Anistropic Debate

Hmmmm.. I've been using the R200 for a few months now, and its anisotropic solution can be just as counter productive as it is productive.

Mip Map lines are soooooooooo 90's :LOL:


When I had my GF3 I just left 64tap turned on constantly, and adjusted resolution according to performance. The performance hit is definatly greater, but to my eyes the IQ difference is night and day most of the time.

Oh BTW I noticed halfway through reading the benchmarks the reviewer (who I assume is just a simple gamer) said this...

"GF4 only...ATI card sold"


:devilish:
 
Livecoma said:
Hmmmm.. I've been using the R200 for a few months now, and its anisotropic solution can be just as counter productive as it is productive.
Click to expand...

Huh? What do you mean by counter productive?

Regards, LeStoffer
 
I suspect he meant that the noise/banding/error (call it what you will) is disturbing to the point of <fill in the blank>.

I wouldn't go quite that far, but there is a definite difference....Anybody who claims that they're on par with one another is simply not being honest.

I would believe the guy who says, "You know...I can see what you're talking about...But honestly...For me, it's just not that big of a deal...and it certainly is worth the minimal performance drop..."

...as opposed to the other guy who swears up and down, "I cannot tell a single bit of difference..."
 
Well as a owner of both cards also I never used 64 Tap Anistropic on our Geforce 3 as the peformance hit was big, especially games like Unreal Tournament. We ran with 32 tap or nothing at all.
Ever since this debate started a couple of days ago I fired up every game I had and started looking, and I mean really looking..
The worst game I saw some mip map borders was some Quake 3 maps, the least was Madden 2002 and Nascar 4. I could not see artifacts on 95% of the games.
I don't want to this to turn into a mine is better than yours debate as it always ends up in a war, I wanted to see what some of the more knowlegeable members here thought about future implementations of anistropic filtering.

:-?
 
I agree, using anisotropic filtering is like night and day in games, at least with NVIDIA's implementation. Not so sure about ATI's ripmapping. Personally just like playing games in 800x600, I'd rather stick bamboo shoots under my fingernails than play games with mipmap lines in them.

When I had the R8500, I did like how it made a few games look better with anisotropic filtering, such as in Madden NFL 2002 where the far end of the field was a little clearer (lines were line across the field) as opposed with the GF3 I had at the time. That was one of the few game where it looked noticably better to me.

Edit: Typos.
 
LeStoffer said:
Livecoma said:
Hmmmm.. I've been using the R200 for a few months now, and its anisotropic solution can be just as counter productive as it is productive.
Click to expand...

Huh? What do you mean by counter productive?

Regards, LeStoffer
Click to expand...

What Typedef said pretty much.

The mipmap lines and all the additional minor to moderate filtering artifacts I've observed while playing games. If I use one word to describe it I would say it is "dirty". Well not super dirty, but defiantly not clean.

NVIDIA's anisotropic filtering so far has suited my taste better then ATi's. That and perhaps the games I play are better geared towards NVIDIA cards. YMMV

Aside from all that, what have we always equated a lower performance hit for a particular feature with? Defiantly not BETTER IQ.
 
Fred said:
The difference Mephisto, is that you don't lose any image quality with multisampling compared to supersampling as long as you are using anisotropic filtering. (well assuming you neglect or precheck alpha textures)
Click to expand...

Well that's not strictly true. Perhaps I'm being a bit picky, but it's quite feasible, for example, that a polygon edge could run along a region of high rate of change in the texture. Ignoring the visibility information in the texture filter would thus give the wrong result.
 
Just a quick question regards the AF on GF3/4:

What happens if you use, say, AF at 2x, and decrease (or is it increase) the LOD? Textures would be sharper, but would there also be a performance hit like in 4x or 8x?

It's a hack alright, but maybe worth it for some...

http://www.digit-life.com/articles/gf4/index6.html

Edit: Added link to digit-life who tried it but didn't benchmark the LOD change.
 
There would be a performance hit, and you'd introduce texture aliasing... which kinda defeats half the purpose of anisotropic filtering in the first place.
 
Fred said:
Adding silicon for ripmapping say on an Nvidia part would be ... questionable. Is the added heat/power cost/clock reduction/die space lost really worth it?
Click to expand...

Would there even need to be any 'added silicon'? AFAIK Rip-mapping is basically using mip-mapping slightly differently - if you have hardware capable of automatically generating mip-maps then you could probably twist them to create the rip-maps(?).
 
ATI banding problem

Could someone give me a screenshot which shows the ATI 8500 aniso banding problem at it's worth and a similar GF3/4 aniso shot?

Thanks (that review has a lot of shots!)
 
PC Gamers took a screenshot of what they percived as the problem
PC%20GAMERS.jpg


I tried to reproduce it but I couldn't
ME.jpg
 
Matt said:
I agree, using anisotropic filtering is like night and day in games, at least with NVIDIA's implementation. Not so sure about ATI's ripmapping. Personally just like playing games in 800x600, I'd rather stick bamboo shoots under my fingernails than play games with mipmap lines in them.
Click to expand...

Err Anisotropic filtering shouldn't be necessary to remove MIP map boundary lines from the texturing - Trilinear should be sufficient for that. It sounds more like the way the particular filtering is implemented.

Aniso' should improve the image where the projection of the texture causes it to be 'squashed' unevenly.
 
Ante ... try using the same resolution of the window and not downscaling the picture. If you include a picture make sure its one for which we dont have to believe you on your word :) If I downscale the PC gamer picture it looks fine too.
 
Ante P,

I took that image right when the program was released using the first set of beta drivers for the ATI 8500. As soon as the angle changes to near 45 like in that image, rip maping will drop back into bi-linear only. There was a nice thread on the old board in which Joe, Dave and others were able to orignally post this. I just verified it and upload it for my own benifit. Of course I have not re-looked at that with newer drivers so it may not be as bad now....

I also have that mirrored:
http://www.pc-gamers.net/jb/reviews/ati8500/mipmap.zip

For me I can see the mip lines, but in the games I play they are not that big of an eye sore. Just my own tastes...
 
Using the 9021 (Radeon 8500) drivers for WinME and with the settings listed above in full screen:

Flat
flat.jpg



45 degree rotation
45degree.jpg



Sorry for the picture size but it's the best way to see this!
 
Thomas,

I tried getting the road to rotate with WinXP ICD 6043 Rev 2 and unless my computer is so fast the and road is rotating at Warp 10 its :

1) Not working in WinXP
2) I still haven't woke up

:p
 
You must log in or register to reply here.

Similar threads

T
  • Locked
New Gamebryo LightSpeed engine for Wii
Replies
3
Views
3K
Shifty Geezer
Shifty Geezer
R
id Software releases Open-Source Wolf3D iPhone port
Replies
13
Views
5K
Ailuros
Ailuros
C
Chat Transcript: ATI's texture filtering algorithms
3 4 5
Replies
89
Views
14K
radar1200gs
R
W
Have you guys seen this?...
Replies
5
Views
2K
Hanners
H
G
Excellent article on NVidia and ATi Optimisations
3 4 5
Replies
90
Views
11K
Quitch
Q
Back
Top