Beyond3D Forum

Ruined · 03-Oct-2004, 02:01

After reviewing some 3dmark05 screenshot comparisons by Driverheaven.net and Bjorn3d.com, it became evident that ATI may be sacrificing IQ for increased performance. Check out these comparisions:

3DMARK05 ATI X800 ARTIFACTS: FOG RENDERING

*** PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREA IN THE RED OUTLINE

NOTE the failure to render all of the fog in the ATI screenshot

EXAMPLE 1
Nvidia GeForce 6800 Ultra 66.70 Beta
ATI X800XTPE 4.9
Reference Rasterizer image

EXAMPLE 2
Nvidia GeForce 6800 Ultra 66.81 Beta
ATI X800XTPE 8.07
Reference Rasterizer image

Source: Beyond3D.com & NvNews.net forums

*********

3DMARK05 ATI X800 ARTIFACTS: FILTERING/MOIRE

*** PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREAS IN THE RED OUTLINES

NOTE the obvious moire patterns in the ATI screenshots
Nvidia GeForce 6800 Ultra 66.51 Beta
ATI X800XTPE 4.9 w/ hotfix

Source: Bjorn3d.com

********

3DMARK05 ATI X800 ARTIFACTS: BANDING

*** PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREA IN THE RED OUTLINE

NOTE the increased banding in the sky in the ATI screenshots
Nvidia GeForce 6800 Ultra 66.51 Beta
ATI X800XTPE 8.07 Beta
Closeup of ATI's banding from above shot

Source: Driverheaven.net

hovz · 03-Oct-2004, 02:13

i see the moire, but i dont see any noticeable banding. that is theres an equal amount of banding on both

jvd · 03-Oct-2004, 02:17

Ruined , if i shut off ai it goes away , but my score doesn't change.

It seems to be the general optimization messing around with 2 areas .

Razor1 · 03-Oct-2004, 02:18

Interesting very similiar to Doom 3

Anyone got shots with 4.8 and 4.9 drivers?

The opts were off on these shots I think.

hovz · 03-Oct-2004, 02:19

what are ur specs and what do u score? no overclock

Razor1 · 03-Oct-2004, 02:21

Quote:

Originally Posted by hovz

what are ur specs and what do u score? no overclock

http://bjorn3d.com/read.php?pageID=1...;form_keyword=

Ruined · 03-Oct-2004, 02:29

Quote:

Originally Posted by jvd

Ruined , if i shut off ai it goes away , but my score doesn't change.

It seems to be the general optimization messing around with 2 areas .

According to the sites which posted those screenshots, AI & all optimizations were set to "off" for the ATI shots showing the artifacting.

ChrisRay · 03-Oct-2004, 02:57

Are you sure this is a result of ATI purposely sacraficing Image Quality for performance? I find it more likely this might just be a result of the internal precision differences between the 2 cards.

And what they are running in different spots.

jvd · 03-Oct-2004, 02:57

Quote:

Originally Posted by Ruined

Quote:

Originally Posted by jvd

Ruined , if i shut off ai it goes away , but my score doesn't change.

It seems to be the general optimization messing around with 2 areas .

According to the sites which posted those screenshots, AI & all optimizations were set to "off" for the ATI shots showing the artifacting.

Not in my case , I don't have a full version of 3dmark 2005 and it will take a very long time for me to line up the screen shots right . I'm sure another ati user who owns the card can do it . ALso the drivers I have don't need a code change or hack to shut off ai

Razor1 · 03-Oct-2004, 02:59

well they shouldn't be having this at x4 aa and x8 af.

Ostsol · 03-Oct-2004, 03:09

Quote:

Originally Posted by Ruined

NOTE the increased banding in the sky in the ATI screenshots
Nvidia GeForce 6800 Ultra 66.51 Beta
ATI X800XTPE 8.07 Beta
Closeup of ATI's banding from above shot

Well, there's clearly something odd in the first set of screenshots, but IMHO banding is equally evident above in both the ATI and NVidia screenshots. That it occurs at different places in the image is the only difference I could see and that's irrelevant.

jvd · 03-Oct-2004, 03:11

IT depends on how they put on 4x / 8x af . If they force it through the cp it wont work. , if they put it on through the ap it will .

Razor1 · 03-Oct-2004, 03:16

they turned it on through the ap

http://www.bjorn3d.com/read.php?cID=702&pageID=1050

gordon · 03-Oct-2004, 03:21

Everything is fine with the r3xxs, must be x800 specific if true.

gokickrocks · 03-Oct-2004, 03:58

looked at the differences between the 2 images posted above...i agree with the moire pattern, but for the other red outlining, i would have to say the ati one is better, it seems more defined to me and if you look at the flooring texture, its clearer as well

i was a bit more curious so i went through the images in bjorn3d and found many differences in each image, which may need a reference rasterizer (i do realize that a ref. rast. may not produce 100% accurate image, but the differences between the cards is big, so at least it will help to determine which card is more correct) to compare against...there are some large lighting and some texture quality descrepencies between the 2 cards

(for those with firefox just open it in a new tab and use the mouse wheel to scroll)

comparison 1 (besides the gun powder from the cannon, take a look at the lighting on the front of the ship)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

comparison 2 (look at the flooring)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

comparison 3 (take a look at the water)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

edit: someone posted on the futuremark forums and compared between applying AA/AF through the CP and APP and found that when forced in the CP, the moire pattern is gone, and with the app, its there

hovz · 03-Oct-2004, 04:29

Quote:

Originally Posted by gokickrocks

looked at the differences between the 2 images posted above...i agree with the moire pattern, but for the other red outlining, i would have to say the ati one is better, it seems more defined to me and if you look at the flooring texture, its clearer as well

i was a bit more curious so i went through the images in bjorn3d and found many differences in each image, which may need a reference rasterizer (i do realize that a ref. rast. may not produce 100% accurate image, but the differences between the cards is big, so at least it will help to determine which card is more correct) to compare against...there are some large lighting and some texture quality descrepencies between the 2 cards

(for those with firefox just open it in a new tab and use the mouse wheel to scroll)

comparison 1 (besides the gun powder from the cannon, take a look at the lighting on the front of the ship)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

comparison 2 (look at the flooring)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

comparison 3 (take a look at the water)...
http://www.bjorn3d.com/photos/showph...cat/503/page/2
http://www.bjorn3d.com/photos/showph...cat/503/page/1

edit: someone posted on the futuremark forums and compared between applying AA/AF through the CP and APP and found that when forced in the CP, the moire pattern is gone, and with the app, its there

besides the moire issue, it seems ati is producing a much better image. probably due to fp24 v fp16

Apple740 · 03-Oct-2004, 04:49

I took some tests with my X800XT. The moire in frame 985 is not there in the default game test, but appears when you set AF in the 3DMark05 CP (Pro version). But if you force AF by the Ati CP the moire dissapears.

Ruined, I would reather see that you make an issue of the fact that Nv uses DST by default in 3DMark05. Without DST, Nv scores decrease by 10/16%, so at this moment it's Nv who sacrifices IQ for speed.

hovz · 03-Oct-2004, 04:50

Quote:

Originally Posted by Apple740

I took some tests with my X800XT. The moire in frame 985 is not there in the default game test, but appears when you set AF in the 3DMark05 CP. But if you force AF by the Ati CP the moire dissapears.

Ruined, I would reather see that you make an issue by the fact that Nv uses DST by default in 3DMark05. Without DST, Nv scores decrease by 10/16%, so at this moment it's Nv who sacrifices IQ for speed.

whats the performance hit when you enable aa and af in the CP?

Sanctusx2 · 03-Oct-2004, 05:42

Good god, I see this post on every forum I frequent. Someone has been busy. =P

hovz · 03-Oct-2004, 06:00

well come on now, its ruined. what do you expect from chalnoths long lost brother?

Apple740 · 03-Oct-2004, 06:03

Screenshot X800XT with AF forced by Ati's CP:

http://members.chello.nl/h.appel1/X800.jpg

I think Ruined's case is closed.

Quote:

whats the performance hit when you enable aa and af in the CP?

Normally I get 5836 (no overclocks/high quality in CP), with 4xAA and 8xAF i get 4562.

RickCain · 03-Oct-2004, 06:36

Impressive he can actually post the same exact thread across multiple forums. L33t copy/paste skillz.

Funniest part is the nvidiots at [H]forum actually jumped on the bandwagon.

Cynips · 03-Oct-2004, 08:38

Quote:

Originally Posted by Apple740

Quote:

whats the performance hit when you enable aa and af in the CP?

Normally I get 5836 (no overclocks/high quality in CP), with 4xAA and 8xAF i get 4562.

Is the 5836 without any AA/AF? Would be interesting to know what the score is with AA/AF set in CP vs. set in the application.

MFZ · 03-Oct-2004, 11:50

Quote:

Originally Posted by RickCain

Impressive he can actually post the same exact thread across multiple forums. L33t copy/paste skillz.

Funniest part is the nvidiots at [H]forum actually jumped on the bandwagon.

Yeah I got a good laugh out of that thread. As for ruined, hard to take him seriously at all. I think the guy sits up all night examining every single detail about ATi cards so he can create threads like this thinking he's found a flaw.

If he's not paid by nV to do this then I worry for him.

Unit01 · 03-Oct-2004, 13:27

So it's down to if AF is forced through app or control panel?
From what has been provided it seems like it.

BTW that zoom in picture rocked

03-Oct-2004, 02:01	#1
Ruined Member Join Date: Sep 2003 Posts: 317	3dmark05 & ATI - Sacrificing Image Quality for Speed? After reviewing some 3dmark05 screenshot comparisons by Driverheaven.net and Bjorn3d.com, it became evident that ATI may be sacrificing IQ for increased performance. Check out these comparisions: 3DMARK05 ATI X800 ARTIFACTS: FOG RENDERING * PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREA IN THE RED OUTLINE NOTE the failure to render all of the fog in the ATI screenshot EXAMPLE 1 Nvidia GeForce 6800 Ultra 66.70 Beta ATI X800XTPE 4.9 Reference Rasterizer image EXAMPLE 2 Nvidia GeForce 6800 Ultra 66.81 Beta ATI X800XTPE 8.07 Reference Rasterizer image** Source: Beyond3D.com & NvNews.net forums ******* 3DMARK05 ATI X800 ARTIFACTS: FILTERING/MOIRE * PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREAS IN THE RED OUTLINES *NOTE the obvious moire patterns in the ATI screenshots* Nvidia GeForce 6800 Ultra 66.51 Beta ATI X800XTPE 4.9 w/ hotfix Source: Bjorn3d.com ****** 3DMARK05 ATI X800 ARTIFACTS: BANDING * PLEASE OPEN THESE TWO IMAGES IN SEPERATE BROWSER WINDOWS AND SWITCH BACK AND FORTH USING ALT TAB AND EXAMINE THE AREA IN THE RED OUTLINE *NOTE the increased banding in the sky in the ATI screenshots* Nvidia GeForce 6800 Ultra 66.51 Beta ATI X800XTPE 8.07 Beta Closeup of ATI's banding from above shot Source: Driverheaven.net __________________ We're all in it together.

03-Oct-2004, 02:17	#3
jvd Naughty Boy! Join Date: Feb 2002 Location: new jersey Posts: 12,731	Ruined , if i shut off ai it goes away , but my score doesn't change. It seems to be the general optimization messing around with 2 areas . __________________ Freexbox 360 !!! Free Psp!

03-Oct-2004, 02:57	#8
ChrisRay R.I.P. 1983-2010 Join Date: Nov 2002 Posts: 2,234	Are you sure this is a result of ATI purposely sacraficing Image Quality for performance? I find it more likely this might just be a result of the internal precision differences between the 2 cards. And what they are running in different spots. __________________ Nzone SLI Forum Administrator NVIDIA User Group Members receive free software and/or hardware from NVIDIA from time to time to facilitate the evaluation of NVIDIA products. However, the opinions expressed are solely those of the members

03-Oct-2004, 03:11	#12
jvd Naughty Boy! Join Date: Feb 2002 Location: new jersey Posts: 12,731	IT depends on how they put on 4x / 8x af . If they force it through the cp it wont work. , if they put it on through the ap it will . __________________ Freexbox 360 !!! Free Psp!

03-Oct-2004, 13:27	#25
Unit01 Member Join Date: Feb 2003 Posts: 190	So it's down to if AF is forced through app or control panel? From what has been provided it seems like it. BTW that zoom in picture rocked __________________ I'll think of something later

03-Oct-2004, 02:13	#2
hovz Member Join Date: May 2004 Posts: 920	i see the moire, but i dont see any noticeable banding. that is theres an equal amount of banding on both

03-Oct-2004, 02:18	#4
Razor1 Senior Member Join Date: Jul 2004 Location: NY, NY Posts: 2,680	Interesting very similiar to Doom 3 Anyone got shots with 4.8 and 4.9 drivers? The opts were off on these shots I think.

03-Oct-2004, 02:19	#5
hovz Member Join Date: May 2004 Posts: 920	what are ur specs and what do u score? no overclock

03-Oct-2004, 02:59	#10
Razor1 Senior Member Join Date: Jul 2004 Location: NY, NY Posts: 2,680	well they shouldn't be having this at x4 aa and x8 af.

03-Oct-2004, 03:16	#13
Razor1 Senior Member Join Date: Jul 2004 Location: NY, NY Posts: 2,680	they turned it on through the ap http://www.bjorn3d.com/read.php?cID=702&pageID=1050

03-Oct-2004, 03:21	#14
gordon Junior Member Join Date: May 2003 Posts: 76	Everything is fine with the r3xxs, must be x800 specific if true.

03-Oct-2004, 03:58	#15
gokickrocks Member Join Date: Dec 2002 Posts: 465	looked at the differences between the 2 images posted above...i agree with the moire pattern, but for the other red outlining, i would have to say the ati one is better, it seems more defined to me and if you look at the flooring texture, its clearer as well i was a bit more curious so i went through the images in bjorn3d and found many differences in each image, which may need a reference rasterizer (i do realize that a ref. rast. may not produce 100% accurate image, but the differences between the cards is big, so at least it will help to determine which card is more correct) to compare against...there are some large lighting and some texture quality descrepencies between the 2 cards (for those with firefox just open it in a new tab and use the mouse wheel to scroll) comparison 1 (besides the gun powder from the cannon, take a look at the lighting on the front of the ship)... http://www.bjorn3d.com/photos/showph...cat/503/page/2 http://www.bjorn3d.com/photos/showph...cat/503/page/1 comparison 2 (look at the flooring)... http://www.bjorn3d.com/photos/showph...cat/503/page/2 http://www.bjorn3d.com/photos/showph...cat/503/page/1 comparison 3 (take a look at the water)... http://www.bjorn3d.com/photos/showph...cat/503/page/2 http://www.bjorn3d.com/photos/showph...cat/503/page/1 edit: someone posted on the futuremark forums and compared between applying AA/AF through the CP and APP and found that when forced in the CP, the moire pattern is gone, and with the app, its there

03-Oct-2004, 04:49	#17
Apple740 Member Join Date: Aug 2004 Location: Rotterdam - NL Posts: 239	I took some tests with my X800XT. The moire in frame 985 is not there in the default game test, but appears when you set AF in the 3DMark05 CP (Pro version). But if you force AF by the Ati CP the moire dissapears. Ruined, I would reather see that you make an issue of the fact that Nv uses DST by default in 3DMark05. Without DST, Nv scores decrease by 10/16%, so at this moment it's Nv who sacrifices IQ for speed.

03-Oct-2004, 05:42	#19
Sanctusx2 Junior Member Join Date: Apr 2004 Posts: 31	Good god, I see this post on every forum I frequent. Someone has been busy. =P

03-Oct-2004, 06:00	#20
hovz Member Join Date: May 2004 Posts: 920	well come on now, its ruined. what do you expect from chalnoths long lost brother?

03-Oct-2004, 06:36	#22
RickCain Junior Member Join Date: May 2004 Posts: 70	Impressive he can actually post the same exact thread across multiple forums. L33t copy/paste skillz. Funniest part is the nvidiots at [H]forum actually jumped on the bandwagon.

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
ATI Supports Microsoft Windows XP Media Center Edition 2005	Dave Baumann	Press Releases	0	13-Oct-2004 10:32
Image Quality Comparisons, Some thoughts.	ChrisRay	3D Architectures & Chips	47	28-May-2004 23:01
NVIDIA Quadro Graphics Win Prestigious Image Quality Awards	Dave Baumann	Press Releases	0	20-Feb-2004 11:43
ATI Solutions for Microsoft Windows XP Media Center Edition	Dave Baumann	Press Releases	0	01-Oct-2003 14:38
A look at IQ at Firingsquad...	WaltC	3D & Semiconductor Industry	128	05-Sep-2003 13:49

Beyond3D Forum

Welcome, Unregistered.