Poll about image quality

[Arun]

I admit to be one of the impopular folks to have voted 1
Although that doesn't quite describe my judgement enough IMO.

The only signifiant rendering difference that I can be sure that isn't due to a slightly different frame is the missing light problem.

I would say it's likely other differences, such as the textures differences, are not due to the frame difference, but since it might be influenced by the lighting quite a bit, I just don't want to say I'm sure of it.

The reasons would most likely be NVIDIA's driver team incompetence - whether that's because 1) they can't tell the difference between a cheat with a big IQ difference or not, or because 2) they honestly write bug-ridden code, I don't know. I suspect at least a part of it would be due to (1) though.


Uttar

 

[Malo]

the fact you posted this poll pointing out the obvious differences not noted by the reviewer is opinion enough

 

[THe_KELRaTH]

I thought it was probably point sampling on the floor. Maybe they're bringing back as the default setting like in 40.xx drivers (can't remember the exact one)

 

[Pete]

I agree that there are at least four differences (possibly overlapping): the lights, the floor, the shadows near the door, and the dotted line in the door frame.

That D-L review flubbed quite a few of their IQ comparisons, though. As I pointed out at Ars and AT, D-L didn't comment on the striking texture (LOD?) difference in the plain SS2 shot, and what looks like deficient AF in nV's AM3 AA+AF shots. Another case of nicely comprehensive benchmarks and lacking IQ analysis. I'm sick of it, but it's mainly all we've got.

 

[darkblu]

But I don't believe the missing lights indicate a texture layer missing. It wouldn't make sense to blend a "lights on" layer onto the "lights off" layer, which is clearly visible.

i.e. unless the 'lights' feature is achieved through a detail texture (which then would be used also elsewhere in the game, and so is supposed to blend with whatever is underneath it). also, it's a fact that on all 4 corridor shots all r350 shots exhibit the 'lights' and all gffx shots don't - don't you find that odd wrt to the 'lights flickering' scenario?

 

[bloodbob]

The lighting would effect the bump mapping so its a bit hard for me to say. It does look like there is a bit of moore effects coming into play on the FX but it could be 3 things A) LOD mipamps B) Something cause by no light ( and the aliasing problems that occur with PS shaders in general ) C) shard mipmap transistion

 

[Zvekan]

Anyway, given who I work for I'm not sure if it is appropriate for me to give my opinions about these images in this forum, which is why I wanted to try to avoid turning this into an 'A is better than B' poll in the first place.

I understand why you wouldn't like to comment, but I'm implore (hope it is the right word ) you to reconsider. I personally write about graphics cards for a computer magazine (name is not important) and am having quite a lot of trouble spotting differences (I manage it only via image swamps, almost never have been able to see lack of AF on single screenshot, except on extreme cases).

Then I noted you discussion about AA methods when commenting on Anandtechs poor article and I understood that you know where to look and more importantly why the result is as it is.

So it would be very nice of you if you could provide more info (at least here on msg board, just say it is all you opinion and knowledge ) that could potentaly result in more accurate reviews.

Although it would be best if you could write a comprehensive guide and publish it via your marketing or offer Dave that you will be writing a short article.

Zvekan

 

[Xmas]

But I don't believe the missing lights indicate a texture layer missing. It wouldn't make sense to blend a "lights on" layer onto the "lights off" layer, which is clearly visible.

i.e. unless the 'lights' feature is achieved through a detail texture (which then would be used also elsewhere in the game, and so is supposed to blend with whatever is underneath it). also, it's a fact that on all 4 corridor shots all r350 shots exhibit the 'lights' and all gffx shots don't - don't you find that odd wrt to the 'lights flickering' scenario?

Flickering lights was just an explanation that came to my mind when I compared those shots. It's not very likely.

The detail texture idea doesn't make sense. You only use detail textures on close objects, and only where appropriate. Furthermore, they're high-frequency and repeated several times across a surface.

 

[Zvekan]

Also you made picks quite difficoult to understand, for instance There are other significant differences, I believe I know what caused them can be understood it at least to ways.

1. nVidia over enthusiastic driver developement team decided that proper lightning is to stressful for their FX cards (would very much like to see the same screen shot on GF4 Ti) and difference is not noticable in normal conditions

2. More technical stuff, like missing texture layers, lower quality filtering and such

I voted for "I believe I know what caused them", but thought about driver team and their optimiziations, while I have no idea what is really causing so drastic visual diference.

So I just wanted to say that % won't be very informative as I believe that some may have voted on same reasoning as me and that would mean that we don't really undestand what is going on.

Zvekan

 

[Pete]

Hmmm. I just played the Halo demo (a trying experience on an XP1700+ and 9100, even at 640x480--is that just a piss-poor port?), and those hallway lights seem to be static, so perhaps the FX card just isn't rendering them. Strange.

 

[dream caster]

As hard as I tried i could only see the light and floor difference; i do not know if my integrated video is showing all from those images but there is no dotted line in any door and no shadow difference.

 

[dream caster]

andypski:

I think it is just fair that you tell us what do you think; we just helped you telling what you are interested in, and you should give back to us telling what do you think of these differences.

 

[Bouncing Zabaglione Bros.]

Lights on the walls and behind all the vents and grates are compeltely missing on the GFX. GFX floor seems to have abrupt transitions and isn't as smooth (more pixelated). There is also more subtle and diffuse lighting on the 9800 on the walls and ceilings.

The 9800 image looks better. It just seems to me that the GFFX is cutting out details and cutting down the lighting & texture quality. If you load up the pictures into a viewer and flip between them, the differences are easy to see,

 

[Fred da Roza]

Some have indicated that it's not the same frame but I'll list the differences that I see.

Lights

Floor textures. Floor seems to shimmer on the FX.

The purple walls have less contrast on the FX

There seems to be some artifacts on the lights at the entrance on the Radeon.

The texture on the left wall at the entrance is much darker on the FX.

The "radar" indicators on the left bottom are different.

 

[Tahir2]

They both look terrible but they are not quite the same frame so comparison is IMHO useless.

 

[kipper67]

I must be blind but what missing lights are you all reffering to?

Apart from the differences on the floor, the only major thing i notice is that on the 9800 shot if you look at and through the grille in the ceiling there appears to be a lot of pipework behind it , whereas on the Geforce shot there is no pipework. Or are these the missing lights you're all on about?

Mark

 

[Xmas]

Have you set your monitor to extremely low brightness, or why aren't you able to see the lights, and the (existing) pipework on the GeForce shots?

 

[OpenGL guy]

They both look terrible but they are not quite the same frame so comparison is IMHO useless.

One thing that's apparent is that the AA quality on the GeForce FX is pretty low.

 

[John Reynolds]

One thing that's apparent is that the AA quality on the GeForce FX is pretty low.

ATI's AA looks pretty bad in Halo too. 8)

 

[Ailuros]

Even better AA wouldn't make me want to play Halo. The thing is not only blah looking, it's way too slow for what it actually delivers.