Radeon 8500 ansiotropic filtering stuff..

I wonder if Firing Squad bothered to read the whole thing.

This quote is from that Digit-Life article:

I must say that on the whole the ATI's anisotropy is better in quality than the NVIDIA's one, in particular, in some games on the Unreal engine. There are a great deal of games where the difference is hardly noticeable, which is a great advantage of the RADEON 8500 taking into account its modest appetite.

So, what can we say about anisotropic filtering of the RADEON 8500? Some will consider these drawbacks unimportant - its cost is more significant for them. Some get irritated at such imperfections. They value quality much more than speed. Well, everyone chooses what he wants.
Click to expand...

It sounds to me that they actually like ATis implementation better because small IQ improvement of Nvidia's method does not justify the huge performance penatly. In other words, Digit Life thins that a little more IQ is not worth a a much bigger performance hit.
 
Further correct me if I am wrong here, but where nvidias Multisampling only does a portion of the work that proper FSAA is supposed to do couldn't you draw the same sort of conclusion about nvidias methods of AA? Wasn't Firing Squad one of the sites that instigated the Quake/Quack feasco last fall during the Radeon 8500 launch. I wonder if this is going to be some sort of repetition of history with the R300 launch? ATI at this point deserves better recognition then that IMO.
 
This is annoyingly ridiculous. :devilish:

And people wonder why sites like FiringSquad who helped "expose" the Quake/Quack thing get accused of bias.

First, none of these sites seem to take ANY notice of the 3DMark "splash screen" thing...and now some are attempting to turn a feature that is implemented differently on different hardware, and relate it to cheating?!

Interestingly, I have not seen FiringSquad's expose on how nVidia cheats with FSAA!! Edge only?! Doesn't even touch the textures! That must be a "another case of nVidia driver teams cutting corners when it comes to quality so that they can get better scores in benchmarks."

Sigh....
 
Rancidm said:
Check out the bottom of that article...

http://www.digit-life.com/articles/anyr2002/anyr2005.html
Click to expand...
The guys at digit-life need to do their homework...
"Again no anisotropic traces of the RADEON 8500 on some surfaces; besides, some almost horizontal surfaces have bugs (because of lacking trilinear filtering with the quite aggressive LOD BIAS)."

I thought everyone knew that the 8500 couldn't do trilinear (i.e. mip linear) filtering with anisotropic filtering.
 
pxc said:
Good article at Digit-Life. It sums up the glitches pretty nicely.
Click to expand...

Funny though as you read the article it appears that he is about to trash ATI but in the end he clearly says that ATIs method of AF is considerably better. If you didn't read his conclusion you would think that he was going to say the reverse. Funny that.
 
Geek_2002 said:
Funny though as you read the article it appears that he is about to trash ATI but in the end he clearly says that ATIs method of AF is considerably better. If you didn't read his conclusion you would think that he was going to say the reverse. Funny that.
Click to expand...

That's because he focuses on quality first, and then looks at it as a whole (quality + performance).
 
Doomtrooper said:
OMG here we go again, the internet truly is turning into nothing but a biased, uniformed piece of trash. o_O
Click to expand...

What did you expect a bunch of 16 year old humans to be like? Honest?
 
Geeforcer said:
Geek_2002 said:
Funny though as you read the article it appears that he is about to trash ATI but in the end he clearly says that ATIs method of AF is considerably better. If you didn't read his conclusion you would think that he was going to say the reverse. Funny that.
Click to expand...

That's because he focuses on quality first, and then looks at it as a whole (quality + performance).
Click to expand...

The reviewer used out of the three benchmarks he used two of the Radeon 8500s worst performing benches. The UT2003 bench is beta software and isn't automatically an accurate bench at all as it really isn't necessarily indicative of how the Radeon 8500 will perform on the final software. The review seems to make the Radeon 8500 look like a poor card really. Additionally the reviewer pokes fun at the hardware by portraying them in a tasteless light with the rasberries and temples and whatnot.
 
It looks like someone confused what has been said HERE about nVidia cards, with Tom's (Lars') incoherent comments on ATi's anisotropic filtering, and blindly and randomly pasted incomplete comprehension into some distorted mess. I think Reverend was psychic when he decided to take a break, as new cards being released are bringing out the worst in websites in full force, and should invest in some lottery tickets. :p
 
I'm glad Ati took the time to respond straight out this time.

I agree with Ati on this one. People are complaining the Radeon doesn't apply ansio to the whole scene. I say, why? If ansio doesn't need to be applied to certain parts of the image, it shouldn't be. That's just a waste of resources. It doesn't make it any less "correct." And it certainly isn't a hack or driver cheat...

As for the 45 degree angle issue, I think it's an acceptable quirk/downside to the implementation. The view doesn't shift to that exact angle too often in most games. The only time I can think of where this might be a pervasive problem is in flight sims.

IMO, Toms hardware is flat out wrong to be ignoring Ati's ansio in their comparisons. If it would different if Ati's implimentation was subpar, but it's arguably the best looking and performing version out there. To pass it off as driver tricks and too difficult to look into is a bit unprofessional.
---------------------------------
"we won't be including ATI's Radeon 8500, because its anisotropic filtering, which involves various driver tricks, is somewhat controversial and would make a comparison unnecessarily complicated."
----------------------------------

Ati's ansio implementation is a lot like Matrox's FAA. It looks good, it's performs good, and it doesn't work 100% of the time.
 
The ignorance in the replies to that story over at firingsquad astounds me. The world just got a little scarier for me realizing that some of these people will actually produce offspring.
 
I've always tried to explain it to people like this:

Anisotropic filtering is Anisotropic filtering

The Radeon 8500 does it as well as the GF3/4, they both do Anosotropic Filtering.

What is different is the way it's implemented, NVIDIA has a 'render all' method which does anisotropy on everything. ATI has a 'render only what's needed' method which only does anisotropic on textures that are needed and seen. This accounts for their fast implementation of Anisotropic. In THEORY there are technical flaws to ATI's method, but in PRACTICE they are not seen (as ATI has pointed out) there is no degradation of image quality.

If you are rendering a lot of 45 degree angles that need anisotropy then by far the GF3/4 will be better suited for that, but if your a gamer, who just wants the best image quality and speed the 8500's method of Anisotropic is better.

Now of course it's always better to judge for yourself, if you can do a side by side in person then you will be able to choose the best option for you.

Sometimes that works, other times the people are too stubborn to see it and stick to their 'brand loyalty' heh.
 
Brent,

It's almost exactly the other way around if you compare ATI's with NV's AA aproach.

Combine aniso with FSAA on both and you're not too far away from an equation on competitive cards.

Imagine for a second how a Radeon would perform with Smoothvision and NV's anisotropic algorithm. :eek:

If one approaches it like that then it was completely reasonable for ATI to implement an algorithm like that.
 
:cry:

It does not matter.. I just read through the responses at firing squad.. the whole lot of them are talking trash about Ati.. Saying this is another "quack" that ati is just using another cheap hack... People are calling for Ati to leave the industry.. on and on and much much worse..

You deliver a technology that is faster, delivers great IQ and offers 16 samples, 8 more than anyone else.. and somehow you are a cheap hack who has cheated everyone..

This is really sad.
 
Brent said:
I've always tried to explain it to people like this:

Anisotropic filtering is Anisotropic filtering

The Radeon 8500 does it as well as the GF3/4, they both do Anosotropic Filtering.

What is different is the way it's implemented, NVIDIA has a 'render all' method which does anisotropy on everything.
Click to expand...

Of course, this isn't true...
nVidia cards determine the anisotropic ratio and applies the degree of anisotropic required for the given portion of the surface - including not applying aniotropic at all.

Actually the only card I've seen that's not smart enough for even this is the KyroII.

ATI has a 'render only what's needed' method which only does anisotropic on textures that are needed and seen. This accounts for their fast implementation of Anisotropic.
Click to expand...

No it isn't.
ATI has an anisotropic implementation that doesn't take a fillrate hit. If there's an decrase of speed, it's because the higher bandwidth requirement, but it's mostly compensated by not performing trilinear.

In THEORY there are technical flaws to ATI's method, but in PRACTICE they are not seen (as ATI has pointed out) there is no degradation of image quality.
Click to expand...

Yes it can be seen.
The lack of trilinear is the most annoying thing. It creates "false edges" which are distracting. It also creates way too much shimmering.

If you are rendering a lot of 45 degree angles that need anisotropy then by far the GF3/4 will be better suited for that,
Click to expand...

Anisotropic is not magicigally disappearing at 45 degree. If you rotate the surface the degree of anisotropy will constantly decrease until at 45 degree it's completely disabled.

but if your a gamer, who just wants the best image quality and speed the 8500's method of Anisotropic is better.
Click to expand...

I'm a gamer and I found the 8500's method worse than disabling anisotropic completely....

Now of course it's always better to judge for yourself, if you can do a side by side in person then you will be able to choose the best option for you.
Click to expand...

True.

Sometimes that works, other times the people are too stubborn to see it and stick to their 'brand loyalty' heh.
Click to expand...

Until the GF3 arrived the best quality anisotropic filtering was present in G400. It had 4x with trilinear, correctly implemented.
It was unplayable though. :(

I was very excited about Parhelia only finding out that it only has 2x anisotropy and takes a performance penalty even for that. Some companies just never learn...

I'm still searching the solution that satisfies me, unfortunately it's yet to be announced. (And I don't know when.)
 
You must log in or register to reply here.

Similar threads

R
Digit life on Parhelia
Replies
15
Views
3K
nggalai
N
Back
Top