted (edited)

 

81FPSwithoutgrassLOD.thumb.jpg.e3b3cab44c610e0884f00654fe228d5e.jpg

First image with no grass LOD I am getting a stable 81 FPS. Besides the grass LOD, the Dyndolod settings are the same for both images.

53FPSwithgrassLODdensity25.thumb.jpg.fb22b236aafd345513cf95dd45b6c593.jpg

Second image with grass LOD I am getting a stable 53 FPS.

For grass cache generation I used NGIO with superdensegrass=false and Dyndolod Mode=1. Veydosebrom mod default INI settings were iMaxGrassTypesPerTexure=15 and iMinGrassSize=20, so I carried forward the same settings in NGIO for generating the grass cache. I also have Grass Helper NG installed with default settings and used it when generating both the grass cache, TexGen and Dyndolod outputs.

The grass billboards were the default res for 1920x1080 and not HD. Then I chose 25 density in Dyndolod and grass mode 1.

Is a 28 FPS drop like this normal? I expected a significant performance hit but this seems excessive and took far more FPS than 3d tree LOD with Nature of the Wild Lands mod installed. The grass LOD's alone hit as hard as an ENB preset so I'm wondering if some setting is problematic or if I'm just underestimating the performance cost of grass LOD.

The grass LOD is also much greener than the full grass in the loaded cell. I will try to fix this later with the INI color setting but for now I am more concerned about performance. It seems like the LOD grass is still more dense than the full grass and doesn't match the grass types of the full grass. I'm not sure if it is the color difference and 2D/3D difference which are creating this impression or if grass LOD generation has other limitations. I'm using Obsidian Weathers but no ENB here.

Edit: Also, from what I understand the NGIO settings 'Overwrite-grass-distance = 7000.000000' and 'Overwrite-grass-fade-range = 18000.000000' only affect the full grass visibility in the loaded cell and make no difference for generating the grass cache and Dyndolod grass LOD generation.

Edited by drift123
extra info
Posted
  On 6/7/2024 at 2:06 AM, drift123 said:

 

81FPSwithoutgrassLOD.thumb.jpg.e3b3cab44c610e0884f00654fe228d5e.jpg

First image with no grass LOD I am getting a stable 81 FPS. Besides the grass LOD, the Dyndolod settings are the same for both images.

53FPSwithgrassLODdensity25.thumb.jpg.fb22b236aafd345513cf95dd45b6c593.jpg

Second image with grass LOD I am getting a stable 53 FPS.

For grass cache generation I used NGIO with superdensegrass=false and Dyndolod Mode=1. Veydosebrom mod default INI settings were iMaxGrassTypesPerTexure=15 and iMinGrassSize=20, so I carried forward the same settings in NGIO for generating the grass cache. I also have Grass Helper NG installed with default settings and used it when generating both the grass cache, TexGen and Dyndolod outputs.

The grass billboards were the default res for 1920x1080 and not HD. Then I chose 25 density in Dyndolod and grass mode 1.

Is a 28 FPS drop like this normal? I expected a significant performance hit but this seems excessive and took far more FPS than 3d tree LOD with Nature of the Wild Lands mod installed. The grass LOD's alone hit as hard as an ENB preset so I'm wondering if some setting is problematic or if I'm just underestimating the performance cost of grass LOD.

The grass LOD is also much greener than the full grass in the loaded cell. I will try to fix this later with the INI color setting but for now I am more concerned about performance. It seems like the LOD grass is still more dense than the full grass and doesn't match the grass types of the full grass. I'm not sure if it is the color difference and 2D/3D difference which are creating this impression or if grass LOD generation has other limitations. I'm using Obsidian Weathers but no ENB here.

Edit: Also, from what I understand the NGIO settings 'Overwrite-grass-distance = 7000.000000' and 'Overwrite-grass-fade-range = 18000.000000' only affect the full grass visibility in the loaded cell and make no difference for generating the grass cache and Dyndolod grass LOD generation.

Expand  

No logs were provided. The performance requirement depends on the load order, hardware and settings. Choose mods, settings and generate the LOD patch according to your needs.

iMaxGrassTypesPerTexure means there can be more grass per landscape. A setting of 15 is 7.5 times higher than the default of 2, though how much grasses are there per texture depends on records in plugins and grass mods. iMinGrassSize of 20 is the default density. Increasing the value means less dense grass.

If grass LOD does not match the full grass, then likely reasons are that a different grass cache was used for LOD generation, the game not using the grass to place full grass or the load order changed after generating LOD.

See https://dyndolod.info/Help/Grass-LOD#Performance

Posted
  On 6/7/2024 at 6:47 AM, sheson said:

No logs were provided.

Expand  

The debug log is 157 mb so I can't attach it but I put the DynDOLOD_SSE_log.txt which is for the generation with grass LOD at 53 FPS.

There was a warning about a missing mesh from ELFX but I assumed that isn't relevant for the grass LOD in Whiterun.

There are some other warnings too but they didn't get listed in the summary page after generating LOD. Several of these warnings including the ELFX warning were present before generating grass LOD where I was getting 81 FPS but I don't know for sure if they were all present.

I'm going to test the extended grass feature in NGIO and see what the performance impact is in comparison to the grass LOD.

DynDOLOD_SSE_log.txtFetching info...

 

  On 6/7/2024 at 6:47 AM, sheson said:

No logs were provided.

Expand  

Here are the INI settings for grass-related mods that were used:

Veydosebrom INI:

  Reveal hidden contents

NGIO Grass Control

  Reveal hidden contents

Grass Cache Helper NG INI:

  Reveal hidden contents

 

Posted
  On 6/7/2024 at 7:09 AM, drift123 said:

The debug log is 157 mb so I can't attach it but I put the DynDOLOD_SSE_log.txt which is for the generation with grass LOD at 53 FPS.

There was a warning about a missing mesh from ELFX but I assumed that isn't relevant for the grass LOD in Whiterun.

There are some other warnings too but they didn't get listed in the summary page after generating LOD. Several of these warnings including the ELFX warning were present before generating grass LOD where I was getting 81 FPS but I don't know for sure if they were all present.

I'm going to test the extended grass feature in NGIO and see what the performance impact is in comparison to the grass LOD.

DynDOLOD_SSE_log.txt 973.33 kB · 0 downloads

 

Here are the INI settings for grass-related mods that were used:

Veydosebrom INI:

  Reveal hidden contents

NGIO Grass Control

  Reveal hidden contents

Grass Cache Helper NG INI:

  Reveal hidden contents
Expand  

As explained in the first post, my signature or https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs,  zip large log files and use a file service to upload them.

You do not seem to use seasons, so it is unclear why you are using Grass Cache Helper NG.

If you want to compare performance requirements of full grass to grass LOD, make sure to extend it to the same distance as LOD Level 4 as explained in https://dyndolod.info/Help/Grass-LOD#Performance

If you want to decrease performance requirements of grass LOD, reduce the density and/or shorten the LOD Level 4 distance as explained in https://dyndolod.info/Help/Grass-LOD#Performance

https://dyndolod.info/FAQ#Long-running-time-or-output-several-GB-in-file-size

Posted
  On 6/7/2024 at 7:35 AM, sheson said:

As explained in the first post, my signature or https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs,  zip large log files and use a file service to upload them.

You do not seem to use seasons, so it is unclear why you are using Grass Cache Helper NG.

If you want to compare performance requirements of full grass to grass LOD, make sure to extend it to the same distance as LOD Level 4 as explained in https://dyndolod.info/Help/Grass-LOD#Performance

If you want to decrease performance requirements of grass LOD, reduce the density and/or shorten the LOD Level 4 distance as explained in https://dyndolod.info/Help/Grass-LOD#Performance

https://dyndolod.info/FAQ#Long-running-time-or-output-several-GB-in-file-size

Expand  

Debug log.txt:
https://ufile.io/m4yuf08u

The grass cache guides I read instructed to use Grass Cache Helper NG. I guess I could set those INI settings manually...

With extended full grass to 60000 blocks on Level 4 I had about 23 FPS compared to 53 FPS with grass LOD in the same range. So I guess grass LOD is just much more taxing than I had expected.

What is interesting is I only lost about 3 FPS in Whiterun tundra moving the slider from 60000 to 0 for Level 4 LOD without grass LOD installed but otherwise the same settings. So the grass LOD appears to be much more taxing than the 3D tree and object LOD in Level 4, because I lose about 30 FPS by enabling just the grass and keeping everything else the same. From what I remember the switch from 2D ultra tree LOD to 3D trees was about a 10 FPS difference.

Would it be possible to create a rule so that Level 4 grass LOD is at 30000, while keeping the Level 4 ultra tree & object LOD at the 60000 range? I'd rather shorten the range of the grass LOD only. I also faintly remember being able to bake grass textures into the terrain LOD before this grass LOD was possible, which could look fine on the more distant terrain.

Posted
  On 6/7/2024 at 11:49 AM, drift123 said:

Debug log.txt:
https://ufile.io/m4yuf08u

The grass cache guides I read instructed to use Grass Cache Helper NG. I guess I could set those INI settings manually...

With extended full grass to 60000 blocks on Level 4 I had about 23 FPS compared to 53 FPS with grass LOD in the same range. So I guess grass LOD is just much more taxing than I had expected.

What is interesting is I only lost about 3 FPS in Whiterun tundra moving the slider from 60000 to 0 for Level 4 LOD without grass LOD installed but otherwise the same settings. So the grass LOD appears to be much more taxing than the 3D tree and object LOD in Level 4, because I lose about 30 FPS by enabling just the grass and keeping everything else the same. From what I remember the switch from 2D ultra tree LOD to 3D trees was about a 10 FPS difference.

Would it be possible to create a rule so that Level 4 grass LOD is at 30000, while keeping the Level 4 ultra tree & object LOD at the 60000 range? I'd rather shorten the range of the grass LOD only. I also faintly remember being able to bake grass textures into the terrain LOD before this grass LOD was possible, which could look fine on the more distant terrain.

Expand  

Grass cache generation with NGIO does not require Grass Cache Helper NG.

Object LOD uses 3 LOD level, 4, 8 and 16. Whatever the LOD meshes of level 4 depict, shows in the area between the edge of the active cells and the object LOD level 8 meshes and can not be made to stop rendering earlier. That would require changes to the engine and/or shaders.

See https://dyndolod.info/How-LOD-Works and https://dyndolod.info/Help/Object-LOD

Posted
  On 6/6/2024 at 9:56 PM, KRZ said:

This one's looking good in xEdit. Unknown 2 only. No crashes in game either. I've attached DynDOLOD logs, just in case.

DynDOLOD_SSE_Debug: https://mega.nz/file/1OVi3ATa#WQN_Iquyedx69e_4LHhKnCn8BMfQCy4DbkwtNf9cHfw

I'll start another run with Sirenroot and SLaWF and share the results once I got them.

Great success. Unknown 2 only.

DynDOLOD_SSE_Debug: https://mega.nz/file/0GdmWY6B#PuPZGFUWPAUm1a1rM7S0YbUGebjXv7GjKfrk2mmc_TU This one's with Sirenroot + SLaWF.

The text-files containing 'Complete' in their title are including Sirenroot + SLaWF.


May I ask how the mismatch came to happen? It sounds like something more people should encounter so I'm still sceptical I messed up somewhere.

Thank you for all the troubleshooting help!

Error Check 170C.txt 1.47 kB · 1 download DynDOLOD_SSE_log.txt 1.15 MB · 1 download

Error Check 170C Complete.txt 1.47 kB · 1 download DynDOLOD_SSE_log Complete.txt 1.15 MB · 1 download

DynDOLOD.esm Complete.zip 1 MB · 0 downloads

Expand  

Let me know if this test version works - no unresolved form IDs and no crashes. https://mega.nz/file/BBAWlCwA#wn7-2iAs803THnasndswcCYH0wFWwV473creTLtoK8I

Let me know if Alpha-171 fixes the issue properly for you.

For speed DynDOLOD creates the large reference list on the worldspace record directly without using the usual xEdit functions that also take care of load order form IDs and the master list being in sync. This requires the master list to be final when DynDOLOD creates the list. A change somewhere else means if circumstances are right, the master list could change after, which is hopefully addressed in this test version.

  • sheson changed the title to DynDOLOD 3.00 Alpha 171
Posted (edited)
  On 6/7/2024 at 1:06 PM, sheson said:

Let me know if this test version works - no unresolved form IDs and no crashes. https://mega.nz/file/BBAWlCwA#wn7-2iAs803THnasndswcCYH0wFWwV473creTLtoK8I

Let me know if Alpha-171 fixes the issue properly for you.

For speed DynDOLOD creates the large reference list on the worldspace record directly without using the usual xEdit functions that also take care of load order form IDs and the master list being in sync. This requires the master list to be final when DynDOLOD creates the list. A change somewhere else means if circumstances are right, the master list could change after, which is hopefully addressed in this test version.

Expand  

Sorry I missed the test version and only now respond to 171 :( It's looking good however. I've attached the xEdit Error Check below but there are no unresolved FormIDs and I've had no crashes either. I think this could be it. Thank you for continously troubleshooting and also for the easy to understand explanation of what's been wrong!! 

Error Check 171.txtFetching info...

Edited by KRZ
Posted
  On 6/7/2024 at 6:05 PM, KRZ said:

Sorry I missed the test version and only now respond to 171 :( It's looking good however. I've attached the xEdit Error Check below but there are no unresolved FormIDs and I've had no crashes either. I think this could be it. Thank you for continously troubleshooting and also for the easy to understand explanation of what's been wrong!! 

Error Check 171.txt 1.47 kB · 0 downloads

Expand  

Looking good. Thanks for testing and letting us know.

Posted
  On 6/7/2024 at 12:12 PM, sheson said:

Grass cache generation with NGIO does not require Grass Cache Helper NG.

Object LOD uses 3 LOD level, 4, 8 and 16. Whatever the LOD meshes of level 4 depict, shows in the area between the edge of the active cells and the object LOD level 8 meshes and can not be made to stop rendering earlier. That would require changes to the engine and/or shaders.

See https://dyndolod.info/How-LOD-Works and https://dyndolod.info/Help/Object-LOD

Expand  

Alright thanks for the info I'll have to keep testing to see what looks and performs best.

Posted (edited)

If I increase the imingrasssize to 120 for grass precache to save performance, could I then change that setting back to 20 after the grass LOD is generated, so that the loaded cell grass density is not affected? I want to save performance on the grass LOD while maintaining the density of full grass in the loaded cell. The idea is to reduce the number of placements drastically.

Would this approach require me to keep bAllowCreateGrass enabled after the fact? As I understand the grass cache files affect both the loaded cell full grass and the grass LOD if bAllowLoadGrass is enabled.

I'm also unsure if the grass density setting in Dyndolod is identical to changing the imingrasssize during grass precaching? Or do these settings work differently? Even at 25 the performance hit was quite heavy and the grass LOD still looked very dense, which makes me think that reducing density for the precache generation may be the solution.

Thanks.

Edited by drift123
correction
Posted
  On 6/10/2024 at 1:15 AM, drift123 said:

If I increase the imingrasssize to 120 for grass precache to save performance, could I then change that setting back to 20 after the grass LOD is generated, so that the loaded cell grass density is not affected? I want to save performance on the grass LOD while maintaining the density of full grass in the loaded cell. The idea is to reduce the number of placements drastically.

Would this approach require me to keep bAllowCreateGrass enabled after the fact? As I understand the grass cache files affect both the loaded cell full grass and the grass LOD if bAllowLoadGrass is enabled.

I'm also unsure if the grass density setting in Dyndolod is identical to changing the imingrasssize during grass precaching? Or do these settings work differently? Even at 25 the performance hit was quite heavy and the grass LOD still looked very dense, which makes me think that reducing density for the precache generation may be the solution.

Thanks.

Expand  

https://dyndolod.info/Help/Grass-LOD#Settings
Change the Density dropdown to a lower value in the advanced mode of DynDOLOD under the Grass LOD checkbox to thin out grass LOD for better performance.

https://dyndolod.info/Help/Advanced-Mode#Grass-LOD
Density sets the percentage of grass LOD billboards that are being generated. 50 means only half of the grass placements are added for better performance but thinner grass LOD in the distance.

Set it to 15 or lower to further reduce performance requirements.

Posted (edited)

On grass density 10 I lose about 15-20 FPS (~80 to 60) in the Whiterun tundra compared to no grass LOD.

It seems that the color of the LOD matches the full grass, except for some reason the LOD only displays the green grass models and not the brown grass. Is this functioning as intended?

Edit: well I tried an ENB and then the green LOD grass changed to light brown...so I think it is just the weather and lighting doing this

image.thumb.jpeg.73b970e25515f6f1f4bc113bff788e44.jpeg

Edited by drift123
typo
Posted
  On 6/10/2024 at 9:18 AM, drift123 said:

On grass density 10 I lose about 15-20 FPS (~80 to 60) in the Whiterun tundra compared to no grass LOD.

It seems that the color of the LOD matches the full grass, except for some reason the LOD only displays the green grass models and not the brown grass. Is this functioning as intended?

Edit: well I tried an ENB and then the green LOD grass changed to light brown...so I think it is just the weather and lighting doing this

image.thumb.jpeg.73b970e25515f6f1f4bc113bff788e44.jpeg

Expand  

If a specfic type of full grass model has no LOD representation see https://dyndolod.info/Help/Grass-LOD#No-Grass-LOD-Check-List

  • Like 1
Posted
  On 6/10/2024 at 10:49 AM, sheson said:

If a specfic type of full grass model has no LOD representation see https://dyndolod.info/Help/Grass-LOD#No-Grass-LOD-Check-List

Expand  

It's just bizarre because when I enable ENB, suddenly both the green and brown grass types become visible in the LOD. I don't know how to explain that but the lighting is doing something funky. Strangely enough, the ENB makes the LOD match the full grass better than no ENB, it just needs to be a little darker I think.

TheElderScrollsVSkyrimSpecialEditionScreenshot2024_06.10-08_11_38_45.thumb.png.25211db9d6a7b88d2dece00a0b959a50.png

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...

Important Information

By using this site, you agree to our Guidelines, Privacy Policy, and Terms of Use.