SMOOTHVISION Not All It's Cracked Up to Be

[AGodspeed]

http://www.anandtech.com/video/showdoc.html?i=1562&p=1

This more thorough look at AA quality and performance raised some very interesting questions; at the top of our list is why ATI chose to go with supersampling AA for the Radeon 8500 and not a much more efficient multisampling implementation such as that found in NVIDIA's GeForce3. There are a couple of major reasons why ATI would make this decision:

1) Better texture quality without specifically enabling anisotropic filtering
2) Much easier to implement than multisampling

Interesting stuff. Anyone want to comment who really knows this stuff.

 

[flexy]

well..don't know if i 'really know the stuff'.....but as far as it concerns 2x quality AA i am totally satisfied with the performance and IQ of the radeon.
( 4 x AA is another subject

i have a question, tho...

Anand tested under OpenGL only ? How does it look under D3D ? This would be interesting, too !!!!!!!!!

And...what if some (in ADDITION) adds anisotropic filtering also on the radeon ? What happens then ?

I mean...he basically says that 2 x quality AA on the radeon EQUALS (even tops!) quincunx AND aniso on the gerforce 3, this is astonishing assuming that the radeon gets the same (or better) quality without even using it's aniso !!!

But what happens if you enable aniso, too, on the radeon ? Even much better picture quality ? Or a significant performance loss ?
Maybe you dont even need antialiasing if you have aniso on...and so on..you get the idea...

(aniso is [T.M.K.] used only to 'clarify' textures which have been blurred by antialiasing before...)

He didnt mention aniso at all for the radeon...i think there might be still huge room for comparisons ?

(And as said....no word about D3D.....)

 

[Demonicon]

Anandtech reviews leave something? Or am I confuzzed

The first article was entirely misleading to say the least. (Through no fault of Anandtech, as many other sites had the same problems)

The second article is confusing to me. I am wondering why they did not enable the aniso filtering on the ATI to get a more equal comparison to the aniso filtered GeForce3?

 

[Adul]

ahem you are the weakest link!


actualy i posted 2 hours before you.

http://forums.anandtech.com/messageview.cfm?catid=27&threadid=637276

 

[Katana]

<< The second article is confusing to me. I am wondering why they did not enable the aniso filtering on the ATI to get a more equal comparison to the aniso filtered GeForce3? >>



From the article;



<< Because of the GeForce3's use of multisampling it needs to have anisotropic filtering enabled to level the playing field, this also reduces performance a bit as you will see in the performance section of this review. >>

 

[Demonicon]

<< Because of the GeForce3's use of multisampling it needs to have anisotropic filtering enabled to level the playing field, this also reduces performance a bit as you will see in the performance section of this review. >>

[/i] >>



I understand that the GeForce3 HAS to have anisotropic filtering enabled to level the field, however, if this is supposed to be a COMPARISON, shouldn't the ATI card be tested with anisotropic filtering enabled also? Or am I missing something.

 

[imgod2u]

In that review, Anand compared the quality of the Radeon 8500 to that of a Geforce 2, and it was almost identical without anistropic filtering. So, really, it's not a matter of ATI having better AA quality, it's just that the Geforce 3 currently uses a lower quality AA method (multisampling). It's quite capable of using the old supersampling method, and hence, matching the AA quality of the Radeon, however, nVidia, for some reason, hasn't enabled that in the drivers.
That asside, the arguement that the anistropic filtering should be added to the Radeon too in a comparison, I guess that would be a nice addition. But the point was to see how both cards perform while providing the same image quality. The Geforce 3 can provide the image quality of the Radeon 8500 and perform better. However, note that that was using the lowest level anistropic filtering on the Geforce 3. And, I believe, the Geforce 3 Ti are capable of 128-tap anistropic filtering, while the Radeons are capable of 64? So theoretically, we can play the up the image quality on one card and then upping it on the other to balance, but when we turn it up to max, the Geforce cards are still capable of more anistropic filtering and we get the same results, even though the Geforce 3 needs a better filtering method to catch up to the Radeon, it can afford to due to it being more powerful.

 

[AA0]

the problem with your arguement is the Ti isn't more powerful than the 8500, it only seems faster in benchmarks because its default IQ settings are lower than the 8500. And when you start to get to 64 and 128 tap, the performance hit isn't small like 16tap, it is massive. So moving up one more notch about the 8500 at the higher qualities will not work for the Ti500. Just on the subject of anisotropic filtering alone the 8500 can produce a similar image (slightly inferior) at its lowest setting, that the Ti500 can at 64, the only different being a 7fps penalty (8500) and a 40 fps penalty for the Ti500. Turning anisotropic filtering on the 8500 in the reviews I've seen looks great, and there is no way I would turn it off in my games.

Anand should have enabled anisotropic filtering for the 8500 than compared the Ti500 to that, and if need be, increase the Ti500 settings to match the radeon to see what happens. To not do that is just plan biased.

 

[wfbberzerker]

repost

 

[flexy]

<< the problem with your arguement is the Ti isn't more powerful than the 8500, it only seems faster in benchmarks because its default IQ settings are lower than the 8500. And when you start to get to 64 and 128 tap, the performance hit isn't small like 16tap, it is massive. So moving up one more notch about the 8500 at the higher qualities will not work for the Ti500. Just on the subject of anisotropic filtering alone the 8500 can produce a similar image (slightly inferior) at its lowest setting, that the Ti500 can at 64, the only different being a 7fps penalty (8500) and a 40 fps penalty for the Ti500. Turning anisotropic filtering on the 8500 in the reviews I've seen looks great, and there is no way I would turn it off in my games.

Anand should have enabled anisotropic filtering for the 8500 than compared the Ti500 to that, and if need be, increase the Ti500 settings to match the radeon to see what happens. To not do that is just plan biased. >>



that basically was my main decision factor to go with the radeon !

Since i of course want a GREAT picture combined with great FPS...and 64tab on the gf3 would render it (more or less) useless......except you would 'accept' a 70% loss of fps in the geforce but instead having a 'nice' picture...lol....ho hardcore gamer would do that....i want speed

And..i really do not think that the gf3 would do 128tab....even the 64tab mode is only possible by hacking the registry, using a tool...since it slows already using 64tab..

The ati uses this dynamic algo (according to ati)...it uses NOT only 16tab...but it uses either no aniso, 16tab, 32tab or 64tab dynamically which is much more efficient.

Btw..show me the number of people in REAL LIFE in front of their computers, playing their games...with Geforce3 aniso tab64 registry hack ON....great quality...but LOUSY fps.....and show me the number of people with a Radeon...the SAME quality picture and neclectible loss of performance. In case of the radeon its no big deal to turn on aniso as DEFAULT and *knowing* it only costs a few fps !

(And thats [i think] no nonsense....this facts come from reviews from tom and or rivastation....no matter what one might think...but if they reviewed it and state the above facts and measued fps....i dont have a reason not to believe then. Result: Radeon if you want acceptable performance AND great IQ !

 

[E3Art]

Im glad you all responded to this post. I just bought a radeon 8500(havent got it yet) and was starting to reconsider keeping it, but thanks to you guys i feel a whole lot better about the card now. Its so hard to find out which one is better because everyone is biased it all boiles down to personal preferance.

keep up the good post's

 

[imgod2u]

I haven't seen the numbers for the Radeon 8500 with anisotropic filtering (either 32 or 64-tap) on compared to the numbers in the Ti 500. On Anand's review, the numbers for the Ti 500 with a low level anisotropic filtering on were higher than the 8500's with anisotropic filtering off. If the 8500 does suffer less penalty for upping anisotropic filtering than the Ti 500, that would give it an advantage, however, I haven't seen numbers for that yet.