To get the most out of AF in opengl for nvidia cards

On something like this? Probably hiding...

Oh, and one of our members started a thread about this over at nVnews...but they IP banned both ranges on my PC so I can't see it, but I imagine it is "interesting" to say the least.

 

This one is right...
http://www.nvnews.net/vbulletin/showthread.php?t=65727

 

Have to give Nvidia value for their AEG dollar.

 

erm... what? Don't be such a drama queen....

I don't have anything to do with AEG but how much you'll get from ATI by spreading this "driver issue"?

Wonder why no one can duplicate 3dcenter's performance hit results. I tried and only got 3-11% performance hit.

edit! typo

 

Good point-due of the obvious bias here at B3D ATI has gotten more "gain" from the DW, BZB, Walt C, Rwolf and the rest of the syncophants here than NV ever did from the AEG crew.
It's jusy too bad that the above can't/won't admit it.

 

Ok I can't read german and the translation gives me a headache. Is there any links to a good english translation?

 

Aw, did they let you out of nVnews for the day to play Ginfest?

Man, I really wish people would quit banging on the AEG people about this. They're just as surprised as everyone else is by this I think, and attacking/mocking/belittling them right now is not gonna help a bloody thing!

I was hoping one of them would join the thread saying something like, "we asked AEG about it, and they're asking nVidia...we're still waiting on a response".

From what I understand of the program AEG might have a direct line to nVidia, but nVidia is still nVidia even to them so it takes a bit for them to answer any questions...if they do.

 

If you look at the benchmark scores, they are done with 4xAA enabled.. which is probably a factor in this (the benchmarking done at NVNews is done with pure AF filtering and AA is not factored in at all). The most current benchmarks include AA enabled.. so the performance hit will probably be closer to 20% than the 10% w/o AA.

The issue is that since the bug has been there for a long period of time, and reviewers not knowing this (we can assume for the most part that they just set the "high quality" setting and then benchmark) will generally skew the score... for ATI, the mere fact that this is happening for the current and past reviews.. they can use this in their "PR slideshow" explaining that lower quality IQ in OpenGL was applied in all major review sites skewing benchmarks in NVidia's favor (especially with the fact that NVidia is known to have the better OpenGL implementations, ATI will most likely mention about the speed specifically, not about the good backwards compatibility). It's not ATI's fault that this bug has existed for over a year.

 

STOP THE PRESSES!

So I sneaked into ChrisRay's house, and here's what my trusted DigiCam came up with:

 

Well, may I say just please give them a break .
If whatever they do are just being a beta tester, they will try to make communicate with NV (hopefully). Anyway, on ATi part of optimization... AFAIK, they always steped out and commited about their optimization in the pass more than that of NV since what I saw anytimes was that NV just only keeps quite about it...

 

Sorry, do I seem to be a bit jaded.

 

Well I did rerun my tests with AA... still performance hit is between 3-11%.

http://www.nvnews.net/vbulletin/showpost.php?p=828827&postcount=10

 

You are using Clamp as well, yes?

The other thing in your post that you didn't mention if AA+AF were forced from the CP or application determined.

You probably want to consult with the 3DCenter authors though...

 

Yes, Negative LOD bias = Clamp.

AA+AF were forced from CP but I will rerun tests again with APP AA 4x/AF 16.

 

There are definately performance and IQ differences when you compare CP forced to application request.

I don't particularly recall what level of filtering NVidia uses when AF is forced.. I'm pretty sure it's at the same level as ATI's or less.

 

Ok, I ran the tests with APP AA/AF but no performance or IQ differences (e.g pixel tests using Photoshop) compared to CP AA/AF.

 

Hanners at EB takes a look-see: http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=23

He finds it to be a control panel issue, rather than an in-game settings issue.

Toss this one in the bug file, return the B52's to the failsafe line, and stand down to DefCon4.

 

Or just use a third party tool for all those settings (like nHancer) and save yourself the headache of managing all those bugs in nVidia's control panel.