Serious Engine Needs higher world polygon limits

[Mailman]

I've been trying for months now to optimize my maps to get higher than 50fps on my PIII-800 system with GF2 GTS 64MB. In low detail areas, I can easily pull >100fps, but if you add any geometric detail in Serious Engine, the framerate chokes. I've tried everything. Good sectoring, detail brushes, mip brushes (hardly useful in tight indoor maps). Nothing works. This engine simply needs higher capacity to render world geometry. It can handle 80,000 model triangles with relative ease, but when you have world geometry over 200-300 polygons (not triangles), it chokes on anything less than a gigahertz machine. I'm trying to make a mod with the widest target audience possible and not everyone has a top of the line machine. High framerates are essential for fast action play, and 60fps on a PIII-800 should not be that hard to achieve.

If I'm missing something, I'd like to know.

[Naythn]

Nope. Everything you said is correct, although the tricount is much more valuable, because polys can be composed of 1+ tris.

The best method is to use models whenever possible.

[sh0dan]

Yes, the Serious Engine polycount is it's biggest graphical problem for mapping.

At the moment you can however still do a lot of thing using models, pretenders, brush-mipping and visibility tweaks, but tricounts detail polygon tricount >1500 or 500 non-detail is a big problem for <1Ghz machines. Every time you get these polycounts, your world traversal time quickly exceeds 10ms - effectively limiting your framerate.

This is the price of dynamic visibility calculations and multiple texturepasses. The reason is that it is very difficult to optimize, since all non-detail polygons can be removed on a per-triangle basis.

[Mailman]

What if the editor had an option for saving maps in BSP compiled format. You'd have all the benefits of fast editing without compiling, and when your ready to make a release quality map, you could compile it for maximum visibility precalculation. Of course, this would be optional as it would be detrimental for large sized maps, but could dramatically speed up indoor maps with lots of detail.

Naturally, this would require a significant engine rewrite, but hey, its an idea :)

[Jedi Outcast]

i wish sectoring large outdoor areas were easier. but it just seems worthless to do so.

[Michael Harris]

Hopefully there are some rules of thumb and optimizations that can help with accomplishing better frame rates purely through working in the editor........ but......

Also looks like a justifiction for getting into modeling ;-)
All models are not active/animated/enemies.....


The Editor can export selected polygons as .RAW 3D files (a common widely supported sort of lowest common denominator triangles file format)...... which could presumably be imported into a modeler and made into something that the Serious Modeler (or the forthcoming SKA Studio) could convert into a Serious model..... which could subsequently be inserted into the world in the editor......

Perhaps there is a specialized modeling procedure that needs to be figured out for routinely creating and inserting large numbers of Serious models into worlds..... Assuming that little or no coding, scripting, etc. is needed for dumb/static architectural models, there ought to be a way to standardize this process and turn it into an assembly line or rubber stamp type process.

One thing to figure out is how to do this in a routine straightforward high volume way.

Another thing to figure out is a set of rules of thumb for how to use/not use models as frequent and major architectural (and landscape) features in worlds. Is it practical to use models for buildings that have interiors, within which player, enemies and weapons will be? If not, is it practical to bust such up into pieces, so that such playing inside of (apparent single piece) models is practical?

Being able to do everything within the editor, without having to use modeling, is understandably desireable..... and a good way to start....... but ultimately, the rule is "hey, whatever works best in terms of the final appearance and playing experience"

[sh0dan]

Originally posted by Michael Harris
The Editor can export selected polygons as .RAW 3D files (a common widely supported sort of lowest common denominator triangles file format)...... which could presumably be imported into a modeler and made into something that the Serious Modeler (or the forthcoming SKA Studio) could convert into a Serious model..... which could subsequently be inserted into the world in the editor......


Unfortunately it's not that easy. You still need to UV-map your model in Milkshape/MAX/Lightwave/Whatever, and you are limited to one texture pass per polygon, and one texture per model. You also loose your per polygon collision, and you don't get any shadowmaps.
There are still a lot of places where models can be used, like the top of the Babylon tower, arches, pillars.


Originally posted by Michael Harris
Perhaps there is a specialized modeling procedure that needs to be figured out for routinely creating and inserting large numbers of Serious models into worlds..... Assuming that little or no coding, scripting, etc. is needed for dumb/static architectural models, there ought to be a way to standardize this process and turn it into an assembly line or rubber stamp type process.


That's why there's a distinction between 'detail' and non-detail brushes. Detail brushes are a lot faster, and gives you everything except visibility. Perhaps a 'superdetail' that excludes more features could be possible, but I don't know the engine well enough to know if it would be possible to exclude more features, so 'super-detail' polygons doesn't affect 'world traversal' time.

Perhaps we should bug Dean or Alen a bit ;)

Originally posted by Michael Harris
Another thing to figure out is a set of rules of thumb for how to use/not use models as frequent and major architectural (and landscape) features in worlds. Is it practical to use models for buildings that have interiors, within which player, enemies and weapons will be? If not, is it practical to bust such up into pieces, so that such playing inside of (apparent single piece) models is practical?

You don't get any collision, and you get very bad lighting, so for buildings it is not an option, unless they are only visible in the background.