'light' models

[esteel]

People that read the change logs for Nexuiz releases might know that my beloved carni model has a little problem. Its standing 'animation' is slightly broken. The model is not total centered but is standing a bit behind where it should be. So when you aim from the left or right at a standing carni you might miss him. Its not a real problem as you will have a hard time to find standing players in Nexuiz but it made me reconsider my playermodel..
I would not deserve the title 'forumaddon' if i would just choose a new model. No no no no!
Well i compared all the current models in Nexuiz 2.2.2. As i have little clue about moddeling i just loaded all the models and used r_speeds 1 to see how much stuff the engines needs to draw in thirdperson view which also includes the map and all stuff so i can't give you absolute numbers and i'm not totaly sure if this is representive but here is what i found: The top six from 'lightest' to 'more stuff to draw':

Shocktrooper (my new one )
Headhunter
Pyria
Insurectionist
Frika
Skadi

Might also be interesting for the planned pro editon. I did make it top six as i think the headhunter does not look that nice so its still top five if that one gets dropped
Also one might use that info to think about making brightskins for the shocktrooper instead of the nexus model.

[GreEn`mArine]

one should rather do a FRESH brightskin with less polygons than Shocktrooper. People like big models, and I consider the size of Nexuz exactly right, "big"

[ai]

What is a Brightskin? Is it a model? I always thought it was some kind of texture slapped onto a model.

[esteel]

Yes the brightskin is just a texture slapped onto the model. But if you vote that feature in nexuiz it uses the nexus model as green only made a brightskin for that one.
Well Green is right bulky models are better as they are easier to see.. i compared headhunter and nexus and there is not much difference both are really big and bulky.. Also it would be easy to make a brightskin for the headhunter as almost all the model is used for shirt/pants colors already. Actually i'm beginning to really like the headhunter so maybe i'll use that model from now on. Would also fit my style of not hiding with small models

[aimless]

Why not simply throw out the broken models or those which are too complex? Perhaps that is the easiest way to improve performance for some people?

Nexus is the biggest model, but isn't it also slightly bigger than the hitbox? At least I heard about that. I can't verify that, since I don't know the size of the models or the hitbox. Is there a way to show the hitbox, would be interesting to see?

Did you use chasecam for your test, or did you log in with another client? Or does the side you view on a model not matter for r_speeds?

Thanks for the list anyway. Good idea.

[esteel]

Why not simply throw out the broken models or those which are too complex? Perhaps that is the easiest way to improve performance for some people?

Nexus is the biggest model, but isn't it also slightly bigger than the hitbox? At least I heard about that. I can't verify that, since I don't know the size of the models or the hitbox. Is there a way to show the hitbox, would be interesting to see?

Did you use chasecam for your test, or did you log in with another client? Or does the side you view on a model not matter for r_speeds?

Well the main problem with the models is that they are about the oldest things in Nexuiz. One should better create new models with modern techniques which will look way better but use less resources then the models we have now. It should be possible to use the current models as a guide or create groups like robots, monsters, humans which could also serve to differentiate the teams in team matches like quake4 does. Just throwing out old models is not really a good solution. And i have not yet tested if it would indeed be a bit speed boost.

I do not know of a way to show the hitboxes.

I did use chasecam yes but i also made the model transparent so the engine had to render it completely.

[003]

The poly counts in the models need to be totally optimized before you go working on other less important aspects of them.

[The mysterious Mr. 4m]

esteel using a different model than carni??? what hathest yon world succumb to?

[esteel]

The poly counts in the models need to be totally optimized before you go working on other less important aspects of them.

Hehe i was almost tempted to respond with 'thanks wise kid'. But honestly.. yeah we know the player models have some damn problems but people following the forums and stuff for longer might have noticed that its not that easy to find modellers willing to do GPL stuff for some reason. Because of that it should be worth to improve the current situation but also continue to try to get new models..

esteel using a different model than carni??? what hathest yon world succumb to?

Yeah i know what you mean.. Well at high quality settings the carni really looked very cool but i did want to change it because of that bug. Well mulder looks also very cool but i wanted to be nice to other peoples FPS I'd really like to see new models.

[esteel]

Though we should replace the models i tried some stuff with the models.. timedemo for demo1:

with unmodified stuff: fps min/avg/max: 37.0370370/96.5360303/500.0000000

now i replaced all models

with the robot model which seems to be the worst of all: fps min/avg/max: 35.7142857/84.7766231/500.0000000

with the headhunter model which should be second 'lightest' but is the biggest and most visible of those light ones: fps min/avg/max: 41.6666667/106.3509749/500.0000000

and with the shock trooper which is the most light model: fps min/avg/max: 43.4615385/108.4597466/500.0000000

Maybe i should do a test with the nexus model but its getting late As one can see there is a 24fps difference on average from using worst vs best model and around 10fps + / - compared to the mixed models that demo1 normally has.
So it seems we will get quite some improvements by using optimized models and this should convince people even more that we need to try to find modelers and get new and optimized models.

Though i'm a bit worried about that reported lowest fps. One even sees that slowdown when watching the timedemo so maybe there is something else to optimize either the engine or the map. Other then that i think the numbers are ok for a map cramped with models