Aquamark3 Preview and Performance Comparison @ Driverheaven

Skell · Sep 15, 2003, 11:12 AM

Has anyone downloaded the hi-res images and XOR'd them? I have, I was completely AMAZED by what I saw.

For convenience, the images are scaled down, although i've uploaded the hi-res XOR's as well here

3.4 v 3.7
Here, both sets of drivers are rendering the scene in the same way, but the 3.7's are able to do it faster. This is the best case scenario, as the ATI driver team have managed to do this without sacrificing image quality

44.03 v 45.23
There looks to be very slight image quality loss when comparing the 44.03's and 45.23's. If you look at the right of the picture, the large building seems to be being rendered different. The 10,000 plus increase in the benchmark probably didnt come as a result of sacrifising image quality.

44.03 v 51.75
In this image however, when comparing the 44.03's and the new Det50's, a massive amount of the scene is being rendered differently...

45.23 v 51.75
... just like in this picture. Without the use of an XOR, you could see how badly parts of the scene were being rendered, but using an XOR, you can clearly see that there is a lot more image degredation than meets the eye.

44.03 v 3.7

45.23 v 3.7

51.75 v 3.7

The last three shots compare the dets quality with the cats. Its hard to come to a conclusion about which driver is rendering the scene correctly as there is no control image (an image whichis deemed by the makers of Aquamark as "correct")

I hope the admins dont mind me using their images, I just wanted to see exactly what was going on. Any problems and I will remove the images. If anyone wishes to correct me on ANYTHING, do it

I dont know if im right or wrong, its just my opinion.

Skell

***Veridian3*** · Sep 15, 2003, 12:00 PM

Skell,

Some very interesting findings there...would you mind explaining to everyone who's reading exactly what you did and what the results mean. (I'm sure not everyone will have seen this type of analysis before.)

Thanks

Stu

Skell · Sep 15, 2003, 12:37 PM

Im just using PSP 8's built in algorithms. an XOR compares two of the images and produces a sorta "map" of where the images differ. Black areas indicate no change (like when you are comparing the 3.4 cats to the 3.7's). Coloured areas indicate changes. As you can see, the dets IQ has gotton worse over the past releases but increased in performance. However, it looks as though ATI's drivers have increased performance and still kept the same image quality.

If anyone can give a better description of what an XOR image is, feel free. I dont claim to know much

All I know is that this is probably the best way of comparing how the different driver sets are rendering from each other.

Skell

EDIT: Heres a better description. An XOR lets u see what you wudnt be able to see by just looking at them

Al_Vampyre · Sep 15, 2003, 12:42 PM

Nice one Skell, that makes thins a little clearer...although the 3.4 3.7 comparison did throw me for a while becuase I din't know what I was looking for!!

Skell · Sep 15, 2003, 12:50 PM

Yeah sorry, I got a bit ahead of myself

Im sure a guru will come along soon and tell people what they mean *waits*

Zardon · Sep 15, 2003, 01:30 PM

of course we dont mind you using the images ! its GREAT to see people analysing these things for themselves, thats why I made sure everyone had available to them the high resolution uncompressed targas.

For those interested XOR is an simple and fast algorithm - a symmetric algorithm, in this case its showing the information which will be rendered differently (inverse). Im not sure if it will be 100% accurate depicting this information but it certainly gives a rough guideline of the differences in some areas with rendering.

Skell · Sep 15, 2003, 01:35 PM

Cheers Zardon

Could you suggest a better way to compare the images? Or anyone for that matter? Be really cool if we could find out whats really going on, and how much of a scenes' IQ has been altered. If theres one thing the XOR is telling us is that we arent going to see everything looking at them how they are atm

Skell

Zardon · Sep 15, 2003, 01:41 PM

well if you guys are finding it interesting enough I can further analyse the image maps and levels and perhaps show more later, im in work atm and I thought if I got too technical with people initially they would get quite bored with the whole thing. Ill get something up in this thread later.

I really like your ideas with the XOR algorithm, very creative.

Skell · Sep 15, 2003, 01:51 PM

Cool! Persoanlly im finding it very interesting.

Hehe, im sure if I hadnt have thought of using an XOR, somebody else would have

Im just a n00b

Skell

Zardon · Sep 15, 2003, 02:32 PM

you are certainly welcome on these forums I hope to see you about here in the future!

Skell · Sep 15, 2003, 02:51 PM

Dude ur too kind!

Looking forward to your more in depth and expert analysis

Skell

BWX · Sep 15, 2003, 03:18 PM

How about some screens with AA and AF at different levels?

Zardon · Sep 15, 2003, 07:06 PM

Terry Makedon in ATI is aware of the texture issue people have been talking about (the stripe section of the ship) and they are looking into it now. If I hear anything back from them ill let you know. He might of course post here himself about it at a later date.

Mac Daddy · Sep 16, 2003, 02:32 AM

Quote:

Originally posted by Zardon
you are certainly welcome on these forums I hope to see you about here in the future!

Very seldom we see these days a newer member with such input and tact ...
Very well done Skell .. and as Zardon mentioned .. you are very welcome on DH

Not a guru here ... but this is my best synopsis ......

" XOR is an algorithm which was run upon two basic comparisons ..
As Zardon mentioned these two images are inversely compared ..... Or in lamens terms super-imposed ..
This comparison allows us to view the addition and deletion (pros and cons) of the subject matter ..

In this case not only ATI vs Nvidia ... but newer Nvidia drivers vs older ... and so forth ....."

Good job Skell ....... And hope to see ya posting more here on DH

Skell · Sep 16, 2003, 09:10 AM

Thanks Mac Daddy!

Kris TPC · Sep 16, 2003, 01:30 PM

"No noticeable differencje in quality on 45.23" ?
Not exactly ...
Where is my shadow !? :-)

http://twojepc.pl/news_zdjecie.php?news03/5847_1.jpg

mixsetup · Sep 16, 2003, 02:44 PM

Ok I bougt a FX5200 ultra as for a ATI card which could do the dx9 and OGL1.4 I would have to pay a heap more and as I have kids I couldn't afford the ATI so at least I could afford the FX 5200 ultra

Zardon · Sep 16, 2003, 08:43 PM

Quote:

Originally posted by Kris TPC
"No noticeable differencje in quality on 45.23" ?
Not exactly ...
Where is my shadow !? :-)

http://twojepc.pl/news_zdjecie.php?news03/5847_1.jpg

do we know the card used to take those images?

Skell · Sep 16, 2003, 08:45 PM

Actually if you check your own shots, (frame 4500 I think off hand) you can see that the 45.23 drivers are in fact missing shadows and other black areas. If it wasnt pointed out to me, I wouldnt have seen it :S

Skell

EDIT: Nope 5000, sorry! Its not shadows as such, just big black blobs. Should be one at the front of the shot, and one to the right of the tank in the foreground.

Kris TPC · Sep 16, 2003, 08:54 PM

Quote:

Originally posted by Zardon
do we know the card used to take those images?

MSI GeForceFX 5600 Ultra Rev.2 & Sapphire Radeon 9600 (non Pro) ...

Zardon · Sep 16, 2003, 09:06 PM

I can confirm its there on several cards, with the 45.23 including my quadro.

Ive spoken with someone in the trade and he has said something to me but wants to confirm it first. It looks like it definately is a driver issue though.

Kris TPC · Sep 16, 2003, 09:07 PM

[QUOTE]Originally posted by Skell
[B]Actually if you check your own shots, (frame 4500 I think off hand) you can see that the 45.23 drivers are in fact missing shadows and other black areas. If it wasnt pointed out to me, I wouldnt have seen it :S

Skell

Check whole stage ... shadow ... no doubt ...

Skell · Sep 16, 2003, 09:08 PM

Cool cool, sorry

Mac Daddy · Sep 17, 2003, 02:12 AM

Quote:

Originally posted by Zardon
I can confirm its there on several cards, with the 45.23 including my quadro.

Ive spoken with someone in the trade and he has said something to me but wants to confirm it first. It looks like it definately is a driver issue though.

Then lets await Zardon's input !!!

Quote:

Originally posted by Skell
Cool cool, sorry

No need to be sorry .. this is a discussion ....
You did a great job in your comparison ....

Kris TPC .. your driver issues have been noted .......

sparkyg · Sep 17, 2003, 04:31 PM

Quote:

Originally posted by Zardon
Terry Makedon in ATI is aware of the texture issue people have been talking about (the stripe section of the ship) and they are looking into it now..

Also check frame 4000. On the craft on the ground on it's right side there is a texture with six circles. It is much clearer on the 4x.x drivers from Nvidia then the ATI ones. You can clearly see a thin line around the circles on the nvidia but not the ATI

Sep 15, 2003, 11:12 AM	#31
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Has anyone downloaded the hi-res images and XOR'd them? I have, I was completely AMAZED by what I saw. For convenience, the images are scaled down, although i've uploaded the hi-res XOR's as well here 3.4 v 3.7 Here, both sets of drivers are rendering the scene in the same way, but the 3.7's are able to do it faster. This is the best case scenario, as the ATI driver team have managed to do this without sacrificing image quality 44.03 v 45.23 There looks to be very slight image quality loss when comparing the 44.03's and 45.23's. If you look at the right of the picture, the large building seems to be being rendered different. The 10,000 plus increase in the benchmark probably didnt come as a result of sacrifising image quality. 44.03 v 51.75 In this image however, when comparing the 44.03's and the new Det50's, a massive amount of the scene is being rendered differently... 45.23 v 51.75 ... just like in this picture. Without the use of an XOR, you could see how badly parts of the scene were being rendered, but using an XOR, you can clearly see that there is a lot more image degredation than meets the eye. 44.03 v 3.7 45.23 v 3.7 51.75 v 3.7 The last three shots compare the dets quality with the cats. Its hard to come to a conclusion about which driver is rendering the scene correctly as there is no control image (an image whichis deemed by the makers of Aquamark as "correct") I hope the admins dont mind me using their images, I just wanted to see exactly what was going on. Any problems and I will remove the images. If anyone wishes to correct me on ANYTHING, do it I dont know if im right or wrong, its just my opinion. Skell Last edited by Skell; Sep 15, 2003 at 01:09 PM.

Sep 15, 2003, 12:37 PM	#33
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Im just using PSP 8's built in algorithms. an XOR compares two of the images and produces a sorta "map" of where the images differ. Black areas indicate no change (like when you are comparing the 3.4 cats to the 3.7's). Coloured areas indicate changes. As you can see, the dets IQ has gotton worse over the past releases but increased in performance. However, it looks as though ATI's drivers have increased performance and still kept the same image quality. If anyone can give a better description of what an XOR image is, feel free. I dont claim to know much All I know is that this is probably the best way of comparing how the different driver sets are rendering from each other. Skell EDIT: Heres a better description. An XOR lets u see what you wudnt be able to see by just looking at them Last edited by Skell; Sep 15, 2003 at 12:47 PM.

Sep 15, 2003, 12:42 PM	#34
Al_Vampyre HardwareHeaven Senior Member Join Date: Jan 2003 Location: Scotland Posts: 1,532 Rep Power: 72 System Specs	Nice one Skell, that makes thins a little clearer...although the 3.4 3.7 comparison did throw me for a while becuase I din't know what I was looking for!! __________________ "Knowledge is knowing that a tomato is a fruit; wisdom is knowing not to put it in a fruit salad" - Brian O'Driscoll - Ireland Rugby Team 2009 Grand Slam winning Captain.

Sep 15, 2003, 03:18 PM	#42
BWX unplugged Join Date: Nov 2002 Location: USNY Posts: 19,669 Rep Power: 110 System Specs	How about some screens with AA and AF at different levels? __________________

Sep 15, 2003, 07:06 PM	#43
Zardon DriverHeaven Founder Join Date: May 2002 Posts: 32,480 Rep Power: 179	Response from Terry in ATI Terry Makedon in ATI is aware of the texture issue people have been talking about (the stripe section of the ship) and they are looking into it now. If I hear anything back from them ill let you know. He might of course post here himself about it at a later date.

Sep 15, 2003, 12:00 PM	Thread Starter #32
*Veridian3* Administrator Join Date: Nov 2002 Location: Cloaked Posts: 5,028 Rep Power: 182	Skell, Some very interesting findings there...would you mind explaining to everyone who's reading exactly what you did and what the results mean. (I'm sure not everyone will have seen this type of analysis before.) Thanks Stu

Sep 15, 2003, 12:50 PM	#35
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Yeah sorry, I got a bit ahead of myself Im sure a guru will come along soon and tell people what they mean waits

Sep 15, 2003, 01:30 PM	#36
Zardon DriverHeaven Founder Join Date: May 2002 Posts: 32,480 Rep Power: 179	of course we dont mind you using the images ! its GREAT to see people analysing these things for themselves, thats why I made sure everyone had available to them the high resolution uncompressed targas. For those interested XOR is an simple and fast algorithm - a symmetric algorithm, in this case its showing the information which will be rendered differently (inverse). Im not sure if it will be 100% accurate depicting this information but it certainly gives a rough guideline of the differences in some areas with rendering.

Sep 15, 2003, 01:35 PM	#37
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Cheers Zardon Could you suggest a better way to compare the images? Or anyone for that matter? Be really cool if we could find out whats really going on, and how much of a scenes' IQ has been altered. If theres one thing the XOR is telling us is that we arent going to see everything looking at them how they are atm Skell

Sep 15, 2003, 01:41 PM	#38
Zardon DriverHeaven Founder Join Date: May 2002 Posts: 32,480 Rep Power: 179	well if you guys are finding it interesting enough I can further analyse the image maps and levels and perhaps show more later, im in work atm and I thought if I got too technical with people initially they would get quite bored with the whole thing. Ill get something up in this thread later. I really like your ideas with the XOR algorithm, very creative.

Sep 15, 2003, 01:51 PM	#39
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Cool! Persoanlly im finding it very interesting. Hehe, im sure if I hadnt have thought of using an XOR, somebody else would have Im just a n00b Skell

Sep 15, 2003, 02:32 PM	#40
Zardon DriverHeaven Founder Join Date: May 2002 Posts: 32,480 Rep Power: 179	you are certainly welcome on these forums I hope to see you about here in the future!

Sep 15, 2003, 02:51 PM	#41
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Dude ur too kind! Looking forward to your more in depth and expert analysis Skell

Sep 16, 2003, 09:10 AM	#45
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Thanks Mac Daddy!

Sep 16, 2003, 01:30 PM	#46
Kris TPC DriverHeaven Newbie Join Date: Sep 2003 Posts: 3 Rep Power: 0	Where is my shadow ? :-) "No noticeable differencje in quality on 45.23" ? Not exactly ... Where is my shadow !? :-) http://twojepc.pl/news_zdjecie.php?news03/5847_1.jpg

Sep 16, 2003, 02:44 PM	#47
mixsetup HardwareHeaven Newbie Join Date: Sep 2003 Location: Australia Posts: 2 Rep Power: 0	Cost Ok I bougt a FX5200 ultra as for a ATI card which could do the dx9 and OGL1.4 I would have to pay a heap more and as I have kids I couldn't afford the ATI so at least I could afford the FX 5200 ultra

Sep 16, 2003, 08:45 PM	#49
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Actually if you check your own shots, (frame 4500 I think off hand) you can see that the 45.23 drivers are in fact missing shadows and other black areas. If it wasnt pointed out to me, I wouldnt have seen it :S Skell EDIT: Nope 5000, sorry! Its not shadows as such, just big black blobs. Should be one at the front of the shot, and one to the right of the tank in the foreground.

Sep 16, 2003, 09:06 PM	#51
Zardon DriverHeaven Founder Join Date: May 2002 Posts: 32,480 Rep Power: 179	I can confirm its there on several cards, with the 45.23 including my quadro. Ive spoken with someone in the trade and he has said something to me but wants to confirm it first. It looks like it definately is a driver issue though.

Sep 16, 2003, 09:07 PM	#52
Kris TPC DriverHeaven Newbie Join Date: Sep 2003 Posts: 3 Rep Power: 0	[QUOTE]Originally posted by Skell [B]Actually if you check your own shots, (frame 4500 I think off hand) you can see that the 45.23 drivers are in fact missing shadows and other black areas. If it wasnt pointed out to me, I wouldnt have seen it :S Skell Check whole stage ... shadow ... no doubt ...

Sep 16, 2003, 09:08 PM	#53
Skell DriverHeaven Newbie Join Date: May 2003 Posts: 9 Rep Power: 0	Cool cool, sorry