Beyond3D Wolves Needed to Tear This One Apart

[Pete]

I (and presumably many others) are still waiting to see some specific examples of that "nvidia centric way of explaining things" that pete insists Josh suffers from.

I may be generally daft, but I'm pretty sure I didn't say Josh was green-tinting things with this write-up. In fact, I said slightly the opposite. digi and some others thought so, tho, and thus I could only assume you were talking to them.

Oh, whoops, you meant PeterAce. Well, if he said he hashed things out with Josh, what more needs to be said? I guess you can PM Josh or the Ace for clarification, if you need. As I said, tho, I thought this article was even-handed, and even verging on not giving nV enough credit for NV40.

I just don't agree with much of your reply to my Big Post. That gamma-correction thing in particular kind of freaks me out, man, as it makes me doubt that we're even on the same page WRT what to expect from a GPU. How you conclude GC is tampering with IQ beyond JG vs. OG AA or MS vs. SS AA--to say nothing of bri vs. tri--is baffling.

But not really important. I don't doubt you're happy with your 5900XT, and surely most owners are, too (especially those who got in on the CoD bundle). It's not worth our time to haggle beyond that.

 

[radar1200gs]

With regard to gamma corrected AA, the GPU is modifying color data without the applications explicit permission.

This doesn't happen with other AA or filtering (they don't modify texture/shader content) merely filter it.

 

[Deathlike2]

Developers should quit trying to force rendering paths down gameplayers necks and instead give them the option of which path they would like to use, just like driver optimizations.

I agree with that, however... they should be given a default choice, if the path gives unacceptable performance.

Giving the users total control and not understanding the performance impact on certain hardware will NEVER be a good thing for the people who deal with these things (Quality Assurance perhaps???)

It's like forcing the DX9 mode on a FX5200 in Half Life 2 - very unacceptable.

 

[radar1200gs]

Yes, nothing wrong with a default choice, so long as the user can change it if desired.

 

[John Reynolds]

With regard to gamma corrected AA, the GPU is modifying color data without the applications explicit permission.

This doesn't happen with other AA or filtering (they don't modify texture/shader content) merely filter it.

Any control panel forced AA is modifying the final output without the application's explicit permission.

 

[hovz]

yes radar was in favor of nvidias shader replacements, yet they clearly modified the game without the games permission. fan-boys these days

 

[OpenGL guy]

With regard to gamma corrected AA, the GPU is modifying color data without the applications explicit permission.

Gamma correction is required only on edges. Interior pixels are completely unaffected.

This doesn't happen with other AA or filtering (they don't modify texture/shader content) merely filter it.

So what you are saying is that AA modes that don't account for gamma correction are incorrect. Thanks, I knew that already.

Apps take care of gamma corrected filtering by either prebaking the texture data with gamma or using sRGB textures.

 

[radar1200gs]

yes radar was in favor of nvidias shader replacements, yet they clearly modified the game without the games permission. fan-boys these days

I was in favor of shader replacements that provide identical visual output to the original shader. My stance has not changed. Have I said one word (prior to this post) about ATi's modified Doom3 shaders?

OpenGL guy, kindly do not place words in my mouth. Non gamma corrected AA methods do not alter the color data of the applications textures etc.

I would have no problem with ATi's gamma correction if it were able to be toggled on and off as the user desires. Unfortunately its like most ATi stuff - forced on and if that doesn't suit you, tough - there's no way around it.

 

[OpenGL guy]

OpenGL guy, kindly do not place words in my mouth. Non gamma corrected AA methods do not alter the color data of the applications textures etc.

Neither does ATI's gamma corrected AA. As I stated earlier, only edge pixels are affected, as they should be.

I would have no problem with ATi's gamma correction if it were able to be toggled on and off as the user desires. Unfortunately its like most ATi stuff - forced on and if that doesn't suit you, tough - there's no way around it.

What a comeback! Since you don't own an ATI card, maybe you should stick with what you know?

 

[radar1200gs]

So, you are telling us that edge pixels don't use texture or shader data??? the things learn on beyond3d from the experts!!!

 

[OpenGL guy]

So, you are telling us that edge pixels don't use texture or shader data??? the things learn on beyond3d from the experts!!!

And you're telling me that doing an ungamma corrected filter is correct?

 

[radar1200gs]

It is doing what the app intended to be done.

 

[John Reynolds]

It's time for the official B3D drinking game! Oh, crap, that means I'm drunk already from this thread alone. 8)

 

[OpenGL guy]

It is doing what the app intended to be done.

That doesn't make it correct. Gamma corrected AA is a quality improvement. The fact that the app can't control this feature (because it's not supported by the API) doesn't make it incorrect.

 

[John Reynolds]

Makes you wonder if Greg ever railed against Quincunx AA. Somehow, I doubt it.

 

[OpenGL guy]

Makes you wonder if Greg ever railed against Quincunx AA. Somehow, I doubt it.

That's not my problem. If I can get him to change his mind or admit he's wrong about one single thing, then I'll consider my life a success

 

[radar1200gs]

This is like so many other issues I have argued on for the past two years.

Yes, gamma corrected AA can be an IQ improvement. However it is not a standard part of the API and should be strictly optional (just like all other optimizations).

Provide all the optimizations you like. I don't mind, just make them optional and let the user or the app decide (user decides through app profiles/cp, app decides through default or app preference setting). Don't force settings on people.

There is nothing wrong with shortcuts as long as the "long cut" or original/traditiona/expected way of doing something is also available.

 

[digitalwanderer]

That's not my problem. If I can get him to change his mind or admit he's wrong about one single thing, then I'll consider my life a success

"A man's reach should exceed his grasp, or whats a heaven for?"

 

[radar1200gs]

Makes you wonder if Greg ever railed against Quincunx AA. Somehow, I doubt it.

It was a little blurry on my GF3, but I quite liked it on my Gf4. Don't use it anymore though.

 

[OpenGL guy]

This is like so many other issues I have argued on for the past two years.

Yes, gamma corrected AA can be an IQ improvement. However it is not a standard part of the API and should be strictly optional (just like all other optimizations).

Provide all the optimizations you like. I don't mind, just make them optional and let the user or the app decide (user decides through app profiles/cp, app decides through default or app preference setting). Don't force settings on people.

There is nothing wrong with shortcuts as long as the "long cut" or original/traditiona/expected way of doing something is also available.

So you're against shader optimizers, early Z rejection, Z compression, color compression, etc.? What a strange person. No one is going to allow such controls because it's simply not necessary.

None of these reasons you stated above is a good reason to allow people to disable gamma corrected AA. The fact that it's transparent to the application is good. The fact that it improves image quality is good for the end-user.

Maybe you should rail on your favorite IHV for not offering image quality improvements like gamma corrected AA.