Can Custom Models Cause FPS Drops?

[Gumberbunbder]

This may be a dumb question but can custom models cause FPS drops? Ive been noticing a map I've been working on has terrible FPS. I tried everything to find out the cause. Eventually I deleted all triggers and still the lag persisted. I started deleting units from the playable map area and narrowed it down to one unit that seemed to be the source, a cannon. I tried changing that units model file and that caused FPS to be normal.
Its weird because I never noticed a major FPS drop until recently even though this cannon model has been a part of the map for a long time.
Does anyone know what kind of an impact models can have on FPS? Does poly count play a role? Any other things about a model that could cause this?

Spoiler: Laggy Cannon

EDIT: Here is the link to the FPS dropping cannon
Old Cannon

I replaced the model with another custom cannon model, here is the link:
Cannon

Even though the new model has far more polygons it performs far better. So is there something wrong with the "old cannon" model?

 

[FUJI]

Of course, custom models can cause lag, if created improperly. Particles can cause major impact on performance if created so many. Also the lights. Are you sure it's model related? Can you specify more details? And post here the model file so we can test it out?

 

[Gumberbunbder]

"CannonVer2" is the seemingly less laggy model, though i keep testing it and it too lags though not as much as the other cannon model.

"OldCannon2" is the laggy model.

I'm not sure about all of this though, I'll keep on testing for now.
It seems certain though, that the original wc3 mortar team model causes zero lag.

EDIT: Im going to try the models in a new empty map and see how they perform.

It's Strange because sometimes when i run the map with the "CannonVer2" i get 55-60 fps the whole time, then other times when i test the map it goes to 30 fps for like 5 or so seconds intermittently. Im going to test like 10 times in a row with mortar team model and record fps

EDIT2: Here are my results so far

Spoiler: Fps Tests

FPS Test, Map: American Colonization

*First number the lowest recorded fps, doesnt last long, 1-2 seconds at most

CannonVer2 (Fantasy Cannon)

-test 1: 56-60 fps
-test 2: 55-60 fps, one dip to 30 fps for a few seconds
-test 3: 55-60 fps
-test 4: 51-60 fps
-test 5: 54-60 fps, one dip to 30 fps for a few seconds
-test 6: 55-60 fps, one dip to 52 for a few seconds
-test 7: 55-60 fps, one dip to 30 fps for a few seconds

CannonVer2 | Deleted Triggers

-test 1: 58-60 fps
-test 2: 58-60 fps
-test 3: 57-60 fps, one dip to 0 fps for 1-2 seconds
-test 4: 54-60 fps
-test 5: 58-60 fps
-test 6: 56-60 fps, one dip to 30 fps for a few seconds
-test 7: 58-60 fps, one dip to 30 fps for a few seconds
-test 8: 58-60 fps

Mortar Team Model | Deleted Triggers

-test 1: 59-60 fps
-test 2: 57-60 fps
-test 3: 57-60 fps
-test 4: 57-60 fps, one rapid dip to low fps, less than a second duration, Alt-tabbed out and then Alt-tabbed back in, now stuck at 45 fps but feels more like 10 fps, very choppy.
-test 5: Restarted previous game, 45 fps persists, feels like 10 or less fps.
-test 6: Closed wc3 and restarted, now 57-60 fps, tried Alt-tabbing again but didnt cause 45 fps again

OldCannon (Model I thought was causing the lag) | Deleted Triggers

-test 1: 30-60 fps, continual drops to 30 fps every 5 or so seconds
-test 2: same as test 1
-test 3: same as test 2
-test 4: same as test 3, even if the camera is far away from the cannons, still happens

OldCannon is consistently the worse performing

I put the map down below, removed all triggers except camera height. Test it and tell me if it performs terribly for you. I'm using Fraps to measure FPS

 

[Strydhaizer]

I see no problem with the models, it has low particles, low quality textures, low poly model, and only have few animation frames in both models, it might be something else that's causing lag in your game. (It only has 30 animation frames and 20 particle counts, shouldn't be an issue even in the classic).

Also are you playing an old version of the game or Reforged?

 

[Gumberbunbder]

I have reforged but am playing with classic graphics. I didn't buy reforged, i just got the update for it back when it came out. The models seems to perform fine in a new map, just terribly in my map, even though i deleted all triggers. I attached the map above if you could test it and tell me if it performs poorly for you too. Theres a bunch of cannons just north of your starting position.

You can try changing the "cannon team" model under Human custom units to something else once youve tested it so you can compare.

EDIT:
I deleted every trigger, deleted all units on map except cannon and some footmen and buildings. FPS drop still occuring.
Deleted all custom model files from map except cannon. So far no FPS drop. We shall see.
I'm truly confused. Sometimes it drops to 30fps, and sometimes it doesn't. But Mortar team model seems to never drop to 30 fps...

 

[Strydhaizer]

Perhaps the problem is the game itself, optimization problem maybe? Have you tried other maps and experience the same issue?

 

[Dr Super Good]

Someone should analyse how the two models are functionally different. The old cannon model might be doing something computationally intensive or slow.

 

[Gumberbunbder]

Perhaps the problem is the game itself, optimization problem maybe? Have you tried other maps and experience the same issue?

When you say "the game itself" are you talking about my map or about WC3 in general? Like reforged could have caused something? I deleted all the triggers in my map so I know it can't be poorly optimized triggers. Normal melee maps perform fine, most custom maps as well. I'd have to do some more testing. I'll try adding the cannons to another map of the same size and zoom that camera out to the same height (2500) to test.

Its ok if i can't figure this out, both models seem to cause the FPS drop, just the "oldcannon" model does it a lot more. I guess there could just be something magically broken about the map, something internally that I can't do anything about, lol

Someone should analyse how the two models are functionally different. The old cannon model might be doing something computationally intensive or slow.

I don't really know if it is the model. Could maps just have gotten damaged, like altered somehow to have performance issues that have nothing to do with any triggers or any content in the map itself? It seems to be the model causing the issues, except the model works fine in the new map I made, although the new map was just a small test map with nothing else in it... not sure

 

[FUJI]

Yes, some of the patches that are already implemented break something as usual.
You can easily notice that by playing the older versions which are not causing fps breakdowns or what.
However, I just consider those normal on my part since it's just revising a 15 year old game, it may not handle some changes.
If leaving the model alone in an empty map doesn't cause fps drops;
Test 1: Try to put 30 of this type inside it.
Test 2: Try to put 30 of mortars inside it.
Test 3: Try to put 30 of the high poly models you have.
And compare. Reply the results here.
(Make sure the map you're testing on is completely empty)

 

[Dr Super Good]

Could maps just have gotten damaged, like altered somehow to have performance issues that have nothing to do with any triggers or any content in the map itself?

No they could not.

It seems to be the model causing the issues, except the model works fine in the new map I made, although the new map was just a small test map with nothing else in it... not sure

New map is smaller, without many other models.

Hence why I suggested it be looked into. It might be doing something that is complex for the game engine. It works at a small scale but when more types of models are in play causes issues.

For example if it required a lot of draw calls. If a scene only has a few dozen this might not make a difference, but when there are a lot of different models so a lot of draw calls happen already the extra might push into poor performance.

 

[SebioL]

Have you finally come to any conclusions... or are there in fact missing fixes with Reforged in the last few patches?
It is not impossible that there is a possibility of some failure with the model itself, even if it is Low Quality. But in this case, precisely because OldCannon is an SD with few vertices, few anims, with little texture and low frame rates, everything seems strange to go wrong.

But within a general context, the units can have different types of impact, from the massive amount used on the same screen (which necessarily relates to the hardware, depending on the demand on HD quality or graphics effects you impose) or issues of mechanical limitations of Wc3 itself, it is not very easy to identify exactly what causes drops in FPS.
Personally, I don't think there can be any big problems even if you zoom in and see a fight with thousands of units clashing in Wc3. As long as you have good hardware and are aware of the game engine limits required by Wc3, because knowing how to deal with it, using "juggles" that deceive the limitations of Wc3, you won't have (or shouldn't) have any kind of problem.

WC3 MECHANICS LIMITATIONS: I will enumerate some main ones related to units, examples of collision pathing that limits a number of units for each Player (this does not necessarily cause FPS drops, but only limitation on movement for significant amounts of units), amounts of units used in relation to amounts of "Playable Players" on the map, many are unaware or little refer this limitation, but it exists, if you really want to have a few thousand units without having any evidence of Lag during big combats, insert more units for Players "unplayable" and force them to be allies and to move whenever necessary to the positions of their "playable" units. Unit edits in OE, using large amount of units to combat at the same time, having high attack speed and still being area of effect to reach large dimensions, if there are many units within this area of influence, it may cause some delay , although not as abundant. Another big or main villain is Triggers (such as Spawns of large amounts of units having the same point as source), but if you tested each one separately or removed all triggers, it can't be.

GRAPHIC/MODELS LIMITATIONS (SD, HD): I hope it's not a problem with the latest Reforged patches. I'm sure in this case it could be some modeling problem or even some corrupted track (3D experts can find this), you use SD and shouldn't have problems even with large amount of units, to be more specific, not even highpolis I believe to be a problem for the performance of the game, but the abundant amounts of frames per second in each animation, I think this is the great perpetrator of the FPS drop problems for most maps that use many custom models in this condition and that regardless of whether they are in SD or HD format, but even if you use many models on the same screen with a high amount of frames, this can be solved with a high-end computer... Note: i never noticed any interference with the game's FPS flow, even when using an abundant number of highpolis units, please if anyone disagrees, prove me know otherwise.
From the experience I had with HD use, I strongly believe that models with a high amount of polygons do not cause lags, they can cause visual discomfort or graphical failures if your hardware is very bad, this story of problems with highpolis really seems to be a myth, they can be a problem in using a large amount on the same screen for the Warcraft Classic engine, I'm not sure about the unit limit, but unlike that with Warcraft Reforged, it supports as many high polygon units as you want, as long as that you have a good Video Card. Using high resolution textures, I'm not sure, but it might be a problem for lower hardware, but if they're within the classic standard and less than 512 pixels, it shouldn't give you any problems. Now, high textures and abusing the graphic effects for lighting and shadows of the new engine, be ready to have a good computer.

 

[Dr Super Good]

Using high resolution textures, I'm not sure, but it might be a problem for lower hardware, but if they're within the classic standard and less than 512 pixels, it shouldn't give you any problems.

As long as the GPU has free video memory then texture resolution adds as good as no overhead. Cost to sample from a 512x512 texture is likely the same as from a 4096x4096 texture of the same format. The exception being from a cache and memory bandwidth perspective where there might be performance gains from smaller textures purely due to more cache hits, assuming the way the texture is used allows such gains to show.