VGAChart 2- shocking NV25 OpenGL Aniso performance

[Col_Kiwi]

For those of you who haven't read it, the VGA Chart II's tests for FSAA and Anisotropic filtering show that the geforce 4 ti's anisotropic filtering is much, <b>much</b> faster in opengl then in direct3d!

Even fast enough for a Ti4200 to beat a Radeon 9700pro (holy [-peep-]!) in JKII with 8x aniso when FSAA is not used!

Don't get me wrong, I'm not saying this is the end of R300's superiority over NV25, because it's not, since when you use FSAA too the R300 leaps far ahead again. But it's certainly something to note for people who already own an NV25.

Many direct3d games can be set to run in opengl, if you bother to look for the option. The two d3d games i play most are WarCraft III, which can be set to opengl rendering by a command line argument, and UT2003, which can be made to render in opengl by changing two lines in it's main ini file. I'd never bothered running either of these games in opengl on my Ti4200-128 before, because there was no need that I saw. But now, you can bet I intend to, since I use aniso and FSAA. The boost in performance is more than worth it!


Comments?

-Col.Kiwi

 

[phsstpok]

That JKII, OpenGL, 8X Anisotropy chart is suspicious.

Look at the <A HREF="http://www.tomshardware.com/graphic/20030120/vgacharts-06.html" target="_new">CPU Scaling chart, which is also JKII but no anisotropy</A>, and compare with the <A HREF="http://www.tomshardware.com/graphic/20030120/vgacharts-08.html" target="_new">8X Aniso chart</A>. The numbers are virtually the same for all video cards. How could all the cards including the slower ones manage nearly the same framerates with and without 8X anistropy?

<b>99% is great, unless you are talking about system stability</b>

 

[Col_Kiwi]

Weird. Quite weird. Doesn't really explain why the nVidia cards did better then the ATI, though... very weird.

I'll have to experiment tonight with fps tests in UT2003 with and without aniso in D3D and OpenGL. I'll post results here afterwards.

-Col.Kiwi

 

[phial]

would visual quality and/or special effects be affected by choosing opengl over d3d?

i mean, sinse UT2003 is a dx8 game, would opengl be able to do the same effects (pixel shaders etc etc)?

i dont know much about the differences between the 2 i admit

 

[vacs]

there is only one thing which can explain why GF4s are faster than R9700s in some games/configurations: --> drivers

even though ATI improved their drivers big time, they are still leagues away from nvidia current driver performance...

 

[Col_Kiwi]

OpenGL also contains advanced shaders. In fact in many cases, OpenGL is technically superior to D3D.


vacs: Yeah...

-Col.Kiwi

 

[phsstpok]

I only have the demo of UT2003 but I found the benchmark was faster with OpenGL. Didn't test this with 8XAA, though.

By the way, what's the command line switch for using OpenGL in UT2003? I've been editing the Benchmark.ini file. A new shortcut would be a lot easier. One interesting feature in the .ini file is the ability to run the benchmark without rendering any graphics. This way you can test your CPU processing power. You can also compare how much your own graphics card is limiting your performance if, in fact, it is.

<b>99% is great, unless you are talking about system stability</b>

 

[LtBlue14]

bump, i want to know the results of your test col kiwi, and what command line to change in the .ini

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[flamethrower205]

ogl is more exact, overall better quality.

"If everything seems under control, you're just not going fast enough."
- Mario Andretti

 

[rubikian]

Here's a <A HREF="http://gamespot.com/gamespot/stories/hints/0,10872,2537710,00.html" target="_new">link</A> to run programs in certain command line like OpenGL. Try to see if it works...

 

[LtBlue14]

this did not work when i ran C:\Program Files\UT2003\System\UT2003.exe /640
in hopes of making the game open in 640x480 res
winxp sp1
thanks for the link but i think it's only for Homeworld (also given some of the commands like "demorecord," which many games don't support)

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[flowmaster]

its beats 9700 pro. hAHAHA!
Just another way to for Nvidia to try and pull some crap that its better then ATI along the same lines of Intel vs AMD were Intel pulls crap in and out.

 

[phsstpok]

Don't know about the command line switch but to change the renderer by editing the .ini files here's how. It works for the demo. I think it should be the same for the full game.

For the benchmark go to the the UT2003 game folder and look for a folder call Benchmark. In this folder is file called MaxDetail.ini. This is the file you need to edit. I forget the exact syntax but there is a place where 3 rendering engines are listed. They are D3D, OpenGL, and None (I think it's "none"). Two rendering engines are commented out with a semicolon ";". Just changes the line by removing the semicolon from the render that you want and place a semicolon in front of two you don't want. Save the file. Now the benchmark will run with renderer you chose.

It's similar for the demo game. From the UT2003 folder you'll see another folder called System. In this folder is a file called UT2003.ini. Edit this file the same as describe above. Select the renderer you want edit appropriately. Save the file and you are set to play.



<b>99% is great, unless you are talking about system stability</b>

 

[LtBlue14]

ok i'm trying this right now by changing
RenderDevice=D3DDrv.D3DRenderDevice
;RenderDevice=Engine.NullRenderDevice
;RenderDevice=OpenGLDrv.OpenGLRenderDevice

into
;RenderDevice=D3DDrv.D3DRenderDevice
;RenderDevice=Engine.NullRenderDevice
RenderDevice=OpenGLDrv.OpenGLRenderDevice

i'm going to take a screenshot of a long slope with no AF in D3D, then with AF in ogl, to know which one to run the game in =)
i'll report back with results

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[phsstpok]

That's it "NullRender" not "None" but glad you found what I mean.

Curious about your results.

I haven't gotten around to testing 8XAA in OpenGL vs D3D. Don't think Radeon 8500 + Tbird + SDRAM results are very interesting but I'll post anyway.

I still think the THG VGA chart for 8X aniso is wrong.

<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 01/22/03 03:18 PM.</EM></FONT></P>

 

[LtBlue14]

i agree with you that it is wrong
here's system setup etc:
p4 1.8 @ 2.3 (127x18), DDR@ (166+27)x2= 386, 2.5-3-3-6-2T
Geforce4 Ti 4400 overclocked to 310/620
Tests done @ 1280x960 @ 60Hz (the nvidia refresh rate fix only works in D3D, so i made it 60 for both)
UT2003 benchmark results:

D3D, no aniso: flyby 119, botmatch 55
D3D, 8x aniso: flyby 44, botmatch 23

OGL, no aniso: flyby 119, botmatch 55 (same as D3D)
OGL, 8x aniso: flyby 44, botmatch 27 (botmatch slightly better)

Applying AF absolutely resulted in a better image on long planes (such as the floor of the arena), where the further tiles looked much sharper with 8x AF on. However, the performance hit is much too great, it's definitely not worth it, and NV25 cards definitely do not process AF tons better on Open GL than on Direct 3D.
my 2 cents.

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[phsstpok]

Here are my results for all the good it does (see my explanation below). By the way, I didn't realize you tested at 1280x960. My tests were at 1024x768.

Athlon Thunderbird 1.0 @ 1.5 (150x10), SDRAM 2-2-2 etc
Radeon 8500 overclocked to <b>[EDIT-->]</b>275/600, Catalyst 3.0 drivers
Win98 SE
Tests done @ 1024x768 (Vsync disabled)
UT2003 benchmark results:

D3D, no aniso: flyby 109.6, botmatch 41.2
D3D, 8x aniso: flyby 106.1, botmatch 40.7

OGL, no aniso: flyby 100.7, botmatch 37.1
OGL, 8x aniso: flyby 98.4, botmatch 36.5

OpenGL was faster but something is wrong. D3D with and without Anisotropic filtering not only ran essentially the same they even looked the same. I could detect no visual quality improvement with AF.

In OpenGL the benchmark with 8X AF was a little slower but looked a whole lot better.

I expected only a small penalty for AF with an ATI video card and this was true with OpenGL. However, D3D + 8X AF should have looked better. D3D AF seems to be broken. Not sure what's wrong here.

<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 01/24/03 11:33 AM.</EM></FONT></P>

 

[LtBlue14]

that's weird that you couldn't detect any improvement, that's got to mean that it wasn't turned on at all...is there any sort of "force" command, in case it's giving the application the choice of on or off? with v-sync, for example, there's "always off" "off by default" and "on by default" in my options, is there possibly an "always on" on yours when dealing with aniso?
otherwise i have no idea unless it has to do with those particular cats (3.0 right?)

i also thought about trying a game that's naturally run in ogl, like JKII, so i tried soldier of fortune II and quake III, and experienced substantial fps hits (though again, better image quality if you're really looking for it instead of trying to dodge those bullets)
so i have again come to the conclusion that crime does not pay, err i mean that GF4 Tis do not run AF 8x very well even in open gl
--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[phsstpok]

ATI drivers for Anisotropy have a choice of Application Preference or an overide setting, 2X, 4X, etc. One choice each for D3D and OpenGL.

OpenGL is working. D3D is not working.

Since the override option didn't seem to be working I switched to "Application Preference" and tried to change anisotropy from withing the MaxDetail.ini file. There's a setting called "LevelOfAnisotropy". This is what I tried changing. The default is "1". I tried "8". No effect. Thinking this might be an <i>n</i>-tap number instead of <i>n</i>X number I tried 32, 64, and 128. No effect.

I'm going to run some other applications to double check what I am seeing (not seeing).

<b>99% is great, unless you are talking about system stability</b>

 

[phsstpok]

This is strange. It's working fine in Serious SAM in both D3D and OpenGL modes. I can switch between API and levels of Anisotropy from within the game and Anistropy definitely is working.

It's still not working in the UT2003 Benchmark in D3D.

Anyway, now I have another game I can test with.

<b>[Update]</b>

OK, I discovered if I change the level of AF for D3D in Display Properties AF doesn't work. However, if I set to "Application Preference" and change the AF setting from with Serious SAM it works fine.

AF works fine for OpenGL either way.

<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 01/24/03 11:39 AM.</EM></FONT></P>

 

[Col_Kiwi]

Here are my (delayed, sorry!) test results:

1024x768x32@60Hz
Hardware:
-P4 1.8A @ 2.4B
-256MB DDR333 CAS2.5
-ASUS P4S533
-GeForce4 Ti4200 128MB running at 275mhz core and 550mhz ddr (same as Ti4400)

OGL-noAA.noAF: 143.8F/50.3B
OGL-noAA.8xAF: 58.6F/36.6B
OGL-2xAA.8xAF: 49.1F/31.8B

D3D-noAA.noAF:146F/51.1B
D3D-noAA.8xAF:58.6F/36.2B
D3D-2xAA.8xAF:49.1F/33B




clearly the same results as you guys.

-Col.Kiwi

 

[phsstpok]

Hey, thanks. I didn't mean for you to re-run tests. I was going to do it but I got sidetracked when I discovered I'm having problems with AF in D3D.

I have partial results in case you are interested.

Tbird @1.5, PC-150 SDRAM, Radeon 8500 @275/300

UT2003 Benchmark (1280x960)

8X AF
Flyby - 72.0
Botmatch - 33.7

No AF
Flyby - 76.6
Botmatch - 35.5




<b>99% is great, unless you are talking about system stability</b>

 

[LtBlue14]

so now onto the bigger question, WHY DO TOMS BENCHES SEEM TO BE RIDDEN WITH ERRORS RECENTLY?

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

 

[phsstpok]

[shrugs]

<b>99% is great, unless you are talking about system stability</b>

 

[Col_Kiwi]

Yeah... i'm wondering that too..

-Col.Kiwi