847 réponses à ce sujet

[Tchos]

Indeed, this looks very nice.  I would ask however that you use different random seeds for the rows of bushes that are right next to each other, so there isn't an obvious repeating pattern.

[kamal_]

Looks good, there are a few sections where the same tree seed is used for bushes like tchos said, but as this is the government district I suppose it could be  thought of as topiary since it's a rich district.

[drechner]

I actually meant to scale the bushes randomly a bit, but I totally forgot before I took the screenshots... I knew I was forgetting something

 

Actually, nearly all the trees still need to be scaled still. I'm avoiding using random SpeedTree seeds because it has a pretty large impact on performance, and the area is already pushing the limits.

[Tchos]

Having just one variant version of plants you're putting in rows can't impact your performance in a significant manner.  How many types of trees, bushes, and grass are you using?

[drechner]

Having just one variant version of plants you're putting in rows can't impact your performance in a significant manner.  How many types of trees, bushes, and grass are you using?

One variation is probably okay, but the unique "trees" count is well over the recommended 5 by Obsidian already.

[kamal_]

I routinely use 25 seeds. Five seeds was for performance reasons at that time.

[kevL]

i have computer 'from that time'. trees are just not the problem. shadows are

so I don't know where they got "5" from .... go with the flow, drech.

[rjshae]

Well I never knew that about the type/seeds performance limitation; I thought it was just a limit on the types. Good to know, although I've never seen a performance hit from using a large number of seeds.

 

As for shadows, the game engine just seems to turn them off if they start getting too burdensome.

[kevL]

As for shadows, the game engine just seems to turn them off if they start getting too burdensome.

i play low-tech: drop shadow only, no bumpmaps, no reflections, etc; sometimes I'll turn it all on for a screenshot, but that risks CTD in heavy traffic. Sight distance stays maxx'd, but if turning any option on gives even slightly noticable lag it stays off.

on the good side, colors look great

[drechner]

i have computer 'from that time'. trees are just not the problem. shadows are

so I don't know where they got "5" from .... go with the flow, drech.

 

Sure, shadows are expensive, but that doesn't mean that additional trees aren't. This isn't an "either/or" situation here

[kevL]

well, shadows are ~1000[0]x more expensive.

 

in the last area I made, every tree was deliberately given a different seed - nearly 50 of them. It wasn't a problem.

[Tchos]

Placeable objects without LOD models are another very expensive thing, as users of the City Hak building placeables discovered.  Much more expensive, I daresay, than SpeedTrees.

 

BTW, Drew, this may mean nothing to you, but I consider myself extremely conservative in terms of whether I use things I consider inefficient in this game.  I've even been excessively careful with heartbeats and AI, and making sure I clean up things that would otherwise continue running after their purpose is over.  From this perspective I say use the seeds.

[rjshae]

A slightly related question I've wondered about is whether tilesets support LoD? In particular, Tupoun's sewer tileset causes just a monster performance hit for me. I suspect it would be better with LoD models.

[kevL]

- just like to point out that in some cases the right thing to do might well be use the same seed, like for that row of bushes in front of the statue.

 

/out

 

( sry, T. an opinion i guess)

[Tchos]

kevL: I agree there's no reason to use extra seeds when not necessary, like when trees aren't right next to each other, but I've never seen actual topiary that had repeating patterns like bad seamless tilework.  The sculpting in all topiary I've seen had order, but the leaves had no detectable pattern.  Yes, you could make up a story around it for why the landscapers would have wanted it that way, but why would you do that?  Do you think it looks good, or is a beneficial effect?  Where are you coming from, here?

[drechner]

I've already scaled all the trees and added a seed variation to the bushes; good idea!

[kevL]

OT,

Where are you coming from, here?

it looks like a 'lawful' area, orderly & structured ... if you find the right seed, if it's a hedgerow of some sort - then next to no player is going to look closely enough to notice whether the leaves match, and the subtle unconscious effect might be to reinforce a 'lawful' orderly atmosphere: a well-trimmed hedge.

but i haven't toyed around with hedges

oh, and you ask why might 'they do that'? well, too much order in a lawful area might be there to conceal 'their' dark side .......

[Tchos]

Okay, I agree that if someone found the right seed such that the tiling isn't obvious and it required looking closely at them to notice a pattern, then yes, that would be a good psychological use for such an effect. 

[kevL]

afterthought: the effect could perhaps be made less blatantly repetitive by overlapping bushes at different deltas (distances from the center of one bush to the center of the next bush). anyway, enough OT fer me.

[Dann-J]

If players are noticing details like tree seeds, then the gameplay must be boring indeed.

 

I find I have to constantly remind myself that the 'problems' I'm spending so much time fixing in my own module are almost certainly minor details that few (if any) players will ever actually notice. More often than not getting bogged down in such minor 'window dressing' issues is really just an excuse to procrastinate the more important issues, like starting that brand new area from scratch, or trying to untangle a complex multi-path quest so that it actually works. Rearranging the window dressing is much less taxing on the brain (or the hairline).

[Eguintir Eligard]

From what I have learned from back before all you were here, the every seed added about a second of load up time for an area. It has nothing to do with the framerate, nor is there any logical software engineering reason why it would... the tree is already generated once in game. It's not being re randomized every frame.

 

Perhaps that will help focus the discussion because a lot of what was being said here is immaterial.

[Eguintir Eligard]

 

I find I have to constantly remind myself that the 'problems' I'm spending so much time fixing in my own module are almost certainly minor details that few (if any) players will ever actually notice. More often than not getting bogged down in such minor 'window dressing' issues is really just an excuse to procrastinate the more important issues, like starting that brand new area from scratch, or trying to untangle a complex multi-path quest so that it actually works. Rearranging the window dressing is much less taxing on the brain (or the hairline).

 

I've been reminding a certain modding buddy about this for four years. Hasn't been heeded, so I guess he's bald now.

[Tchos]

Perhaps that will help focus the discussion because a lot of what was being said here is immaterial.

 

I'm pretty sure that discussion has run its course, but that additional information could be useful in future discussions elsewhere.

[drechner]

From what I have learned from back before all you were here, the every seed added about a second of load up time for an area. It has nothing to do with the framerate, nor is there any logical software engineering reason why it would... the tree is already generated once in game. It's not being re randomized every frame.

 

Perhaps that will help focus the discussion because a lot of what was being said here is immaterial.

 

Whenever you introduce a new mesh, you increase mem usage and the number of draw calls, the former potentially causing perf issues (though unlikely in the case of NWN2) and the latter being the main cause of perf issues. I don't want to sound like a contrarian here, but to state otherwise is just flat-out wrong.

 

NWN2's biggest issue with GPU-related perf is the culling (or lack thereof) of static meshes. Therefore, introducing more meshes into a level will decrease performance, generally not because of the impact on mem usage but because of the increase to draw calls.

[Dann-J]

Whenever you introduce a new mesh, you increase mem usage and the number of draw calls, the former potentially causing perf issues (though unlikely in the case of NWN2) and the latter being the main cause of perf issues. I don't want to sound like a contrarian here, but to state otherwise is just flat-out wrong.

 

NWN2's biggest issue with GPU-related perf is the culling (or lack thereof) of static meshes. Therefore, introducing more meshes into a level will decrease performance, generally not because of the impact on mem usage but because of the increase to draw calls.

 

True - but they're called 'speedtrees' for a reason. The format has been designed to be as low on resources to render as possible. They don't have actual model meshes of their own, but rather are generated on-the-fly by algorithms.

 

The biggest cause of lag in adding speedtrees to an area is not in rendering the actual trees, but rather the shadows they cast. I always set small shrubs that are low to the ground to not cast shadows at all. They'll often end up in the shadow of a larger tree or some sort of placeable, so they rarely need shadows of their own. In a forest area with ten times more undergrowth plants than taller trees, that speeds up shadow rendering considerably.