[jpyrett]

Last night I noticed an important tweak (dont know if its been posted or not).. In UT 2003 (patched) and the UT2004 demo (w/ demo patch),
the VRAM is not fully used. It is set to 32 MB as default! To check this:

Open UT2003.ini or 2004.ini and find "32" (no quotes) and you'll see a line VAR...(cant remember exact string) =32 and looking in the log file UT2003.log will have a line similiar to "Using 32 MB VRAM"

Change this line to VAR.....=64, save and rerun and now the log file will say "Using 64MB of VRAM for OpenGL" .. much better!

Cheers!
Jacob, Cincy

[Horsepoo!!!]

I'm interested in knowing why Ryan Gordon or Epic decided to ship with configs that don't fully take advantage of graphics cards. Is there a technical reason? Is it because the game can't detect the card and make on-the-fly changes to switch from 32 to 64 or even 128 megs of RAM?

[Judge_Fire]

Originally posted by Horsepoo!!!:
I'm interested in knowing why Ryan Gordon or Epic decided to ship with configs that don't fully take advantage of graphics cards. Is there a technical reason? Is it because the game can't detect the card and make on-the-fly changes to switch from 32 to 64 or even 128 megs of RAM?

AFAIK, these tweaks need to be done on the PC side (DirectX instead of OpenGL), too. I've wondered about it myself; how many people out there aren't using their hw to the max...?

J

[jpyrett]

Thought about this some more.. From the in-game menu settings such as low, higher, highest are written over the old values in the .ini file. This is easy to see. If the VRAM setting was a default setting, it should be automatically overridden. My 9000 64 MB should force the .ini to change, but it doesnt.

The logs are saying the app is enabling << value>> from the .ini file. And is only allocating this buffer once according to the log.

Performance-wise, its hard to tell the difference as the FPS rate is so varied in the game depending on the action and randomness of action. My 1 GHZ TiBook runs frame rates around 40fps, but will dip in the low 20's and in the mid 60's if nothing is going on with the action. Either way, I am very happy with performance of ave fps ~55, by turning down some options (UT2003).. Havent played UT2004 much because going to buy the full version.

- Jacob

[Vash]

In both UT2k4 and UT2k3 setting the VRAM to 128 slows the game down extremely. Even though the Radeon 9800 Pro and Geforce 4 Ti both have it, the FPS is about 60% lower.

[Hydra]

AFAIK the VAR setting in the ini file only relates to caching of geometry in the video cards memory. If you fill it up with only geometry ( 128 of the Radeon Pro's 128MB) you have no room left for textures etc. this would result in everything but geometry having to be put in slower system memory. This is to the best of my understanding and I could be wrong. I have tried to put the VAR setting on MY dual 2.0/9800 to 128 and it degrades the performance greatly.

-Jerry C.

[Hydra]

is VARsize what you are talking about? http://www.thehaus.net/Tips/UT/ut2k4tips.shtml At the bottom of this page it mentions what I posted above.

" VARSize controls how much static geometry in megabytes can be cached in AGP memory."

-Jerry C.

[Johnny_B]

Okay, so except from the var, one should also turn up "CacheSizeMegs=" ???

Could anyone post ALL tweaks know for this game Would be great

[emark]

Sorry : edited : i need to read more before I type / speak...


So then, there are 2 different editable caches to think of:


"Chasing Apple" had this one about loading MAPs and what not into RAM...since you have to wait for everyone else anyway, for guys like me with older/slower machines (particularly bus and hard drive access at old ATA/66 this may be worth considering too:

link (see midway to bottom of 1st page)

anyway, I benchmarked my system witht the standard bot and flybyss in UT 03:

My dual 800 QS, and the performance was increased by setting this at 512 and even 256 mb...more the lowest frames number was increased than the average or max....which is really most important, right?

I have to admit to being suprised that the video card caching would be controllable, but interesting to see the results.

As far as UT 04 benchmarking, I have been a little confused as to running the standard tests....yes, I've seen the Bench UT at MacBidoulle (sp?) but I may not be using it correctly, I can get into levels, but it isn't automatically putting the actor(s) through their paces...if anyones worked through this, or can spell out a simple reproducible way to run the old standards at fixed settings we could generate some interesting tests across machines, and w/ varying cache settings.....

pipe up if you've had time to figure out ways to do this...

plus, I would expect one patch relatively soon to fix some bugs...though the ones I'm thinking of are more specific glitches than thinking something else will improve (retail) performance to any great extent....but just my hunch...

[emark]

I don't think it's conclusive, and nothing would replace good ol' benchmarking, but here is some further info re: original topic. seee bottom of page 2.


other thread

[RooneyX]

Whatever happened to AGP and sharing main system memory? Does this feature even exist on the Mac? Then why need 32MBs minimum for Quartz Extreme or even having to fiddle with ini files for games? I seriously think this hardware feature isn't properly enabled on Macs or programmers aren't taking advantage of it.

Is there any app to let me test if AGP texturing is working on a Mac?

[lenox]

Originally posted by RooneyX:

Is there any app to let me test if AGP texturing is working on a Mac?

Sure, play any game that uses OpenGL and has textures. Do you see them? Also if you have quartz extreme enabled, each window is treated as a texture by your agp graphics card.

[Olorin]

so in other words changing the VRAM size to 64 is going to slow things down cause there will be no room for textures?