First few GFX benches

Status
Not open for further replies.
The whole problem with subjective measures such as IQ is that they can usually only be quantified in terms of a step function.

I think a step function methodology is what Sharkfood and others are advocating (which I think can and should be done) whereas others such as Democoder are thinking in terms of a continuous function (which I think cannot and should not be done).

EDIT: Whoohoo... turned the page :D
 
Sharkfood said:
So then, by your basis of judgement, there is no quantifiable difference possible between these two images:
<Q3 hyperblaster images snip>
The general hypothesis of A > B can be derived. No review board of scholars is needed.
Click to expand...
Actually, I would disagree. In the Q2 example you posted I personally think the first one looks somewhat overbright and cartoonish, while the second looks washed-out and dim, but probably more 'photoreal'. Also note that the HUD is only clearly visible in the B image.

Neither I would say is clearly superior to the other...

Of course, the second image with the AA is far more obvious.
 
Ideally, yes. But I think that we're all old enough to realize that popularity does not equal "truth".
Click to expand...

Agreed. I did not say popularity=truth. Again, as it pertains to subjective matters, "truth" is in the eye of the beholder.

Roger Ebert may be the most "popular" movie critic. If I have completely different criteria for determining what makes a good movie, then I don't care how popular he is, I don't pay his reviews too much mind.
This doesn't mean he is wrong or not speaking the truth.

They aren't matters of opinion or subjectively arguable.. or at least not with any amount of common sense or logic.
Click to expand...

There are indeed some cases like that, but other matters are certainly subjectively arguable.

What is better image quality wise? 2X supsampling AA, or 4X Multisampling AA?

We have "sources" willing to compare as "comparable" (due to this whole "subjective" nonsense theorem) images of smooth and correct color, smooth edges, near perfect alpha blends....
Click to expand...

Yes, we do have that problem. And this is why it is important for ALL reviews to have the subjective analysis. Because the more people giving subjective analysis, the easier it is to root out the non-sense reviewers who would come up with something completely different than "everyone else."

The problem is when that point is stretched beyond the surreal in order to maintain equivalence in favor of one IHV versus another, or otherwise allow completely apples vs. oranges comparisons.. that would be obviously unanimous dissention by anyone.. well, other than Stevie Wonder or Helen Keller.
Click to expand...

Yes, we all know what the problem is. ;) But if it's of such "obviously unanimous dissention" by anyone, then that reviewer gets rightfully slammed and its credibility called into question.

If 10 sites give some subjective analysis that says similar things, and one site says something completely different....then that one site's analysis is obviously suspect. Credibility goes down.
 
Dio said:
Actually, I would disagree. In the Q2 example you posted I personally think the first one looks somewhat overbright and cartoonish, while the second looks washed-out and dim, but probably more 'photoreal'. Also note that the HUD is only clearly visible in the B image.

Neither I would say is clearly superior to the other...

Of course, the second image with the AA is far more obvious.
Click to expand...

That's fine, but that still doesn't excuse the reviewer from providing his opinions. You are perfectly free to disagree with those opinions if you like, of course (I often do, but I still have more respect for a reviewer who provides his opinions and attempts to justify them than I do for one who plays the "above-it-all game" and pretends to have no opinions--usually those kinds of reviews are surprisingly free of valuable information. )

One of the most egregious violations of review "ethics" I have ever seen came when Anand reviewed 3dfx's V3. The entire review was plastered with the most horrible screens shots you've ever seen--the quality was *horrible*, providing the impression that the V3 was truly a terrible product (at least I thought so at the time.) But there was one sentence in the review that aroused my curiosity, and that was when Anand said--almost regretfully--that the screenshots did not look like what he saw on the screen when he actually used the card. So I went immediately out and bought a V3 to see what was what. The image quality was so good that I had no qualms about keeping the V3 and removing my former TNT/V2 combo that had worked so well for a long time. In fact, the V3's image quality at the time was simply the best I'd ever seen in a 3D card.

Of course it turned out that Anand's screenshot software was not up to the task of properly grabbing a frame from the V3, and with the right screenshot software very good representations of what the V3 displayed were possible.

The question haunts me to this day: how could *any reviewer* knowingly publish screen shots that *he knew* were far below the caliber of the IQ he saw from the product on his own review screen? The first thing anyone would suspect in such an event was that the frames weren't being grabbed properly. To this day it is my opinion that he deliberately exploited a problem with screen-grabbing software & the V3 which existed for a time shortly after its release to make the product look much, much worse than it was. Perhaps, though, I give him too much credit and he simply did not know at the time that improperly operating screen-grabbing software would explain the discrepancy between the frames he grabbed and what he actually saw on the screen. He was, after all, very young and inexperienced at the time. But to my knowledge he never officially rescinded the screenshots even after it became public knowledge that the V3 needed a little different screen-grabbing software to get accurate representations of its image quality. So I still cannot rule out that it was an intentional exploitation, done for reasons I can only surmise.

This is a great example, albeit an extreme one, of what can happen when a reviewer attempts to remain aloof about IQ.

Edits: Typos
 
I will repost this quote and add some emphasis:

That's fine, but that still doesn't excuse the reviewer from providing his opinions. You are perfectly free to disagree with those opinions if you like, of course (I often do, but I still have more respect for a reviewer who provides his opinions and attempts to justify them than I do for one who plays the "above-it-all game" and pretends to have no opinions--usually those kinds of reviews are surprisingly free of valuable information. )
Click to expand...

It is the reviewer's justification that is key. Because that is what gives the reader the ability to judge weather or not the opinion is reasonable.

Copare these two "opinions":

1) Saying "I think Product A's image looks bettter."

2) Saying "Overall, I prefer product A's image quality. Though product A's image lacks AA on textures, it has better edge AA quality. In my opinion, the always present increased edge quality outweighs the times when lack of texture AA is noticable." Here's some screenshots:

With number 2, the reader knows what the differences between the two are, and has some data to come up with a reasonable basis for agreeing or disagreeing with the opinion of the reviewer.
 
Joe DeFuria said:
I will repost this quote and add some emphasis:

That's fine, but that still doesn't excuse the reviewer from providing his opinions. You are perfectly free to disagree with those opinions if you like, of course (I often do, but I still have more respect for a reviewer who provides his opinions and attempts to justify them than I do for one who plays the "above-it-all game" and pretends to have no opinions--usually those kinds of reviews are surprisingly free of valuable information. )
Click to expand...

It is the reviewer's justification that is key. Because that is what gives the reader the ability to judge weather or not the opinion is reasonable.

Copare these two "opinions":

1) Saying "I think Product A's image looks bettter."

2) Saying "Overall, I prefer product A's image quality. Though product A's image lacks AA on textures, it has better edge AA quality. In my opinion, the always present increased edge quality outweighs the times when lack of texture AA is noticable." Here's some screenshots:

Now, the reader knows what the differences between the two are, and has some data to come up with a reasonable basis for agreeing or disagreeing with the opinion of the reviewer.
Click to expand...

Precisely.... :)
 
I think I've been unclear.

My criticism was that it was said to be blindingly obvious A was better. I personally disagree with that particular statement; I think it's difficult to say that because there was nothing 'wrong' or 'bad' with the second image that is immediately 'obvious' (or, at least, nothing where alternate points can't be made against the first). I'd say it's hard to create a quantitative test for cases such as these.

However, his second case was an excellent example of where there IS quantitative differences between two images.
 
I don't believe any of the 9700 's FSAA shots published so far are accurate either with the gamma correction feature of the 9700..I believe its applied after the frame buffer so screen captures from the frame buffer would not show the true output ?? :?:

gammas.gif
 
WaltC said:
The question haunts me to this day: how could *any reviewer* knowingly publish screen shots that *he knew* were far below the caliber of the IQ he saw from the product on his own review screen?
Click to expand...
I think you stated it yourself: inexperience and just plain not thinking. As we've been discussing, the quality of web reviews isn't good even now, and at the time it was worse.

I hesitate to see malfeasance, when incompetence is so much more prevalent :)
 
Dio, which of the Quake2 images that Shark posted represents more accurately what the developer wanted? Is A or B blindingly and obviously worse/better to the developer? Which is more important - what you think you like/prefer or what you think the developers intended? Um... maybe we can substitute the word "important" with "correct"...

It's a big can of worms :)

Anyone and everyone can have their own preferences. The question is whether their subjective tastes constitutes :

A) This is what I know is correct/intended ; or
B) Correctness be damned, this is what I prefer

I have so much more to say (and respond based on many good points raised by Shark, Joe and WaltC) but at 1.40AM I really prefer to have a good night's sleep :)
 
What a great thread this has become...... This is why I really love this site...... Nowhere else I know of would this go on without getting abusive and personal.....
 
Shark,

I dont think Democoder is talking about IQ differnces of in the order of V3 screenshots from UT2003 in 16bpp bilinear 256k textures non-AA'd v Radeon9700Pro 1024k textures 16xTri AF 6xAA but rather quantatively measuring when the quality delta is much much smaller i.e. as a GfFX v 9700pro should be, though 4xOGMS v 9700 (4xPJMS is it now?) should be discernble from decent screenshots.

Is the GfFX AA gamma corrected as well?
 
Reverend said:
Dio, which of the Quake2 images that Shark posted represents more accurately what the developer wanted?
Click to expand...
That's probably the $64k question. Terribly hard to get the answer.
 
Yes but where is the gamma correction applied, as if its after the frame buffer then only taking a actual digital picture of the monitor would show any differences...
 
Doomtrooper said:
I don't believe any of the 9700 's FSAA shots published so far are accurate either with the gamma correction feature of the 9700..
Click to expand...
I disagree: Recall the pinwheel shots that were posted here before. The 9700 Pro had much better results than non-gamma corrected results.
 
I was hoping you would answer opengl_guy..as I wasn't sure..so gamma corrected AA on the 9700 is captured then...capturing images from the frame buffer using Hypersnap (my preference) or fraps properly displays what we see on our screen.

I do remember that test, but wasn't sure if it was 100% accurate.
 
Well

my feeling is that although some form of IQ is always useful, Xabre series anyone?, it shouldn't be the primary focus of the review. The problem is with 3D graphics is that the closer we get to reality perfect imagery the more we notice any inconsistencies.

When reviewing cards everything needs to be taken into account. How each aspects is weighted I'm not sure as different people find different aspects of the card as important.

However for what it's worth I really like the reviews here, if I want a shootout style review I can go to any of the other hardware sites. Very few have the depth and intelligence of B3D.
 
However for what it's worth I really like the reviews here, if I want a shootout style review I can go to any of the other hardware sites. Very few have the depth and intelligence of B3D.
Click to expand...

On the contrary, IMO, it's because B3D is one of the few with enough depth and intelligence to actually do an informative "shoot-out," is exactly why I'd like to see B3D do them. ;)
 
Doomtrooper said:
I don't believe any of the 9700 's FSAA shots published so far are accurate either with the gamma correction feature of the 9700..I believe its applied after the frame buffer so screen captures from the frame buffer would not show the true output ?? :?: [/img]
Click to expand...

As long as the same hardware is used to create the image that is displayed as the hardware that produces a screenshot, the gamma correct FSAA should definitely be visible in a screenshot.
 
Status
Not open for further replies.
Back
Top