PDA

View Full Version : 3dmurk2k3: Images & The Truth


Pages : [1] 2 3

dohcmark8
06-06-03, 11:14 AM
I present you with images of Tech-Reports false claims of nvidia using application detection for 3dmark03, thus lowering quality. I will allow you to make your own judgements on this. These are Tech-Reports images, not my own.

Left is 3dmark, right is 3dmurk.

1. First i have a side by side comparo of the two images.
png version available here: http://mywebpages.comcast.net/dohcmark8/3dmark/murkmarkbignorm.png
http://mywebpages.comcast.net/dohcmark8/3dmark/murkmarkbignormj.jpg

2. This image is using the solarize filter to show differences in lighting.
png: http://mywebpages.comcast.net/dohcmark8/3dmark/murkmarkbig.png

http://mywebpages.comcast.net/dohcmark8/3dmark/murkmarkbigj.jpg

3. I have zoomed the two images 400% to verify any quality differences.
png: http://mywebpages.comcast.net/dohcmark8/3dmark/zoommark.png
http://mywebpages.comcast.net/dohcmark8/3dmark/zoommarkj.jpg

4. I solarized the zoomed images
png: http://mywebpages.comcast.net/dohcmark8/3dmark/zoommarksolarize.png
http://mywebpages.comcast.net/dohcmark8/3dmark/zoommarksolarizej.jpg

Therefore i conclude Tech-Reports report of the 3dmurk issue is indeed false.;)

Don't tell me the differences are hard to see. Tell me, if you see any differences with your own eyes when looking at the images?.

Hanners
06-06-03, 11:21 AM
Originally posted by dohcmark8
Therefore i conclude Tech-Reports report of the 3dmurk issue is indeed false.;)

How do explain the image quality differences shown in the screenshots in their article then?

Not to mention the 1000 point drop the 5800 Ultra suffers when 3DMark is renamed?

dohcmark8
06-06-03, 11:25 AM
Originally posted by Hanners
How do explain the image quality differences shown in the screenshots in their article then?

Not to mention the 1000 point drop the 5800 Ultra suffers when 3DMark is renamed?

I think they made thos numbers up, atleast until i see another site verify this, these are Tech-Reports images, do you see any differences?.

solofly
06-06-03, 11:28 AM
Originally posted by Hanners
Not to mention the 1000 point drop the 5800 Ultra suffers when 3DMark is renamed?

You mean ONLY when 'AF' is used...

Now I would like to see the same tests done using ATI card too...

digitalwanderer
06-06-03, 11:44 AM
Originally posted by solofly
You mean ONLY when 'AF' is used...

Now I would like to see the same tests done using ATI card too...
I've been trying that on me 9700 Pro using 2x, 4x, 6xAA & 8x, 16x AF all this morning and the most difference I got was 1 point.

The poster sort of forgot to mention what card, OS, & drivers he was using. I'm not flaming, I'm honestly curious. (BTW-I'm benching me GF4 now to see if there's any "surprises" in it. ;) )

dnoyeB
06-06-03, 11:53 AM
if its about AF, then rech report should have used some of the AF tools to show us the difference in the AF planes. I guess they will do a part 2 because so far they only showed the difference, but no suspected reasons.

dohcmark8
06-06-03, 11:54 AM
Originally posted by digitalwanderer
I've been trying that on me 9700 Pro using 2x, 4x, 6xAA & 8x, 16x AF all this morning and the most difference I got was 1 point.

The poster sort of forgot to mention what card, OS, & drivers he was using. I'm not flaming, I'm honestly curious. (BTW-I'm benching me GF4 now to see if there's any "surprises" in it. ;) )

"These are Tech-Reports images, not my own."

euan
06-06-03, 12:00 PM
I checked the two frames in 3dmark2003 then 3dmurk2003. The two pics were entirely identical. I can post pics is requested but nvidiots would say they were wrong. I would also post benchmarks, but I can't be bothered watching the scenes, and it has already been tried over at rage3d. :)

I present you with images of Tech-Reports false claims of nvidia using application detection for 3dmark03

Yes, how to start an unbiased, and impartial "post your opinion" thread. 10 out of 10. :angel2:


Although the differences are minimal, there is definetly differences on the edges, which looks more like lighting tricks to me. IE reduced shader precision. The fact that AF controls it is simply the symptom, not the cause. IMHO of course.

GlowStick
06-06-03, 12:05 PM
Yeah, their article was pritty poorly done.

When they showed you the two pictures side by side then said

Differences between the two are not immediately obvious here, even when you're looking at the full-sized PNG images.

The only way they can even try to reproduce this is by useing photoship filters? That truely dose make me laugh.

Also, Nvidia and ATi drivers dont look for exe file names anymore, their programmers are not that bad.

A good example is AOL Kids. They make those stupid marco scrolling programs out of VB, and they do it two ways.

THe first way is called SendKeys, witch looks for the window caption, then simulates typeing. That is easlily circumvented by havieng random caption names, eg adding a random number of spaces.

But how a REAL programmer would do that, is to use api calls find the preset window class relations that they cannot change with out a patch.

Trying to detect a game by its exe is just, silly.

R.Carter
06-06-03, 12:10 PM
Originally posted by dohcmark8
Tell me, if you see any differences with your own eyes when looking at the images?.
Of course I can see the difference. The edges of objects look different in the two pictures.

*points* Just take a look for yourself!

Hard to say which is "better" though, but there is a difference.

EDIT: In case you can't see anything just look at # 4 images with the desk. The one on the right has green specks on the edge, while the one on the right has NO green specks at all. The edges are a lot darker. k thx bye.

If you can't see it with your eyes then I suggest you mess with your monitor settings or something or write a program to do a pixel by pixel comparison of the two raw screen shots.

Of course, it's a good idea to "optimize" in scenes that are rather dark to make the "optimizations" harder to spot.

DivotMaker
06-06-03, 12:14 PM
Originally posted by euan
Although the differences are minimal, there is definetly differences on the edges, which looks more like lighting tricks to me. IE reduced shader precision.

Euan,

Not picking on you here, ok...

Sorry, but I fail to see any difference in ANY of the pics. If people are resorting to taking static screenshots, then magnifying them and running them through complicated image programs and THIS is what they find, then I am afraid that there are alot of people with WAY too much time on their hands.

Last time I looked, games were in motion at 30+ frames per second and there is no way in hell that people can tell the differences between the screenshots posted above in this thread. I am looking for ANY differences that I can (20/15 vision here) and I can't find any. I hate to say it, but it really looks as if someone is trying really hard to find something that just is not there.

R.Carter
06-06-03, 12:18 PM
Originally posted by GlowStick
But how a REAL programmer would do that, is to use api calls find the preset window class relations that they cannot change with out a patch.

Well, you could always edit the class and window names by hex editing the .EXE file itself.

However, I'm not sure if a display device driver can really access the Win32 API.

You can easily scan the imports of the driver files to detemine what calls it is making and I'm pretty sure that Microsoft does scan the files before giving them WHQL certification to check that they aren't calling stuff that they shouldn't be calling.

So if the following was in a driver it might not pass

HWND hwndPrevInstance = FindWindow( szClassName, szWindowName );
if ( hwndPrevInstance != NULL )
{
// Leet driver "optimizations" enabled
}


Hmmm... I wonder if you enabled the DriverVerifier (Verifier.exe) on Nvidia's drivers if it will complain very loudly.... hmmm.... best to have a kernel debugger installed before messing with DriverVerifier though.

digitalwanderer
06-06-03, 12:21 PM
Originally posted by BigBerthaEA
Euan,

Not picking on you here, ok...

Sorry, but I fail to see any difference in ANY of the pics. If people are resorting to taking static screenshots, then magnifying them and running them through complicated image programs and THIS is what they find, then I am afraid that there are alot of people with WAY too much time on their hands.

Last time I looked, games were in motion at 30+ frames per second and there is no way in hell that people can tell the differences between the screenshots posted above in this thread. I am looking for ANY differences that I can (20/15 vision here) and I can't find any. I hate to say it, but it really looks as if someone is trying really hard to find something that just is not there.
You're entitled to your near-sightedness/nVidia colored glasses, but there ARE differences between the two pictures and how much is irrelevant.

This is a BENCHMARK, not a game. Meant to be run ONE way. :rolleyes:

I 'spose the fact that there is such a broad error on the scores by just renaming the .exe is just "coincidence"? Puh-leeeese. :rolleyes:

I'd give it a rest at defending them too much this time, it looks like they REALLY got caught with their trousers down! :lol:

Hellbinder
06-06-03, 12:21 PM
The point is... It does not have to be Viewable for it to be inapropriate. Just like Clip-plane issue from before.

Hellbinder
06-06-03, 12:23 PM
You know, It is totally True that the IQ is really not that greatlly affected. I Personally dont see a problem If their AF looked like that in any game anywhere. But that is not really the point here.

the issue here is they are doing specific application detection because they have found a way to Fiddle With the Texture Stages or other things. Which obviously cant be used accrossed the board. But only in this or other Rail type Benchmarks. Thats my take on it.

It artificially inflates the score and is clearly Trickery or Hackery. I am not going to call this cheating per se. However it does decieve Buyers and OEMs into thinking the card is regularly capable of something its not. I have a Strong feeling they are likely pulling the same Kind of things in some of the other major benchmark apps/Games as well.

What it really is, is like Bait and Switch technique, Or 3 card monty. Its the consumers who are losing becuase they think they are getting the highest performing card with AA+AF in said applications. when in reality they are not.

saturnotaku
06-06-03, 12:23 PM
I think this is totally psychological. If you give someone two exact images and tell them to find the differences (instead of "do these pictures look alike to you") they will always find something that's different.

DivotMaker
06-06-03, 12:24 PM
Originally posted by digitalwanderer
You're entitled to your near-sightedness/nVidia colored glasses, but there ARE differences between the two pictures and how much is irrelevant.

Yawn....:rolleyes: :rolleyes:

DivotMaker
06-06-03, 12:25 PM
Originally posted by saturnotaku
I think this is totally psychological. If you give someone two exact images and tell them to find the differences (instead of "do these pictures look alike to you") they will always find something that's different.

EXACTLY...

Hanners
06-06-03, 12:25 PM
Originally posted by GlowStick
Also, Nvidia and ATi drivers dont look for exe file names anymore, their programmers are not that bad.

Then why does renaming the EXE make a difference to the scores and image quality?

I would have agreed that it's a small, petty thing for anyone to notice if it wasn't for the massive change in score - Something is definitely up, and I'm amazed that anyone can leap to nVidia's defence any more after the events of the past few weeks.

Hellbinder
06-06-03, 12:26 PM
See point again that IQ is not the Crux of the issue.

Kihon
06-06-03, 12:27 PM
Originally posted by BigBerthaEA

Sorry, but I fail to see any difference in ANY of the pics. If people are resorting to taking static screenshots, then magnifying them and running them through complicated image programs and THIS is what they find, then I am afraid that there are alot of people with WAY too much time on their hands.



Look at the edges of the second table. on the 3dmurk image, you can see several "layers" on the surface of the table" on the 3dMark one, the layers are not noticable.

Seriously, if you cant see that, you need to check your eyes.

Not sure how this is related to aniso though.

R.Carter
06-06-03, 12:28 PM
Originally posted by BigBerthaEA
Last time I looked, games were in motion at 30+ frames per second and there is no way in hell that people can tell the differences between the screenshots posted above in this thread. I am looking for ANY differences that I can (20/15 vision here) and I can't find any. I hate to say it, but it really looks as if someone is trying really hard to find something that just is not there.

Well, one could argue that a LOT of stuff isn't noticable when it is going at 30FPS.

The fact is

1) When doing a pixel by pixel comparison there shouldn't be ANY differences in the two screenshots if the name of the executable was changed, right?

2) There shouldn't be a performance difference if the name of the executable was changed, right?

However, there do appear to be differences from simply renaming the executable and some people want to know why this is the case.

Originally posted by saturnotaku
I think this is totally psychological. If you give someone two exact images and tell them to find the differences (instead of "do these pictures look alike to you") they will always find something that's different.

Well, to prevent any "human factors" a program could be written that does a pixel by pixel comparison to detect any minute differences in the two images. Computers aren't usually affected by psychological issues.

So, just write one and prove that there are NO differences in ANY of the pixels.

Hanners
06-06-03, 12:33 PM
Seeing as we're talking about the 'truth', how come nobody has posted this image from the Tech Report article? This is the one that says it all at the end of the day, whether you believe it or not:

http://www.tech-report.com/etc/2003q2/3dmurk03/480-diff.jpg

saturnotaku
06-06-03, 12:33 PM
Originally posted by R.Carter
Well, to prevent any "human factors" a program could be written that does a pixel by pixel comparison to detect any minute differences in the two images. Computers aren't usually affected by psychological issues.

So, just write one and prove that there are NO differences in ANY of the pixels.

Exactly. Anyone out there brave enough to write one? :D

digitalwanderer
06-06-03, 12:33 PM
Originally posted by saturnotaku
I think this is totally psychological. If you give someone two exact images and tell them to find the differences (instead of "do these pictures look alike to you") they will always find something that's different.
Mebbe, but that doesn't really go a long way towards explaining the massive scoring difference now does it? I'm pretty weird when it comes to computers, but even I don't think they have much of a psychological bias when it comes to rendering... ;)

Originally posted by R.Carter
Well, to prevent any "human factors" a program could be written that does a pixel by pixel comparison to detect any minute differences in the two images. Computers aren't usually affected by psychological issues.

So, just write one and prove that there are NO differences in ANY of the pixels.
I was under the impression that was what they had done in the article with that funkadehlic black & white spritely looking pictures...those were all the different pixels right there. :)