Tom's GF4 anisotropic scores in his Parhelia review

[noko]

Did a quick series of test once again in SS:SE. GF3Ti200 clocked at 240/500, DX API vice OpenGL this time (I started testing and then realize I wasn't in OpenGL ), 20.42 drivers in W2K, I wonder if the drivers behave different in WinXP? Hmmmmm. Anyways here is what I got in Demo "The Elephant Atrium" with 8x anisotropic filtering using Tri samples at 1152x864x32.

Game setting/test Texture LOD bias set to:

• max blur - 58.9 FPS
  normal - 51.3 FPS
  max sharpen - 45 FPS

Reference images for each setting with mipmaps colorize:

Max blur mipmaps
normal mipmaps
Max sharpen mipmaps

I was surprised at the performance differnce from 58.9 to 45 FPS by just adjusting LOD. That is a %31 difference.

 

[OpenGL guy]

Did a quick series of test once again in SS:SE. GF3Ti200 clocked at 240/500, DX API vice OpenGL this time (I started testing and then realize I wasn't in OpenGL ), 20.42 drivers in W2K, I wonder if the drivers behave different in WinXP? Hmmmmm. Anyways here is what I got in Demo "The Elephant Atrium" with 8x anisotropic filtering using Tri samples at 1152x864x32.

Can you compare the older drivers to the new? Then we could see what differences there are in IQ.

 

[noko]

Can you compare the older drivers to the new? Then we could see what differences there are in IQ.

I did do that, there was no noticeable difference that I could tell. I have some 8x shots on this thread with mipmaps colorized. I also have shots with no anisotropic, 2x, 4x and 8x images and colormipmap images which I didn't publish except the 8x images. There was no significant difference, to me, they where like mirror images of each other. If you want me to post those let me know.

Update, opps it was in a different thread, here you go for 8x:

Reverend,

Looks like it maybe a GF4 thing more so then a GF3 change. Well in any case the performance and quality between 28.32 and 29.42 is virtually identical in all respects. Here is a graph doing a performance comparisons using anisotropic filtering between the 28.32 and 29.42 drivers on my Gf3 Ti200. Using my own configuration for all test which stayed the same, this is what I got. SSSE, The Elephant Atrium, modified Max Addon configuration (same in each test except for anisotropic filtering settings), 245/530 core/mem.

Here are some comparison images for anyone who is interested, good luck in finding anything significant.

28.32 8x anisotropic
29.42 8x anisotropic

28.32 8x anisotropic, colored mipmaps
29.42 8x anisotropic, colored mipmaps

 

[OpenGL guy]

Thanks for the information, noko.

Anyone else have a GeForce 4 so we can get a full range of comparisons?

 

[Ailuros]

noko,

Here´s a snippet from my initial post on page 1 of this thread:

29.42 WHQL
Rivatuner RC11
demo001/1024x768x32/LODcurveError@30000/max detail
4xOGMS

64tap quality optimisations: 73.9 fps
64tap performance optimisations: 86.9 fps

16tap quality optimisations: 82.3 fps
16tap performance optimisations: 92.1 fps

Assume there is a possibility that the drivers with no third party tweaking utility involved switch by themselves to "performance" mode in highest aniso settings how irrational is my theory since the performance delta as shown above is ~87 vs 82 fps?

Given the fact that 28.32 drivers did not expose performance/quality settings for anisotropic filtering in Rivatuner and appeared only in 29.42, I still see no other logical explanation according to what I´ve been able to collect so far.

When I get home I´ll switch drivers to default settings and rerun 4xFSAA/16 tap vs 4xFSAA/64 tap to see what if they truly match the highlighted scores above on absolute driver defaults. Also I´ll try to see if between quality/performance with 8 level aniso any changes in lightmaps can be reproduced with screenshots.

 

[noko]

I don't use Rivatuner, I do use PoweStrip, so I am at a lost on Performance/Quality settings since that is not available, at least yet in PS3. I only used PowerStrip to overclock in my tests posted, didn't use the other PS settings to tweak except refresh rates for my different W2K resolutions.

 

[Ailuros]

I reset drivers to default settings and deleted in the registry every reference to Rivatuner.

I ran three times the demo 001 in q3a (same settings as mentioned before) with 4xFSAA/2 aniso and three times with 4xFSAA/8 aniso. In two occassions (for each aniso setting once) FSAA wouldn't kick in at all. All other cases had a performance gap between 2x and 8x level aniso between 5-6fps.

Anyway here are a couple of screenshots from DM1 (note: neither in the benchmarks or screenshots was any TC used). One with 4xFSAA/L8 performance and one with 4xFSAA/L8 quality. I tried to shoot to get as close in position as I could. There is a difference but it's so small I would be hardpressed during gameplay to notice it.

Ack I'm giving up

 

[Seiko]

perhaps the difference is that there is no difference

I did a very quick article, using MOHAA.

http://www.beyond3d.com/reverend/gf4aniso/index.php

Rev,

Took a look at the images, is it just me or is the difference between 4* and 8* aniso no existant. Surely that would account for not loosing any more FPS when shifting from 4* to 8* ?

Maybe it's my eyes but the level between 4 and 8 is not worth the loss?