PDA

View Full Version : r_speeds


patchx
06-25-2002, 05:04 PM
i know that the tris count should be kept under 1000 with the r_speeds thingo but can anyone tell me what the maximum number of surfaces/shaders should be in a map according to the r_speeds thingo?

cause i know that too much shaders can effect framerate and im trying to make my new map with that in mind

-patch

Xcom
06-25-2002, 05:30 PM
Originally posted by patchx
i know that the tris count should be kept under 1000


That's a bit extreme don't ya think. For a SP level, I'd say 20K is a moderate limit if you wanna make ppl with lesser systems play your map. :D 20K including NPCs because they add a great deal to tris count.
About shader count, I frankly don't know. I have pretty crappy system and I noticed that if the texture memory hits 6-7 MB, the framerate drops dramatically.

Hubris
06-25-2002, 05:31 PM
Magic number on the r_speeds is generally 10k at the absolute worst, not 1000, 1000 would be a pretty plain map :p.

http://www.claudec.com/lair_of_shaders/tutorials/rspeeds_claudec/rspeeds_claudec.shtml

Good tutorial there.

Maps like CTF_Bespin which is very simple performance wise, is 2-4k.

Can't answer your question about shaders, just less is more is the general rule of thumb for me, also try and stay away from grass and vine shaders that protrude perpindicular from the brush, they seem to really adversely effect performance, if you played through the Yavin level SP you'll know what I mean. :p

Hubris
06-25-2002, 05:33 PM
Also, don't test r_speeds in third person, the polys on your model/lightsaber will effect your r_speeds and give you a false number.

patchx
06-25-2002, 05:40 PM
i try to keep it under 10000 before npc's because as soon as you put 7or 8 stormies into the picture it starts to go crap especially if there using weapons like the flack or rocketlauncher which produce heaps of little things for the game to draw to

what about the little thingo thats says for eg. 4.6mb tex on the r_speeds stuff

does that tell you how much ram or whatever is being taken up by textures?

and if so does anyone know what the max should be?

-patch

Xcom
06-25-2002, 05:43 PM
Originally posted by Hubris
Also, don't test r_speeds in third person, the polys on your model/lightsaber will effect your r_speeds and give you a false number.

But isn't it kinda "unwise" to ignore this "false" number because the majority plays in 3rd person anyway, w/ a lightsaber. :)

Xcom
06-25-2002, 05:46 PM
Originally posted by patchx
what about the little thingo thats says for eg. 4.6mb tex on the r_speeds stuff

does that tell you how much ram or whatever is being taken up by textures?

and if so does anyone know what the max should be?



Yeah. Keep under 6mb. :D

patchx
06-25-2002, 05:52 PM
thanks x

you scared me for a second with that whole;
About shader count, I frankly don't know

my faith in the walking encyclopedia of mapping knowledge that is you was shaken

-patch

Vitigis
06-25-2002, 06:57 PM
Originally posted by Hubris
Also, don't test r_speeds in third person, the polys on your model/lightsaber will effect your r_speeds and give you a false number.

Actualy, that's the reason you should ALWAYS use 3rd person to test your r_speeds.....because people use 3rd person when playing the map, as in MP, you have to use 3rd person for the saber. Remember, that 10k count should be 10k in the heat of battle! That means, sabers, guns, force powers...all that - though I think that systems now a days should be able to handle 15k fairly well, as quake 3 was released when the latest and greatest CPU was around 500MHZ and I beleive the baddest video card was the mighty TNT2 32 meg and that 10k mark was set aimed at the systems that were out at the time.

Large areas where centralized fighting will take place should never get close to the 10k limit, while smaller areas with limited fighting should be 10-15k or so.

VIO
06-25-2002, 09:53 PM
20K wher in the r_speeds does it have a number like that....

Hubris
06-25-2002, 10:23 PM
Well, if you're going to test with in third person, following the philosophy of "Well that'll happen" then also test with 15+ bots goofing off too, as they will certainly be there as well. You'll then reach problems of getting false readings as the model traffic moves through the level, which is why you test by yourself, against a bare minimum number.

The point is that it is a baseline measurement testing just the poly stress of your individual level, without other factors interfering such as less or more complex models, the Tavion model will probably give slightly higher r_speeds than the Ugnaught model when in third person.

If you wanna test in third person with other arbitrary factors, go for it, I'd rather do the math in my head and get proper readings. :p