IQ Observation - not what you think.

phsstpok

Splendid
Dec 31, 2007
5,600
1
25,780
OK we have all seen the GamersDepot comparison, "Detonator 50: First Look", and we know that IQ is horrible with nVidia's 51.75 beta driver but I noticed something else.

With the 9800 Pro the image is more detailed with pixel shader 1.4 than with 2.0. Look at both ATI images of Tomb Raider (you may need to adjust brightness to see it well). At the end of the hall there is a small diameter pipe sitting on top of a large diameter pipe. To the right of these is a metal box with all kinds of detail on it. You can clearly see the detail with the PS 1.4 image. The PS 2.0 image is blurry and the detail is lost.

Why would this be, sharper with PS 1.4 vs PS 2.0, that is?

Also, exactly what improvements should I be looking for with PS 2.0?. I don't see any.

Here's a link to the article, once again.

<A HREF="http://www.gamersdepot.com/hardware/video_cards/ati_vs_nvidia/dx9_rel50/001.htm" target="_new">http://www.gamersdepot.com/hardware/video_cards/ati_vs_nvidia/dx9_rel50/001.htm</A>

<b>56K, slow and steady does not win the race on internet!</b>
 
Man, you really look at very high detail! But you are right. I don't thing PS 2.0 implies blurry images, maybe is a problem when working with them. Maybe ATI don't have drivers correctly "adjusted". Another possibility would be a defective architecture ... but I don't want to think about that possibility!

Another thing is that luckily you won't notice too much the diference when playing ... or another shiny effects will bring your attention :wink:

Very good work!


Still looking for a <b>good online retailer</b> in Spain :frown:
 
LOL! Normally I don't look at that much detail but I was trying to see the difference between PS 1.4 and 2.0. I was looking for improvements but I found the opposite in this case.

I think I need a scene other than a dimly lit hallway to see what I am looking for.

As for that bluriness. It has me concerned. I don't think it should be there since testing was supposedly done with the intent of maximum detail.

Would it part of the normal filtering? I mean it is a distant object that is blurry. It's also supposed to be in shadow. Might the detail be lowered intentionally? Is it a mistake? Could it be a subtle cheat? Who knows?

<b>56K, slow and steady does not win the race on internet!</b>
 
Also remember that the section captured is actually VERY small and the picture is blown up alot (when I ran he becnhieI looked for that segment).
It would be nice to see more in depth look once the main (Beyond3D, Extremtech, etc.) players get their hands all over it. Expect some pretty good comparisons in the near future. The more issues exposed the better IMO.



- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <A HREF="http://www.redgreen.com" target="_new"><font color=green>RED</font color=green> <font color=red>GREEN</font color=red></A> GA to SK :evil:
 
That's without aniso on. I wonder what difference that would make.

Neways, in ps2.0 mode Tomb Raider: AOD uses depth of field, much like that in that DX9 demo with the balls, and the real time lighting (u know Grape!?). The purpose of the effect is to blur things far away (our eyes can't focus on everything at the same time), maybe thats what it's doing, or maybe the effect isn't being used properly.

"Mice eat cheese." - Modest Mouse

"Every Day is the Right Day." -Pink Floyd
 
If I have to look that far for anything I'll need a bigger monitor. I can't see it on a 15" at work.

<font color=blue>other people's lives
seem more interesting
cuz they aint mine
</font color=blue>
<font color=green>Modest Mouse</font color=green>

TKS
 
Also remember that the section captured is actually VERY small and the picture is blown up alot (when I ran he becnhieI looked for that segment).
I didn't consider that possibility so I re-read the article. I think you are wrong. The article implies that the photos are direct screen captures, "All photo's are "untouched" - they're taken right from the screen cap folders and saved as high-res JPG's".

However, since you brought up the subject I just realized the images are 1024x768 captures and I have been viewing at 1280x960. At the original resolution the difference is even more apparent. It stick out like the proverbial sore thumb. Now I can see that the whole latter portion of the hallway is out of focus.

Furthermore, in the PS 2.0 image Lara is standing farther down the hallway. She is actually closerto the object I mentioned than in the PS 1.4 image yet the latter has more detail.

Something is definitely wrong here.

I also look forward to a more in depth examination of this issue (but I don't look forward to the nVidia fanboy I-told-you-so's).

<b>56K, slow and steady does not win the race on internet!</b>
 
Anisotropy? Perhaps GamersDepot made a mistake and left AF enabled with the PS 1.4 capture.

Depth of field is an interesting idea/technology but it would sure mess up IQ comparisons.

If that's what I am seeing then it sux. I hope we can turn it off.

(Yeah, that's it. Now that we have a nice image with anisotropy let's go ahead and blur it to make it look "realistic").

<b>56K, slow and steady does not win the race on internet!</b>
 
If you are using 800x600 on that 15-incher then you can't miss it. At 1024x768 I don't think you would be able to pick out the details.

<b>56K, slow and steady does not win the race on internet!</b>
 
Actually I saved the file and looked at it with Lview (only one I have at work).

I Set the desktop to 10x7 (from my beautiful usual 16x12) on my P260, and I viewed (or Lviewed) the 4 in slide-show.

Yes, the difference is apparent, but the blur does seem intentional. I occurs on both cards, and both cards provide much greater definition in PS 1.4. But that also doesn' look as good if you loof the strictly defined transitions at the bottom right edge of the wall anf foor nearest laura and the light on the wall. It's like a pencil line in the PS 1.4 and more like a fan paint-brushed blended border on the PS 2.0 shots
However by the same token the shadows don't appear as nice in the PS 1.4 shots likely because of that. (also not that the R9800 PS 1.4 shows the overhead light, which you don't see in the others).

The main thing would be to get the reference shots (what he programmers were intending) and comparing them. Here there is no way of telling which one is doing what the programmers intended.

Look at the floor in the PS 1.4 shot of the FX, is that trail of lights supposed to look like that or more diffused. It's more diffused in the R9800 in both PS 1.4 and PS 2.0 as well as the Ps 2.0 of the FX.

I wish it were more of exact copy since alot of things may simply be based on the camera position and laura's position. That overhead light definitely plays a role, and it would also explain why Laua's so dark (based on a camera view [which we are used to looking washed out due to the background lighting]). So it's almost impossible to tell what meant to be there and what's not.

I see isues in them all, some look better to ME, but the question is what did the Developers WANT me to be seeing?

I'll wait for better/more detailed reviews.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <A HREF="http://www.redgreen.com" target="_new"><font color=green>RED</font color=green> <font color=red>GREEN</font color=red></A> GA to SK :evil:
 
I just got a reply from Duane Pemberton, the article's author.

He say's the effect is due to "motion blur". Lara Croft was sliding down the ramp at the time of the screen capture.

Now that kind of bothers me too. For a first person view you might expect blurring as seen from the character eyes. However Tomb Raider games use a third person view. Why would the camera have the blurred effect? (OK I'm just nit picking but I wouldn't choose blurring for the camera's view, ).

I'm going to hate PS 2.0 in FPS games if we are going to get blurring every time we move.

But that also doesn' look as good if you loof the strictly defined transitions at the bottom right edge of the wall anf foor nearest laura and the light on the wall.
I don't mean to be a wise *ss but could you translate that.

<b>56K, slow and steady does not win the race on internet!</b>
 
<font color=blue>But that also doesn't look as good if you loof the strictly defined transitions at the bottom right edge of the wall anf foor nearest laura and the light on the wall.</font color=blue>

Sorry I was looking away while typing.

What I was trying to say was that the sharp edges don't look as good as the smooth edges at the transition near the floor and the wall which is most distinct near the the bottom right side of the frame.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <A HREF="http://www.redgreen.com" target="_new"><font color=green>RED</font color=green> <font color=red>GREEN</font color=red></A> GA to SK :evil:
 
I'm looking at the wall-floor transitions. It looks very much like aliasing vs anti-aliased but anti-aliasing is not enabled as can be seen when looking a Lara's right leg. I switched to 800x600 to see better detail.

I think I need to read up on PS 2.0

<b>56K, slow and steady does not win the race on internet!</b>
 
I think I need to read up on PS 2.0
Yeah, don't we all. :smile:

The transition does looked aliased and it also looks very blended (sort of the opposite of AF). The transition on both of the PS 1.4 shots looks very harsh, and if the surcafe is suposed to look like a smooth rounded cave floor, then the screen is done correctly in PS 2.0 ubt not in 1.4, the question is what are they TRYING to do, or what do they Expect you to see?




- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <A HREF="http://www.redgreen.com" target="_new"><font color=green>RED</font color=green> <font color=red>GREEN</font color=red></A> GA to SK :evil:
 
I don't know what was intended but the transition from ceiling to wall looks fairly well blended.

I wonder if we are giving too much credit to artistic intent.

Suppose the designer just sketched it out, deciding the hall would be dark and walls and floors would be roughly the same color and texture and meet at a right angles. What if he didn't give any thought to the transition and just got what he got?

<b>56K, slow and steady does not win the race on internet!</b>
 
Well that may be the case, but if we are discussing IQ, then the fact that we don't know what the reference frame looks like we really can't say what is doing it right or wrong on such a minute issue. Glaring errors are obvious, but whether a transition should be clurred, or sharp may be dependant on the desired effect.

I don't know, I just know that there are alot of differences, and what we may THINK looks better is actual the card not being able to do what the programmers intended, therfore, what do we know until we know what they knew. :wink:


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! <A HREF="http://www.redgreen.com" target="_new"><font color=green>RED</font color=green> <font color=red>GREEN</font color=red></A> GA to SK :evil: