Should "tcg" command brutally drop fps? and physics question.

General discussion regarding the OpenMW project.
For technical support, please use the Support subforum.
Post Reply
User avatar
jirka642
Posts: 112
Joined: 23 Aug 2014, 11:39
Location: Czech Republic
Contact:

Should "tcg" command brutally drop fps? and physics question.

Post by jirka642 » 23 Aug 2017, 18:07

I was testing this old heavily modded setup, and was getting very low fps when I tried to run into urns and baskets in Balmora.
I wanted to know if there is some problem with them, so I enabled colision geometry with "tcg". Fps dropped drastically. :o
Should that happen? I would not expect that having few more lines to draw would drop fps so much. It's also using physics thread to draw them which is a bit wierd...

images: http://imgur.com/a/GWw8S

Also, wasn't the physics fps drop problem (when walking into things) fixed in last openmw version? Or is this completely caused by the creator of that urn model? (Too many polygons?)
On GitHub as kunesj, on GitLab as kunesj.
OS: Linux Mint 18, CPU: Intel Core i5-3330, RAM: 16GB, GPU: GeForce GTX 650Ti, OpenMW: from source

User avatar
DestinedToDie
Posts: 1056
Joined: 29 Jun 2015, 09:08

Re: Should "tcg" command brutally drop fps? and physics question.

Post by DestinedToDie » 23 Aug 2017, 19:18

The urn does have a rather excessive amount of vertices. It's more high-poly than any model I've seen in modern games.
Last edited by DestinedToDie on 23 Aug 2017, 19:30, edited 1 time in total.

User avatar
lysol
Posts: 701
Joined: 26 Mar 2013, 01:48
Location: Sweden

Re: Should "tcg" command brutally drop fps? and physics question.

Post by lysol » 23 Aug 2017, 19:23

jirka642 wrote:
23 Aug 2017, 18:07
Also, wasn't the physics fps drop problem (when walking into things) fixed in last openmw version? Or is this completely caused by the creator of that urn model? (Too many polygons?)
It wasn't "fixed", but it was rather reduced. It is a lot better now than before, but it is still a bit of a problem.

That said, this model is horrible really. Sorry to whoever made it, but there are triangles everywhere in no logical order. I'm not expert on 3d modeling really, but from my limited knowledge, this model is really poorly made. So yes, too many polygons.
Normal mapped texture replacers, exclusive for OpenMW:
My Nexus page

raven
Posts: 47
Joined: 26 May 2016, 09:54

Re: Should "tcg" command brutally drop fps? and physics question.

Post by raven » 23 Aug 2017, 20:09

Should that happen? I would not expect that having few more lines to draw would drop fps so much. It's also using physics thread to draw them which is a bit wierd...
It is meant to be used for debugging, so won't be optimized much.

Bullet pretty much makes a call for each line to be drawn, if I remember correctly. Then it depends on how well your graphics system handles that, batches lines into a buffer or something...

User avatar
jirka642
Posts: 112
Joined: 23 Aug 2014, 11:39
Location: Czech Republic
Contact:

Re: Should "tcg" command brutally drop fps? and physics question.

Post by jirka642 » 23 Aug 2017, 20:23

OK, thanks for the responses everyone.
I expected it would be mostly the models fault, but it's good to get a confirmation.
On GitHub as kunesj, on GitLab as kunesj.
OS: Linux Mint 18, CPU: Intel Core i5-3330, RAM: 16GB, GPU: GeForce GTX 650Ti, OpenMW: from source

User avatar
Pherim
Posts: 139
Joined: 27 Aug 2014, 15:37

Re: Should "tcg" command brutally drop fps? and physics question.

Post by Pherim » 24 Aug 2017, 11:05

DestinedToDie wrote:
23 Aug 2017, 19:18
The urn does have a rather excessive amount of vertices. It's more high-poly than any model I've seen in modern games.
Many "better meshes" mods for Morrowind are like this. They simply take the vanilla models and add a subdivision surface modifier in Blender to smoothen it. Of course, this creates a huge amount of polygons. However, the vanilla engine is actually very forgiving when it comes to high-poly meshes (and as I've pointed out before, polygons alone are not the biggest factor when it comes to performance with today's hardware. Modern GPUs are able to push a large number of triangles per object, so it mostly doesn't really matter if an object has 100, 1,000 or possibly even 10,000 triangles, as long as they all use the same material. that means texture, not NiMaterialProperty, in this case, as in a Nif file, several objects can share the same NiMaterialProperty. Modern GPUs can handle literally millions of polygons - if the CPU can keep up with the draw calls). The main problem in this case is, that the vanilla mesh doesn't have a separate collision mesh, so the new, hig-poly mesh is used for collision - and while once again the vanilla engine doesn't really have any problems with that, we all know OpenMW does. I've run into this exact problem myself, and what I did was take the vanilla mesh and use it for collision with the more detailed mesh - but that was a different, less excessive one.
lysol wrote:
23 Aug 2017, 19:23
That said, this model is horrible really. Sorry to whoever made it, but there are triangles everywhere in no logical order. I'm not expert on 3d modeling really, but from my limited knowledge, this model is really poorly made. So yes, too many polygons.
This is what happens if you apply Catmull-Clark subdivision to a triangulated mesh. This method is usually meant to be used with quads, and works much better if the object is converted into quads (if possible) before using the modifier. Actually, what happens is that the modifier takes every single triangle and divides it into two quads, which gives really weird results, and you end up with these somewhat irregular patterns after the mesh is triangulated again.

User avatar
lysol
Posts: 701
Joined: 26 Mar 2013, 01:48
Location: Sweden

Re: Should "tcg" command brutally drop fps? and physics question.

Post by lysol » 24 Aug 2017, 12:34

Pherim wrote:
24 Aug 2017, 11:05
This is what happens if you apply Catmull-Clark subdivision to a triangulated mesh. This method is usually meant to be used with quads, and works much better if the object is converted into quads (if possible) before using the modifier. Actually, what happens is that the modifier takes every single triangle and divides it into two quads, which gives really weird results, and you end up with these somewhat irregular patterns after the mesh is triangulated again.
Ah, yes that's right. Makes sense of course, I've always read heard that working in tris can mess up your models when adding modifiers. Thanks for the explanation.
Normal mapped texture replacers, exclusive for OpenMW:
My Nexus page

User avatar
Pherim
Posts: 139
Joined: 27 Aug 2014, 15:37

Re: Should "tcg" command brutally drop fps? and physics question.

Post by Pherim » 26 Aug 2017, 16:18

It's actually even three quads per triangle (as opposed to four quads per quad), which means that besides this weird pattern, you'll end up with six times the triangles for every iteration of the modifier. If converted to quads first (which is possible with many Morrowind meshes), it's only four times the triangles - and a much more regular mesh.

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests