The mission has low fps. Searching for a solution.

[OrbWeaver]

On 4/21/2021 at 11:04 AM, ddaazzaa said:

I have visportals around the manor although I don't know how optimally placed they are.

If anyone knows if there are better ways to visportal/optimize the exterior I'm happy to try and create better performance.

I recommend Springheel's video series on mapping, in particular the video about designing your map with the placement of visportals in mind.

From the screenshots @stgatilov posted, it looks like some of your portals are worse than useless. Portals are not a magic fairy dust that you can just sprinkle over your mission in a criss-cross pattern and hope that they will automatically improve performance — they won't (and might even make it worse). A portal which never closes, and does not substantially restrict the visibility of objects behind it, achieves nothing whatsoever.

If you specifically want an outdoor area which does not lend itself to visportalling, you might need to consider other techniques for performance optimisation, such as reducing draw calls by simplifying models, reducing light counts and shadow-casting objects, etc.

[stgatilov]

Just now, OrbWeaver said:

If you specifically want an outdoor area which does not lend itself to visportalling, you might need to consider other techniques for performance optimisation, such as reducing draw calls by simplifying models, reducing light counts and shadow-casting objects, etc.

There are almost no lights outside, and almost no shadows. @Gadavre has already discovered that shadow settings have absolutely no effect on this scene.

[ddaazzaa]

2 minutes ago, OrbWeaver said:

I recommend Springheel's video series on mapping, in particular the video about designing your map with the placement of visportals in mind.

From the screenshots @stgatilov posted, it looks like some of your portals are worse than useless. Portals are not a magic fairy dust that you can just sprinkle over your mission in a criss-cross pattern and hope that they will automatically improve performance — they won't (and might even make it worse). A portal which never closes, and does not substantially restrict the visibility of objects behind it, achieves nothing whatsoever.

If you specifically want an outdoor area which does not lend itself to visportalling, you might need to consider other techniques for performance optimisation, such as reducing draw calls by simplifying models, reducing light counts and shadow-casting objects, etc.

Yeah, I wasn't sure how to place visportals in a large open area. The concept of visportals is easier to understand when looking at smaller rooms but I never had any luck trying to wrap my head around how they work with larger open spaces.

I'm going to try what stgatilov has suggested and see how that goes.

[Springheel]

This might help, specifically the part about hide_distance.

 

 

[nbohr1more]

21 minutes ago, ddaazzaa said:

Yeah, I wasn't sure how to place visportals in a large open area. The concept of visportals is easier to understand when looking at smaller rooms but I never had any luck trying to wrap my head around how they work with larger open spaces.

I'm going to try what stgatilov has suggested and see how that goes.

As an addendum, with the extensive use of modular content you will likely need to embed caulk brushes inside some of the modular models so that the visportals can properly seal.

[Araneidae]

Is The Painter's Wife well optimised?  It murders my system, and buying a new graphics card is off the table until the semiconductor market recovers from its current doleful state (I want a reasonably high end AMD card ... good luck with that today!)  The Painter's Wife is really the only mission that I find unplayable, so it would be interesting to know where it fits in this conversation.

[Gadavre]

4 hours ago, ddaazzaa said:

I have tracked down a performance issue due to the map wide fog. It seems to increase draw calls by 40%. There is a no fog version in the link below which should improve performance quite a lot. Let me know what performance is like with this version:

https://drive.google.com/file/d/1iXE4-W0jucFSUrDQ415XsBehCLuX8gzd/view?usp=sharing

I'm currently seeing if I can reduce the performance impact of the fog.

 

Thank.  Removing the fog particles in the mission removed the low FPS in the castle grounds. blackgrove_no_fog.pk4 works better

a variable with this value improves visibility, but does not increase for me performance:

seta r_skipFogLights "1"

FPS drop is still there between the sea and the castle fence.

 

Edited April 22, 2021 by Gadavre

[duzenko]

4 hours ago, Araneidae said:

Is The Painter's Wife well optimised?  It murders my system, and buying a new graphics card is off the table until the semiconductor market recovers from its current doleful state (I want a reasonably high end AMD card ... good luck with that today!)  The Painter's Wife is really the only mission that I find unplayable, so it would be interesting to know where it fits in this conversation.

I'd say the original version is not bad, and the newer fixed version should be faster

In any case you already have a thread about that, so let's not spam other threads with that again

[Araneidae]

My apologies

[ddaazzaa]

On 4/22/2021 at 9:47 PM, stgatilov said:

Also I believe these bushes eat too much performance for their relatively unimportant role:

  Reveal hidden contents

When I stand the shore and rotate camera, I get severe FPS drop at the moment when my frustum starts including this thing.

Yeah, I didn't know those hedge models are over 5000 polys each. That's a major optimization I can make with hedge LODs.

23 hours ago, nbohr1more said:

As an addendum, with the extensive use of modular content you will likely need to embed caulk brushes inside some of the modular models so that the visportals can properly seal.

I don't think there's any problems with rooms in the manor sealing properly. Is there a reason caulk would seal any better than a brush with a standard texture?

I noticed Sir Taffsalot's Sword using the same modules and has caulk on all of the internal brushes.

[nbohr1more]

Any brush that is not a func_static should work.

Caulk is just an invisible brush and it will ( to some degree ) optimize away to the minimum needed surfaces for sealing as compared to other materials.

[stgatilov]

2 hours ago, ddaazzaa said:

Is there a reason caulk would seal any better than a brush with a standard texture?

Caulk surfaces are not rendered at all, normal surfaces are rendered, even through you don't see them.

No idea how big the performance difference is. I bet the actual rendering of invisible brush surfaces takes negligible time, but additional frontend processing and shadows might take some time. I'm not sure it is worth wasting your time replacing the brushes now. But it is probably a good idea to use caulk in future missions.

[Springheel]

2 hours ago, stgatilov said:

Caulk surfaces are not rendered at all, normal surfaces are rendered, even through you don't see them.

No idea how big the performance difference is. I bet the actual rendering of invisible brush surfaces takes negligible time, but additional frontend processing and shadows might take some time. I'm not sure it is worth wasting your time replacing the brushes now. But it is probably a good idea to use caulk in future missions.

We've run those tests before, and there was no performance benefit to using caulk.  

[JackFarmer]

@ddaazzaa

Cull away as many entities as possible with the "hide" property.

Look into HHVF and check the outside enviroment. Most of the small entities (including plants) have a hide property on them.

[ddaazzaa]

13 hours ago, JackFarmer said:

@ddaazzaa

Cull away as many entities as possible with the "hide" property.

Look into HHVF and check the outside enviroment. Most of the small entities (including plants) have a hide property on them.

Thanks, I'm in the middle of using hide at the moment and I'm seeing some very large performance improvements. I've found the best way to work out the hide distance is to temporarily insert a speaker right next to the entity and adjust the diameter to the preferred cull size, then divide the diameter in half and that becomes the hide_distance number.

[ddaazzaa]

@Gadavre

I have released a new version that should improve performance by a lot. Let me know what your performance is like now.

1.02:

https://drive.google.com/file/d/1wZc_nqHoX7kQvzfg08EpoRy2hyuH7mw3/view?usp=sharing

Edited April 25, 2021 by ddaazzaa

[Gadavre]

On 4/25/2021 at 7:35 AM, ddaazzaa said:

@Gadavre

I have released a new version that should improve performance by a lot. Let me know what your performance is like now.

1.02:

https://drive.google.com/file/d/1wZc_nqHoX7kQvzfg08EpoRy2hyuH7mw3/view?usp=sharing

@ddaazzaa Тhаnk! I've tested with the fog, the area between the sea and the castle fence is already fine....

But on the castle grounds  itself there are places with low fps

Edited April 26, 2021 by Gadavre

[Gadavre]

[duzenko]

@GadavreYou seem to have triple buffering on

Either turn it off or do your tests with VSync off

[Gadavre]

56 minutes ago, duzenko said:

@GadavreYou seem to have triple buffering on

Either turn it off or do your tests with VSync off

i dont use triple buffering. I don't use the video driver settings

I can't turn off VSync. since there will be Screen Tearing

[ddaazzaa]

6 hours ago, Gadavre said:

i dont use triple buffering. I don't use the video driver settings

I can't turn off VSync. since there will be Screen Tearing

Due to the nature of those outdoor areas it will be hard to improve performance much more without introducing noticeable amounts of pop in. I'm happy to release a different version with more pop in that should have better performance.

[stgatilov]

38 minutes ago, ddaazzaa said:

Due to the nature of those outdoor areas it will be hard to improve performance much more without introducing noticeable amounts of pop in.

I am quite against pop-in and LOD stuff in general.

Most of the screenshots include those bushes again, but at close distance now.
You added LOD to turn them into boxes from far distance, but it does not save players when they come close.
Why didn't you simply reduce the complexity of the model (without any LOD at all) ?

[ddaazzaa]

There's only 2 versions of those models (low and high poly). The low poly model isn't really suitable for up close, so I'd have to create my own. Is there an easy way of lowering poly count on models without knowledge of 3D modelling software?

[stgatilov]

42 minutes ago, ddaazzaa said:

There's only 2 versions of those models (low and high poly). The low poly model isn't really suitable for up close, so I'd have to create my own. Is there an easy way of lowering poly count on models without knowledge of 3D modelling software?

Did these models come from core TDM assets?

If yes, then asking someone from TDM team is viable approach 
It is not good having such heavy models in core assets, since mappers will have same problem with them again and again.
If some average-sized models is created, it can be bundled in this FM and added to core assets for 2.10.

 

[ddaazzaa]

They are from the core TDM assets. The hedge archway alone is 9600 polys. The rectangle hedges are 5200 each. Add a few of those in a row and an archway and the poly count starts to skyrocket.

models/darkmod/nature/hedge01_***.lwo

What's the best way of asking someone? Should I create a thread in the Art Assets forum requesting if someone can create lower poly models?