PDA

View Full Version : "Trilinear" Filtering in Motion


Pages : [1] 2 3 4

MikeC
08-01-03, 07:30 PM
I've posted two 5MB video clips.

One is from a GeForce FX 5900 Ultra using the 44.03 Detonator drivers:

http://www.nvnews.net/temp/geforce_fx_5900_ultra.png

The other is from a Radeon 9800 Pro using the 3.6 Catalyst drivers:

http://www.nvnews.net/temp/radeon_9800_pro.png

Video was taken from UT2003 at 800x600 using FRAPS and compressed in Microsoft's MPEG4 format.

Download the Codec here for Windows Media Player -
http://www.cabcconnection.com/support/CodecDownLoad.asp


http://www.nvnews.net/temp/1.avi

http://www.nvnews.net/temp/2.avi


UseTrilinear=True and LevelOfAnisotropy=1 in UT2003.ini file.

MoveForward speed in the User.ini file was changed from the default of +300 to +1 when capturing the video.

-=DVS=-
08-01-03, 08:20 PM
First of all images are NOT identical gun is looking slightly in different location at start and throu out the test buts its ok

Second very hard to see any differences would need bigger resolution , both looks ok.


EDIT: looked at it from zoom first file <1> Looks better , what card is it Rad or Fx ?

MikeC
08-01-03, 08:38 PM
Originally posted by -=DVS=-
Second very hard to see any differences would need bigger resolution , both looks ok.

Yes, you have to have a keen eye to spot the differences, but they are there :)

Hint: Look at the "smoothness" of the mipmap transitions between the textures on the ground. Imagine that there's a horizontal line across the video that passes through the cross hair.

Also save both videos to your PC and play them back one after the other multiple times. The most important tip is to stay focused on the ficticious horizontal line as the crosshair moves. If you're still having trouble, decrease the playback speed.

Kruno
08-01-03, 08:43 PM
Mike, maybe if you had a high quality 200-500MB video we would be able to spot the difference. :)

StealthHawk
08-01-03, 08:49 PM
Video 1 looks better to me. Clear mipmap levels in video 2. Viewed each video once.

SH64
08-01-03, 08:49 PM
Cant see a difference .. both good .

MikeC
08-01-03, 08:52 PM
Originally posted by K.I.L.E.R
Mike, maybe if you had a high quality 200-500MB video we would be able to spot the difference. :)

Heh. They were about 100MB before I compressed them. The quality of the compressed videos is very good though.

Maybe if I smacked your ass a couple of times, you could see the difference. Some of the folks at B3D were able to point them out :)

reever2
08-01-03, 08:56 PM
Video 1 looks better

MikeC
08-01-03, 09:03 PM
Originally posted by SH64
Cant see a difference .. both good .

I've noticed that some people say they can't tell a difference even when I changed the forward speed from 300 to 1. Plus, the video was recorded at 30 frames per second.

Malfunction
08-01-03, 09:03 PM
Ya, have to agree there... quality looked about the same but at the horizon, vid 1 looks alot better.

saturnotaku
08-01-03, 09:09 PM
Boy won't there be egg on a whole lot of people's faces if it is in fact the 5900 that folks are saying looks better. :eek:

I'm not gonna vote, though. ;)

-=DVS=-
08-01-03, 09:13 PM
Originally posted by saturnotaku
Boy won't there be egg on a whole lot of people's faces if it is in fact the 5900 that folks are saying looks better. :eek:

I'm not gonna vote, though. ;)

Well difference is small hard to see realy my vote goes on 1 even if it is GeforceFx ( but in case it is you need to use aplication setting to get such results ?)
Now second question would be what FPS were you getting on both ? :p

And why didn't MiceC made comparision without aplication settings , make them on Quality settings in Control panel with 8xAF maybe with 4xAA also , that is how moust people play :rolleyes:

creedamd
08-01-03, 09:52 PM
1 looks better, much better of a blend transistion. You can see the line clearly on #2.

SH64
08-01-03, 09:56 PM
Originally posted by creedamd
1 looks better, much better of a blend transistion. You can see the line clearly on #2.

Yup ! i agree with that ... finally my retard eyes noticed a difference .

now can we conclued that vid 1 is the 5900 ?? :D

Malfunction
08-01-03, 09:59 PM
Originally posted by saturnotaku
Boy won't there be egg on a whole lot of people's faces if it is in fact the 5900 that folks are saying looks better. :eek:

I'm not gonna vote, though. ;)

I just woke up, dl'd the vids and clearly saw the horizon blurr...lol. I guess it is easy for me to see since I get that in Performance mode/0xAA/AF/Mip Map Detail set to "Blend" in BF1942 w/High Quality Game Settings.

However with Quality Mode and settings set the same, the horizon is very well hidden or should I say *blended very well and makes it difficult to locate.

Vid one looks like it is on Quality Mode using Det 44.67's

Vid two looks like it is on Performance Mode using Det 44.67's

Will be interesting if Vid 2 is on ATi card. :) If not... uh oh...lol. :p

Peace,

:)

*EDIT- I fixed what I said... I am still waking up..lol whoops. Vid one = Quality mode/ vid two = Performance mode...lol. Sorry for the confusion.

Malfunction
08-01-03, 10:12 PM
Oops...lol. :)

SH64
08-01-03, 10:17 PM
Originally posted by Malfunction

Vid one looks like it is on Quality Mode using Det 44.67's

Vid two looks like it is on Performance Mode using Det 44.67's



now THATS more like it ;)

walkndude
08-01-03, 10:21 PM
Here's what seems a bit sketchy to me as far as all the ut comparisons lately...

With ati cp settings set to application preference the user should be able to adjust af in the .ini and the card will filter the texture stages as requested by the .exe. Well if the .ini has trilinear set to true and "level of anistropy" set to 1 then the image being rendered is going to be trilinear(true AFAIK) filtered AND will have some degree of anistropic filtering applied.

At this time nV does not have an "application" setting for anistropic filtering in the control so the degrees of anistropic filtering are forced by the driver and NOT controlled by the application as per ATI... meaning with the same .ini settings the rendered image will only be trilinear filtered with NO anistropic filtering(its been forced off in the cp).

Not an apples to apples comparison in the least.

Has anybody taken the time to compare images (both with and without highlighted mipmaps) of either card with LevelOfAnistropy set to both 0 and 1 inorder to make sure it isnt having an impact when trying to compare trilnear filtering?

Carbon Unit
08-01-03, 10:56 PM
I can't tell any difference between those 2 videos, and I have been playing UT2003 and all the full settings with no problems, all the frames look great, its a major leap over mt Ti4200....just wait till the DX9 games come out and Nvidia's Ultra shadow technology kicks in

CaptNKILL
08-01-03, 11:15 PM
Originally posted by Carbon Unit
just wait till the DX9 games come out and Nvidia's Ultra shadow technology kicks in
And then the pixel shader performance drags it down :D

-=DVS=-
08-02-03, 12:16 AM
Originally posted by CaptNKILL
And then the pixel shader performance drags it down :D

:rofl

By the time real games come out we gonna have new cards for sure ;)

reever2
08-02-03, 01:19 AM
Originally posted by Carbon Unit
just wait till the DX9 games come out and Nvidia's Ultra shadow technology kicks in

Im sorry but what does ultrashadow have anything to do with dx9 games?

madshi
08-02-03, 03:16 AM
Originally posted by -=DVS=-
By the time real games come out we gonna have new cards for sure ;)
So you'll have a new card on 30. September? HL2 will be a very shader intensive game, when being played on a DX9 card...

Alan_HU
08-02-03, 07:31 AM
Clip 1 is clearly superior. Is it the GeForce or the Radeon?

volt
08-02-03, 07:40 AM
In 2nd vid, the transition is very noticeable. OMG if you guys can't see it :)