ATI Filtering Optimisation - Genius or Disingenuous

[pocketmoon_]

It has to be done

[edit]

Perhaps ignoring the fact that the PR appears to have messed up on this, from a gamers point of view, how do you feel ?

 

[Dave Baumann]

You might want to make it clear whether you are talking about the technicals or how it has come across....

 

[mjtdevries]

Another option might be that you can be talking from the consumer standpoint (no visible IQ loss, better performance) instead of from a technical or PR standpoint.

 

[Diplo]

IF the optimisations improve performance without any noticeable degredation in image quality then I'm all for it. How could any gamer not be? (By noticeable I mean with the human-eye and not using high-contrast enlargements that are a few pixels different.)

I guess the trouble occurs where the optimisation is always enabled via hardware and it introduces a small, but noticeable, difference in IQ. In these cases I'm all for it so long as it can be controlled by the user/application and they are made aware of it.

 

[bloodbob]

Not disclosing it is very bad.
Though as far as technically its not really that brilliant there was dithered trilinear before this and actually did all the screen as opposed to about 1/2 the screen and dithered trilinear had a bigger preformance saving ( 2/3's of bandwidth of brilinear ) if you ignore any texture caching.

 

[DemoCoder]

I think there should be an option for "nice trick".

Anyway, how do people define noticable? Angle-dependent AF was definately noticable in screenshots, tests, and even visible to the naked eye in-game if you knew what to look for, but ATI was given a free pass by the community for it, because it wasn't noticable enough.

I happen to think there are many circumstances were brilinear is also "not very noticable", not NVidia was dragged through the mud for not making it an option.

Now we have the situation where ATI has provided a method by which it will dynamically switch on/off true-trilinear, on a per-texture basis I guess. (and if believed, patented it, I hope, only the specific image analysis technique used, not the general idea of "dynamically enabling/disabling it" which is a pretty obvious technique IMHO)

I think this feature is cool, but I would like the ability to control it. Given that there is no api to control it (via the APP), it should atleast be a control panel option. Why? Because for some scenarios, I might really want, absolutely, 100% real-trilinear (especially as a developer testing some stuff), and because there may be scenarios where it does cause strange bugs (hell, AF broke some games). Thus, it's always nice to have the option.

 

[DSC]

A cheat is a cheat. It's not applying full trilinear as the application/control panel requests, but their reviewer's guide has the gall to suggest the competitor's card be set to full trilinear.

 

[pino]

A cheat is a cheat. It's not applying full trilinear as the application/control panel requests, but their reviewer's guide has the gall suggests the competitor's card be set to full trilinear.

Amen.

 

[madshi]

For me it's too early to decide. Let's wait a few days/weeks to see whether any noticable IQ problems will be found.

The idea is not bad (if it really works as advertised), but ATI should have documented it. Also I'd like to be able to turn it off - at least via a registry hack.

 

[christoph]

technically i think there is still too much to be discovered/cleared up before one can judge. but then hey they can add whatever they want to as long as its an option the user is informed and has control over.

the way this came across is dissapointing to me at least

 

[Evildeus]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

 

[jb]

I think there should be an option for "nice trick".

Anyway, how do people define noticable? Angle-dependent AF was definately noticable in screenshots, tests, and even visible to the naked eye in-game if you knew what to look for, but ATI was given a free pass by the community for it, because it wasn't noticable enough.

I happen to think there are many circumstances were brilinear is also "not very noticable", not NVidia was dragged through the mud for not making it an option.

Not wanting to start something with DC....

However they are night and day differences. The Angle Dependent AF was a hardware "limitation" built into the hardware from day one. We knew during the first reveiws that this was the case and there was nothing we could do about it. NV Brilinear was something forced on us after the cards were shipping and it was not a built in limitation. And we could do something about it (just not upgrade drivers). I am not saying either is right or wrong. Just they are two different things and thus should be treated differently.

 

[christoph]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

ppl are working on it.....its only a matter of time

 

[jb]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

If you can not see it, nor tell any IQ loss. Then does it matter?

 

[christoph]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

If you can not see it, nor tell any IQ loss. Then does it matter?

who concluded you cant? i dont think theres a consense as of now

 

[Evildeus]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

If you can not see it, nor tell any IQ loss. Then does it matter?

Of course not, but if it's not the case, doesn't it matter? If you don't have the methodology, how can you verify if it's true or not.

christoph, yeah, i hope it won't take to much time.

 

[Wunderchu]

a bit more discussion on the topic, here:

http://www.anandtech.com/news/shownews.html?i=22196





... oh, and BTW, in case any of you haven't heard of this news yet:

http://www.ati.com/online/chat/

(source: http://www.beyond3d.com/forum/viewtopic.php?t=12486&postdays=0&postorder=asc&start=700 )

 

[Bjorn]

The Angle Dependent AF was a hardware "limitation" built into the hardware from day one. We knew during the first reveiws that this was the case and there was nothing we could do about it.

Does it matter if it's a hardware or software limitation ?

And i don't remember that the first reviews of the 8500 contained info about AF + bilinear and the angle problems. I certainly don't remember that Ati had info about that on their homepage. But i might be wrong.

 

[Geo]

Or?

And.

 

[mjtdevries]

I think the most problematic right now, is to find a correct methodology to verify the quality of this filtering.

If you can not see it, nor tell any IQ loss. Then does it matter?

who concluded you cant? i dont think theres a consense as of now

There is a 37 page discussion with lots of people angry at ATI, who are doing their very best to find some IQ loss, so they can accuse ATI of cheating. But there is still nobody that has found and IQ loss, and some who claim better IQ.
What conclusion can you draw from that?