[Proclone]

Seeing as my computer is barely able to run generals properly (ok that's an exageration, but if does like to hang after a while, a hang-over from the w3d engine and it's dislike for my system configuration) I did some testing for Blitz 2 on the reaction the game had to different methods of constructing units and their material application.

I forgot about this "research" until I starting trying to think of something to contribute. I never really wrote any of it down though and most of it isn't 100% proven but I thought I'd post some of the information here in hopes others might eventually help pin down the exact reaction the game has to different things.

For example, right now there seems to be a big uproar/concern in the n00b modelers and their obsession with polycount.

Honestly I've found the polycount to be the LEAST important element in a model for Generals. The ironic element about all teh special effects in generals is that to run tehm you need a pretty powerful computer.. so a lot fo people are running nice machines for generals (heck ti won't run on anything too crappy). Because of this, for people who can run Generals their computer HAVE NO ISSUE with polygons. Heck I think I rmemebr reading someone got a 50k poly unit into the game and it ran alright. I remmeber the Renegade devs testing 100k+ models in Renegade. The engine laughs at polygones it seems.

The question then becomes, what's slowing the game down? It's obvious SOMETHING is doing it. The first thing that came to mind was the particle effects. While they DO slow the system down (they're hardware rendered) for the most part they're a seperate element from the models inserted by modders. What I mean is that they're a seperate case.

I found the best way to test for system impact was to use the w3d viewer. It comes equiped with a system clock monitor which can be used to benchmark a model's efficentcy.

From beta testing E&B I knew that the majority of lag generated within w3d games is caused by the MATERIALS, to be specific any special effect playing on a material. Alphas are DEADLY to the refresh speeds and this is double true in generals. That's not to say they should be avoided though. I tested them further and found that you can crash generals pretty easily via un-careful use of alpha. Basically if you enable alpha blending over an entire model you increase it's system load by about 30x. What was odd was that if you break off sections of a model and apply the alpha blending mode to them the system load hardly takes a hit.

What I determined was that ADDITIVE alpha blends slow the shit out of the game. Weather theirs actual transparency or not is irrelevant. a SOLID model with the OPTION of being transparent will cause the same load as one that's 50% invisible.

While I haven't made any extensive tests on animated skins, I found that while viewing some of teh boss model from E&B one of them that looks like a glowing jellyfish with lightning unning through it caused a significant memory hit. It was heavily transparent though so I can't be certain about the impact it's animation had.

All in al theres lots of other interactions to test, and I'd love to find out what others find... all I know right now if that Blitz 2 uses more polygones than generals on average and yet has a smaller memory footprint (and doesn't hang my computer).

[Curator]

That is interesting and brings up some good points. Speaking of ADDITIVE, guess what 80% of all particle effects use as their shader. You can test out this theory by adding the command line '-nofx' to your generals shortcut. This disables all particles regardless of the settings in the game options. You should see a significant increase in performance.

Another major hogger are dynamic shadows. These require so much processing it's no wonder the computer can handle them as well as it does.

[Flyby]

Interesting topic, Proclone, although i'm not totally surprized about the alpha slow down :
When you read the manual of World Builder, there is a specific paragraph that warns not to use too much 3 point blending.
Those are areas where 3 different terrain textures are blended together :

a short quote from the manual :

BLENDED TEXTURES
-You must blend textures to create more natural effects, but do not blend textures unnecessarily. There is
always a resource hit on blending textures.
-Three-way blended textures are resource hogs. Limit yourself to no more than 300 three-way blended
textures in your map. Fewer such blends make for better game performance.
-There should never be more than 50 three-way blends in a single shot of your map from the standard camera position.

This would confirm your findings about additive alpha blending...

[DeeZire]

Hmm, I still cant get a 2K poly model to show up in the game.

You could adjust the priority of each particle system so that only the really essential ones get displayed (like super weapon effects etc) but the others get displayed on super dooper machines I suppose, but that would make for a pretty dull looking game on most specs.

[Zoop]

for testing purposes, i had a 30k poly geosphere loaded in Generals, i could make 3 of em before the game became unbearably slow ... but jsut one of em seemed to be no problem, my system specs

P4 2.6ghz
512mb DDR
Geforce 4 Ti 4400

[Flyby]

come to think of it :
The most obvious impact on game speed is the AA setting you're using for D3D.

Setting it on high filtering, seriously degraded gameplay.
I recall setting the maximum quality "4xs" made it virtual impossible to play with a 2.6ghz/1mb system+GF4200.
Decreasing sampling quality to Quincunx (in combo with anisotropic filters)made it occasionally bog down,
AA of 2x or turning it off all together never caused a problem.
It even gets worse when combining with high anisotropic filter settings.