Latest Riva Tuner Aniso optimizations...

[Ante P]

I've been experimenting for half an hour or so to get good results.

Without getting any degredation in quality I managed to boost the performance by 2%.

What I would like to know is what the default settings for each level are and what you other guys have tried.

Well, I don't have the time to play "mix and match" with these settings, but I'm sure someone else will do it.

Well let me know. Hopefully I'll find some good settings to use as I am doing a "Aniso + FSAA shootout" between a Radeon 8500 128 and a GeForce 4 Ti 4600.
(Though I must admit, even though there's no quality degradation it still seems like "cheating" as no "normal" users will use these settings and thus the results are pretty misleading.. on the other hand a user MUST use a tweaker in order to even access D3D aniso options so I dunno..)

 

[Typedef Enum]

A couple of things...

- I'm going to take a wild stab, but I _believe_ that what Alexey has done with RT is a somewhat low-level approach to what will eventually become standard with the nVidia drivers. In other words, there will be some algorithms which will utilize these different settings to more optimally use A.F. Perhaps it will be an adaptive approach, similar to ATI...

- You have to really play around with those settings. You can get _far_ better than 2%...I had to email Alexey about setting up the more optimal configuration, as something didn't seem intuitive...Once I understood the interface a little better, I got beter results.

- OpenGL Script. Did you execute the patch script? Do it. As an example (I'm @ work, so these numbers aren't totally accurate)...

On my P4 2.7 GHz. system, I got like 173 FPS with 8x A.F. enabled + patch. Without the patch, I got like 165 FPS...Under both scenarios, I also had enabled the "performance" setting. When I used "quality" the perf. dropped down to like 130 FPS.

So, between the patch + perf. setting, I was able to boost perf. by almost 50 FPS...Not having done any scientific I.Q. analysis, I didn't seem to notice any quality degredation either...

 

[alexsok]

I also got an increase in perfomance, both in Direct 3D & OpenGL.

I wasn't actually benchmarking to see the exact boost I got, but it just felt smoother!

Also, I agree that we need to play a bit with the options presented in this new release, cause the boost achievable with different settings is pretty substantial and WILL make a difference!

 

[Ante P]

I'm mainly interested in which levels you vhose to alter and to what Aniso level?

Even if you aren't done tweaking just throw some settings at me.

 

[AlexUnwinder]

I've been experimenting for half an hour or so to get good results.

Without getting any degredation in quality I managed to boost the performance by 2%.

What I would like to know is what the default settings for each level are and what you other guys have tried.

Well, I don't have the time to play "mix and match" with these settings, but I'm sure someone else will do it.

Just my two cents: first, I strictrly recommend to read context help on these options (just click RMB on the list of texturing stages) before using them, you must perfectly understand the settting you use in order to acheive good results.
Second, you must understand that such AF optimization (limiting the maximum degree of anisotropy on per-texturing stage level) cannot be absolutely free. The resulting image quality totally depends on your understanding of these options and specific of the game's engine. For eaxample, I know that UT's d3d renderer utilizes two texturing stages and it always uses texturing stage 0 for filtering base textures and texturing stage 1 for lightmap filtering. In this case, I can safely go to RT's optmization settings, leave texturing stage 0 unchecked, check texturing stage 1 and select AF limiter to 1. In this case, the driver will filter base textures with the level you've forced, but lightmaps will be filtered with level 1 (i.e. bilinear filter) only. But these settings can be bad for the rest 3D engines, everything depends on the game engine's texturing stage utilization behavior.
Basically you've to leave only one texturing stage unchecked and limit AF for the rest texturing stages, so the most usual settings are:

stage 0 - unchecked
stage 1 - checked
stage 2 - checked
stage 3 - checked

AF limit - level 1

Just run 3DMark2k1 with these settings and compare IQ with the original AF. You'll see that thes settings are really good for Game2, but they are bad for Game1 (texturing stage 2 brings the most of details in this test).

 

[Hellbinder]

Second, you must understand that such AF optimization (limiting the maximum degree of anisotropy on per-texturing stage level) cannot be absolutely free. The resulting image quality totally depends on your understanding of these options and specific of the game's engine. For eaxample, I know that UT's d3d renderer utilizes two texturing stages and it always uses texturing stage 0 for filtering base textures and texturing stage 1 for lightmap filtering. In this case, I can safely go to RT's optmization settings, leave texturing stage 0 unchecked, check texturing stage 1 and select AF limiter to 1. In this case, the driver will filter base textures with the level you've forced, but lightmaps will be filtered with level 1 (i.e. bilinear filter) only. But these settings can be bad for the rest 3D engines, everything depends on the game engine's texturing stage utilization behavior.

My first thought... WHAT A HASSLE!!

Is it possible for you to add some kind of applications preference tab.. so that when ut.exe is detected it applies said values?? or at least a quick reference tab of pre defigned games... like Ut engine games.. Q3:TA etc,...Soemthing that could be pre defigned and then right-clicked from the taskbar icon???

Just a thought.. Why dont all companies do like PowerVr and have individual game settings that you set once and then it just "knows" when to apply it as teh exe is run...

 

[Ante P]

Guess I'll definitely not be using this in any reviews. I dare to state that most (90+%) of our readers couldn't be bothered. (And of course most of them don't use ANY tweaker anyways..)

But I do think it's great for experimentation though. Personally I think I'll just leave it as it is if it indeed needs changes on a per-engine/game basis.