ze_cathedral_css

Started by Joris Ceoen, August 01, 2013, 03:03:31 PM

Previous topic - Next topic

Joris Ceoen

ZE_CATHEDRAL (CS:S)
URL: http://gamebanana.com/projects/34608




This cathedral is being built for the Zombie Escape community (mainly for PlagueFest, but any other ZE server is completely free to use it!) and is entirely built by me. It uses Google Sketchup, 3DS Max and WWMT to their full potential for delivering the best possible results on the Source Engine (the CS:S and HL2 version).

This map has a precedor, currently available for download, but inefficient to host on a server: http://csgo.gamebanana.com/maps/168170
Why? Because it was created under severe pressure, in a timespan of 5 days and back in the time it were my very first models, where I didn't know everything about polycount and methods to increase performance.
I believe the entire cathedral when rendered exceeds a million polygons with all of the models together, making it combined with many players and dynamic elements a pain in the butt for computers to run. It runs at 45fps in average when fully rendered, which is a no-go.

In time with many other projects until now I have learned so many methods to optimize from the roots up that I was finally able to re-start a project that is very hard to realize, a fully standing cathedral where 64 players can run into with at least 100fps or more at dynamic moments, with the necessary aesthetics. This new cathedral has many different approaches over the first version:


  • The entire cathedral architecture consists of 1 texture. All of the models are UV-mapped with one 1024x1024 sized texture containing the 3 most important textures, being the cathedral walls, marble pillars and dark-stone-like concrete supports. The older one had 3 albedos/textures (512x512) one all of the models, making it more expensive. This may look like a ridicilous way of increasing performance, but for a map at this scale every little bit counts for something!
  • The polycount plain and simple is 3 times less than the initial models, being much better organized and looking much better, if not being incomparable. Certain other models are also differently designed over the old one, being more realistic towards real-existent cathedrals. This makes the difference between everything bein rendered at 170FPS instead of 45...
  • The frontwing has a double corridor-structure instead of the older single one.


  • The windows now are all 2 polygons while the older windows had around 50 polygons (counting 50 polygons x 60 times...). Also the windows frames are around 400 polygons while the older ones where 3500+ being incredibly expensive for little to no better results.
  • Brushwork is now integrated in a work-friendly environment instead of an impossible-to-fix result over the older one

I cannot predict when this project will be done, as I will also create a complete outside framework for the cathedral (something which I haven't done in the older version for obvious reasons) and also add many more gameplay elements than before. I did learn that taking all the time pays off instead of rushing anything with projects as fragile as these, so never rush anything, even if it takes years! Currently there is around one month of work in this cathedral (working arguably 3 weeks onto it, about 3-4 houres a day).

This wouldn't be possible without WWMT, I mean I need to align everything perfectly, and with the easy illuminating origin, texture export and model tweaking tools WWMT provides it makes everything so easy. Sometimes when I realise such a tool wouldn't exist, there would be no way for me to be standing where I am now. Many options such as collision models, sequences, illuminating origin, attachements, LODS, Custom Breakable Gibs and so much more were totally inexistent to me until I discovered them with WWMT.

I will post extra screenshots to share so far. Except for the first 4 screenshots (including the 2 above) I quickly took the effort to go back into the map and take updated screens of the work so far:









I'll keep this thread up-to-date when something important is to be shared!

wallworm

Excellent discussion on your project. Perhaps your next project will even do the world geometry in Max too :)

When you get time, please share with the community the exact steps you're taking to optimize various elements as it will be helpful for those people who don't understand how/why. For example, it is very common for a new user to import a 300k + model from some external source then try to add it into a level. Showing how and what you've done specifically might be helpful for some aspiring artists.

Joris Ceoen

Quote from: wallworm on August 01, 2013, 07:36:23 PM
Excellent discussion on your project. Perhaps your next project will even do the world geometry in Max too :)

When you get time, please share with the community the exact steps you're taking to optimize various elements as it will be helpful for those people who don't understand how/why. For example, it is very common for a new user to import a 300k + model from some external source then try to add it into a level. Showing how and what you've done specifically might be helpful for some aspiring artists.

Sure, would you like me to put it here or somewhere else on the forums? Lol a 300k model, I think Source limits around 50K for one model, and even then Valve themselves never used more than around 10K at max for a single model. One exception I know of is the very last helicopter in HL2EP2 which has 37k polygons, but it was an exception.

I really would indeed, but not this evening, since it's very late and I'm almost going to sleep. I'll see tomorrow, when I certainly will have the time. I will probably more explain why a model should be low-poly for many reasons and why going expensive is bad (because nowadays many people tend to think polycount isn't important, and yes RAM memory is also important but doesn't make polycount irrelevant) and why LODS aren't always a solution.

See you tomorrow!

Joris Ceoen

This evening I'm going to write about optimizing in modeling, however beforehand I would like to mention that I'm not a professional modeler. I mean it's not my work/job, but it certainly is my hobby, and experience with multiple engines learned me the way of optimizing everything. There are many obvious, but also unobvious ways of optimizing models and I will try to adress the most important ones in what I'm going to write very soon.

I'm writing it now in an external program to later post it here, as I think it will take some time before it's finished.

Neil


SMF spam blocked by CleanTalk