Gingerbread Man on 30/11/2004 at 20:03
Following up on oft-repeated cries that with the editor in-hand we will finally be able to make huge, sprawling, Unreal-style levels for T3, I am engaged in experiment.
A simple, single, one-texture / one mesh subtracted brush space measuring 2048 x 16384 x 16384 (a very reasonable size in UEd for an outdoor area when I have gone as high as a 49159 cube with terrain and all sort of goodies) appears to have taken roughly 20 minutes to build. NavMesh process ended up with around 16383 Adjacent Links, and the Spatial Database started wheezing horribly.
Obviously a 16384 x 16384 space is hardly something you'd call a Level, and with judicious zoning etc it might be possible to create long chains of areas in a more traditional building paradigm, but I'm wondering if anyone can ballpark a reasonable maximum map size in whatever relevant parameters (mesh count, lights, pathnodes, etc) before I resort to building / meshing / populating a gigantic network of streets and buildings in the next experiment. I'm also fully aware of the idea that a huge, open space might be the sort of thing to grind the engine, and that a carved network of "rooms and halls" with a far greater spatial area -- and with zoning -- might not make the engine flinch in the slightest.
Although, unless people are interested in a completely unpopulated set-piece without very many links and emitters etc (and virtually no AI), there's still the problems associated with populating a large level to deal with.
Basically, my question boils down to this:
How much larger than the shipped levels could an FM realistically be? Are the relatively small levels a function of console limitations and a desire to create one product that would run well on both platforms, or are they simply the most that could be squeezed out of the engine without losing efficiency?
<small>And are there no Anti-Portals in T3Ed, or am I just a retard and haven't found them?</small>
Harwin on 1/12/2004 at 18:22
If you try to make a large, populated level that's too much bigger than the shipping ones you're likely to run into the property max. Although a significant chunk of that is for archetypes, so if you don't increase the *variety* of objects, just the number, you can probably get some growth.
Try pasting the geometry from one city section map into another if you want a quick and dirty experiment.
I'll check on anti-portals.
Edit: Nope.
David on 24/2/2005 at 00:00
Bump!
godismygoldfish on 24/2/2005 at 00:11
Dave just sticky the damn things ;)
Gingerbread Man on 24/2/2005 at 00:13
They're so old, they need to be refreshed anyway. :D
Macsen on 24/2/2005 at 00:16
Quote:
If you try to make a large, populated level that's too much bigger than the shipping ones you're likely to run into the property max.
Any ideas about what seem to take up the most space? Just the physical size of the level, or the amount of objects contained therein?
doctormidnight on 24/2/2005 at 04:20
I have no experience with editing (unless you count papers about the Soviet prison system, then I'm the bomb), but what about vertical building? Is the limitation based on land area, space in 3 dimensions occupied, or something else?
Gingerbread Man on 24/2/2005 at 04:22
The important limitations are intangible. Light maps, scene complexity, pathfinding databases, and object property maximums.
Building vertically, horizontally, even in spirals won't save you from those ;)
doctormidnight on 24/2/2005 at 04:28
Quote Posted by Gingerbread Man
The important limitations are intangible. Light maps, scene complexity, pathfinding databases, and object property maximums.
Building vertically, horizontally, even in spirals won't save you from those ;)
If we made this large wooden badger...
Renzatic on 24/2/2005 at 04:31
I've noticed that with most engines that use realtime lighting that lights per surface is usually a big performance factor...I'm assuming this is the same with T3, right?
