The truth about UT2003 and nVidia

What you say is all nice and fine Walt, and in principle I agree on a lot, but as you put it yourself: too many games simply don't even offer us the options we'd need to set our IQ as we'd like it. Nor do they seem to care much about setting e.g. the optimal filtering level for the different texturing stages, or there would be no need for user or IHV optimizations in the first place, wouldn't you agree? Thus many users (including me) have gotten into the habbit of controling IQ the only way they can: setting options through the control panel. IMO until the day arrives where the majority of applications lets us configure our desired IQ settings properly and tells the drivers explicitly what to do when it comes to filtering, the CP will continue to be the center of our IQ needs.

The CP gives us the option to choose in several steps between performance and quality, so when I set the slider to its highest quality option available, I expect to get the highest quality possible, it really is (or should be) that simple. If performance would suffer so horribly, then please at least give me the option to choose "ultra quality" or whatever and put a note there (warning: increases IQ slightly, but in many cases at the cost of performance). But as you can see when you have a look at 3DCenter's benchmarks though, the majority of games doesn't even suffer in performance at all, some even appear to run faster with full trilinear forced!

So, while I agree when you're saying that the application should ideally tertermine the filtering of textures itself, we're simply not there yet and I'd like my drivers to be as functional as I need them for today's games, not for an utopian tomorrow. As this UT2003 debacle serves to illustrate not even the latest and greatest games really tell the drivers what should be filtered and what not, or if they do the IHVs simply choose to ignore or override it, neither of which is desirable IMO...
 
WaltC said:
Bottom line is that the driver guys aren't idiots. Lots of things they do have specific and well-justified reasons behind them.

Yes, to get that next design win. Performance vs. price is the whole game.
 
Gollum said:
What you say is all nice and fine Walt, and in principle I agree on a lot, but as you put it yourself: too many games simply don't even offer us the options we'd need to set our IQ as we'd like it. Nor do they seem to care much about setting e.g. the optimal filtering level for the different texturing stages, or there would be no need for user or IHV optimizations in the first place, wouldn't you agree? Thus many users (including me) have gotten into the habbit of controling IQ the only way they can: setting options through the control panel. IMO until the day arrives where the majority of applications lets us configure our desired IQ settings properly and tells the drivers explicitly what to do when it comes to filtering, the CP will continue to be the center of our IQ needs.

The CP gives us the option to choose in several steps between performance and quality, so when I set the slider to its highest quality option available, I expect to get the highest quality possible, it really is (or should be) that simple. If performance would suffer so horribly, then please at least give me the option to choose "ultra quality" or whatever and put a note there (warning: increases IQ slightly, but in many cases at the cost of performance). But as you can see when you have a look at 3DCenter's benchmarks though, the majority of games doesn't even suffer in performance at all, some even appear to run faster with full trilinear forced!

So, while I agree when you're saying that the application should ideally tertermine the filtering of textures itself, we're simply not there yet and I'd like my drivers to be as functional as I need them for today's games, not for an utopian tomorrow. As this UT2003 debacle serves to illustrate not even the latest and greatest games really tell the drivers what should be filtered and what not, or if they do the IHVs simply choose to ignore or override it, neither of which is desirable IMO...


Gollum the point I was making is that with UT2K3 and the Catalyst Cpanel; we are there, already....;) If you set the Catalysts to Application Preference, go into the simple UT2K3.ini textfile, and look here:

[D3DDrv.D3DRenderDevice]

and make sure the following value is set

UseTrilinear=True

You'll get full trilinear, as I understand the issue.

There are many other settings in the UT2K3.ini that you can change to impact the way the game renders as well--which are not adjustable either from the game's GUI or from any videocard's driver interface, such as LOD in the game, for instance.

What I'm trying to communicate is that as long as 3D games are written by different programmers using different approaches, which is most likely forever, forcing these things from driver control panels will always be done from a "best guess" scenario on the part of the IHV--nothing improper here at all. In any event, you can do nothing about UT2K3's rendering LOD from any driver control panel--this adjustment as well as that of many other parameters can only be accomplished from within UT2K3's internal configuration files. I have my LOD set to a higher-than-default level which impacts frame-rate performance but provides better IQ because I like it that way. The driver Cpanel can't help me with that at all--so like it or not, even if the Cpanel allowed me to set correct texture treatment stages for UT2K3, I'd still have to go into the game's configuration files to work with LOD and all of the other settings the Cpanel (by any IHV) doesn't address.

I've often wondered how many people realize that if they set their control panel to 2x FSAA, but set up the internal controls for a game to 4x FSAA, that the Cpanel setting will force 2x FSAA instead? Or what about the fact that if you have the ATi vsync Cpanel set to "Always On" and if you try and turn off vsync from within the game, the driver ignores the command and keeps vsync turned on anyway? Then, suppose you have one game which you like to run at 2x FSAA, and another you like to run at 4x FSAA, and another you like to run at 6x FSAA? If you have to use the Cpanel to force those levels of FSAA because there are no in-game controls, OR you forget to set the Cpanel to "Application Preference"--you've got to constantly go back to the Cpanel to set the changes you want for each game you run. So you can see that the use of the Cpanel to force modes can be inconvenient, but it's most important to realize how *incomplete* this approach is to setting up IQ within individual games.

I think this problem in large part is caused by software developers. In UT2K3 for instance there ought to be an accompanying text file which lists all of the options provided in the internal configuration files, enumerates their conventions, and explains them in layman's terms so that end users can use these settings to enhance their enjoyment of the software. The lack of such explanations is probably why many people stay away from the game's configuration files and use the IHV driver overrides instead, which they think they understand.

But, I can certainly agree with you in saying "we're not there" from the standpoint of software developers understanding that people want to set these parameters in-game, whenever possible. That's the developer's fault--not the IHV's. The IHV's have done for the most part a terrific job in trying to compensate for developer laziness in this regard, I think. But it's only proper to blame the developer when the occasional imperfect situation develops despite the best efforts of the IHV. With respect to the Catalysts and trilinear in UT2K3 the problem is only one of convenience, not function support. (But it really isn't "inconvenient" when you think of the fact that you have to go into the .ini configuration files anyway to set up other IQ settings that aren't addressable through the Cpanel.)

As an aside, I wonder with amazement how it is that any software developer can understand why end users might wish to see the kinds of IQ possible though advances in the APIs, and work to support such advances in their software, but then turn around and declare that matters of IQ are "subjective" and not worth the trouble for them to directly address. That seems to be a blind spot some of them have, apparently. The differences in trilinear and bilinear support in an application might well be "subjective" insofar as how they are perceived by different people when rendering; however, implementing tri and bilinear support (as well as LOD and lots of other things) in your software is a purely objective exercise for the developer. You either do it and do a good job of it, or you do it poorly, or you don't do it at all. The cpanel approach from IHVs is just an effort to work around software developer deficiencies. As explained earlier in this post, we ought not ever expect IHV Cpanels to replace software developer responsibility for rendering IQ, IMO.
 
I thought the article was, as is typical of 3DCenter, thoroughly benchmarked and nicely documented. Well, the pics at least--my German is nonexistent, so the text could have been Moby Dick for all I know ;). Thanks for posting this here, Demirug. Sure, it may be old news to members of this board, but not everyone reads this board. It's good to see more people taking IHV's to task for lowering IQ in order to raise their benchmark numbers, particularly when the IQ hit is as pronounced as it is with nV at 8xAF.
 
Back
Top