First few GFX benches

[Sharkfood]

There is a huge difference between no-AA and 2x/4x AA. But as you go higher and higher, the improvements in quality at each level aren't as dramatic. Some people may have hypersensitized themselves, but the vast majority of people won't really notice. Unless some vendor makes a huge leap to 16x or 64x sampling, I doubt we will get that "wow" factor again.

This sounds more to me like a precursor excuse for trying to discount 4x and 6x AA scores for a favored IHV that may soon follow.

I can't fathom how anyone that has actually used AA could say 4x or 6x AA has no stark or dramatic difference compared to 2x. Obviously, if you look long and hard for a 3+ year old game with fairly low contrast and minute details, you might find one that shows little benefit above 2x AA, but for most all modern games, there is a stark and incredible improvement over 2x AA.

Thus, we are going to go through another round of people 16x zooming into screenshots and arguing over minute aliasing details that 90% won't notice or care.

Zoomed screenshots are usually used to identify problems that aren't readily visible in still screenshots, but are starkly noticeable in motion. Edge crawling in motion just doesnt show up in still shots. It's either zoomed screenshots or people start capturing 500MB MPGs at true resolution to illustrate the problems.

As far as 90% wont notice or care.. not care I'd believe since, depending upon IHV brand recognition, people over time have proved themselves to artificially not care about things if they effect their favored IHV. Notice? I can only assume you mean people that have trained themselves long enough to overlook the blaringly obvious and absurdly visible long enough to truly not notice things anymore. I've seen this behavior with rocket trails in benchmarks being completely missing, horrible banding in skies, and poor performance with extreme dips in framerate. It's amazing what enough positive re-inforcement can burn into someone's ability to overlook the starkly obvious.

 

[Jamm0r]

To be perfectly honest with you, the jump in quality from 2x to 4x on both my 9700 Pro and Voodoo 5 is MUCH more dramatic than the jump from no AA to 2x. This is also true on the 8500, but mostly because of the unfortunate vertical 2x AA issue. As a matter of fact, 4x (and especially 6x) on the 9700 is so good that I somehow, if having to choose between No AA and 2x AA, would definitely choose the former over the latter on the 9700.

 

[Reverend]

I post one sentence in this forum in a while and we get responses such as those by Hell and Doom (hmm... the two of you must belong to the same cult judging by your names...).

FWIW, John's "guess" in response to my post was what I meant. With the difference in grid patterns between the two products, the IQ differences should prevent such "4xAA performance shootouts" but I'm sure most websites will do so anyways.

As for Doom's post about my lack of coverage/articles on non-NVIDIA companies/products, this has been discussed before and I will leave it to those who think my opinions matter to make their own decisions about whether my "reasons" are/were legit or not.

Oh, and ATI's dev rel did "suck" at the time I made those comments (based on comments by the various developers I talked to back then). Not anymore it would appear. And ALL PR suck. There was once when I posted on 3DPulpit that PR and Marketing are evil and it was in reference to the way Derek Perez handled a particular situation involving myself (it was during the time of the original GeForce). Some people take what I say/said too seriously... when Brian Burke (still with 3dfx then of course) read what I wrote (about PR being evil) he emailed me with "Well then, I guess the honeymoon's over for us... " . The smiley was his. When you are all grown up and mature you know how to handle things well/appropriately when you know you are talking to someone you understand clearly. People like Hell and Doom are not such examples IMO.

It's just too bad things didn't work out between VE and myself. I had hoped that my 9700Pro review would be useful to readers (my conclusion was a simple "The Radeon 9700Pro is IT") and how my other 9700Pro-vs-GF4Ti4600 performance+IQ article showed examples (Mafia, F1 2002) where the 9700Pro wins in almost every AA situation (jaggies and texture quality). Maybe Dave will give me back a reverend@beyond3d email and my efforts won't go to waste.

 

[Doomtrooper]

I think you need to read my replies to you before you go off on your insult tirade (cult..lol) Reverend, I was following Johns train of thought also...quite obvious by looking at my posts.

As said before, you are a competent reviewer and my only problem with you as a technology reviewer was the constant one IHV favortism for whatever reason (lack of contacts...whatever).

As for the AGE comment..I'm older than you.

 

[martrox]

Yea......well, I'm older than both of you....and that really sucks....... I think everyone needs to take a step back and try to look at what we might all agree on...... PR sucks, guessing at benchmarks suck, being so blinded by one's allegence to one manufacturer not only sucks, but is stupid, too. It keeps you from really enjoying the fruits of everyones labors. While I have never been very keen on how nVidia conducts itself - and never have been shy about mentioning it - fact is I've owned many of their products, and will probably own more in the future. ATM, I own ATI, because they have the best products, by far. No amount of fanboi BS will change that. 6 months ago, I owned nVidia, because they had the best products, and, again, no amount of fanboi BS can change that. I really do try to be objective about all products. That way, I can make an educated purchase, and not miss out on the really good stuff because it's made by XYZ, and they suck.......... oh, and by the way, I really don't care, at the moment, about some benchmarks that someone has pulled out of their nether regions......... because it's still vaporware, and will be until it's on the shelves. Then, we can read all about how good/bad it really is....... And Rev, please find someone to publish your 9700 review. While I may not always agree with you, I have the upmost respect for you and I always come away from a review of yours feeling like I've learned something of value about the product.

 

[DemoCoder]

I can't fathom how anyone that has actually used AA could say 4x or 6x AA has no stark or dramatic difference compared to 2x. Obviously, if you look long and hard for a 3+ year old game with fairly low contrast and minute details, you might find one that shows little benefit above 2x AA, but for most all modern games, there is a stark and incredible improvement over 2x AA.

I have a R300 PRO and for me, there is no "stark" difference akin to no-AF vs AF, or pointsampling vs bilinear, or no-AA vs some-AA. Perhaps you have a different definition of "stark", but for me, it means you don't need to squint, and you will instantly realize the difference.

Primarily examples: Q3, Counter-Strike, and BattleField 1942.

I'd love it if someone would do the "pepsi challenge" with a focus group in front of real live audience, with identically calibrated systems, etc. My gut instinct would be that majority of people wouldn't notice the difference, just like most people can't tell the difference between a good audio system and a great audio system, or MP3 vs CD.

My point is, 90% of the people on this planet won't notice the difference between 4X, 4XS, pseudo-stochastic 4X, gamma or non-gamma connected AA. They won't unless you specifically point out the minute differences.

I know people who honestly don't see the big difference between VCD, SVCD, and DVD, even though to me, all of them have major artifacts on seens with a large contrast ratio.

Spin that however you wish as a pro-NVidia statement. I'm not gonna be around during the screenshot fest that's going to occur next month, since I am sick of subjective arguments that have zero impact on what people are going to buy. I have consistently maintained that both the NV30 and R300 are great chips and are roughly "on par", and the attempts by people to blow the differences between them way out of proportion are just getting tired.

 

[Hellbinder]

The same Cult.. The truth has been discovered...

FWIW, John's "guess" in response to my post was what I meant. With the difference in grid patterns between the two products, the IQ differences should prevent such "4xAA performance shootouts" but I'm sure most websites will do so anyways.

What i dont understand is.. Why not? differences in technology and approach are there to give one company an advatage over the other. This is why we test at standard settings like 4x or 2x. To see what each company offers in performance/quality and to come to a decision on which is better for the money.

I just do not understand the mentality of not comparing features because of varried approaches. Its what the entire enchalada is about.

Now, Either.. the Nv30 has a better sample pattern and its slower, the Nv30 has an inferior pattern (iq wise) but its faster.. or its better or worse in both areas. Right?

So the Question remains.. which is it? Why dont you think the Nv30/R300 should be compared at 4x? and if not.. then what can they be compared at? I suspect that Nvidia has totally cheesed out and sacrificed Quality for performance. So they win some more benchmarks. Thats my guess. Which is what they have a history of doing.

Should make their bar graphs look a lot better eh?

 

[Doomtrooper]

Martrox,
Unless you are in your mid 40's I doubt it....I'm sure Dave will do something if not to many Bridges are burned.

 

[martrox]

Oh, Demo, CDs suck too...... vinyl is where it's at, but you need valves (tubes to the barbarians) to really appreciate it And my old, abused 52 year old ears can hear the difference........ of course, the top end isn't as crisp as it was 20 years ago.......

BTW, Gold Lion/Genelux rule

 

[Reverend]

Hell, please re-read what I wrote. I wrote that there should be obvious AA IQ differences between a R300 and NV30 at 4xAA and hence any 4xAA performance comparisons between the two shouldn't be done. I have no doubt that the R300's 4xAA will look better than the NV30's 4xAA. My stating the grid pattern difference between the two at 4xAA (OG vs RG) merely serves to say all of the above.

 

[martrox]

Martrox,
Unless you are in your mid 40's I doubt it....I'm sure Dave will do something if not to many Bridges are burned.

Mid 40's....I wish........ turned 52 in November.......

 

[Doomtrooper]

I have a R300 PRO and for me, there is no "stark" difference akin to no-AF vs AF, .

Come on coder you are trying to tell me you can't tell if Anisotropic filtering is on in BF 1942 or say RTCW???

I'll take the Pepsi challenge, and I'm sure many that chose the "choice of a new generation' could tell which pic above had slight AA and AF on :?

 

[Doomtrooper]

Martrox,
Unless you are in your mid 40's I doubt it....I'm sure Dave will do something if not to many Bridges are burned.

Mid 40's....I wish........ turned 52 in November.......

Wohooo..I don't feel like Grandpa no more..congrats gramps you are the new elder.

 

[Hellbinder]

Spin that however you wish as a pro-NVidia statement

There is no need to spin your statements at all IMO. You have done a fine job making the case yourself. Suddenly a select group of people here are all about the superiority of 2x AA, and how we cant compare 4x FSSA between products anymore..Imagine that....

I own a 9700 also.. and there is a HUGE difference in some games between 2x FSAA and 4x FSAA. but admitedly in others not so much. It all depends on how much contrast between color patterns there are on the edges. Strong light on dark clashes and its very noticable.. very blended and similar colors not so noticable. I also admit that even I have a coupple friends that have no flipping idea what im pointing out to them when i show off FSAA and AF. But im still not going to change my position in mid stream and say we dont need 4x fSAA anymore.. (intentional exageration)

Funny thing is.. EVERYONE notices the purdy pixel-shaded water in games tho.. heh..

 

[DemoCoder]

The point HellBinder, is to test cards at nearly equivalent IQ and see how fast they run. Imagine if the NV30 didn't have anisotropic filtering at all, but was 2x faster than R300 at the same res. You'd scream bloodly murder that it's not a fair comparison. Imagine comparing a 32-bit precision pipeline to a 16-bit one that runs 2x as fast. Ignoring the artifacts from 16-bit, is it fair?

What you really want to benchmark is not IQ and not performance but IQ / Performance. Of course, that assumes IQ could be "quantified", which it can't. And of course, there are oodles of problems even with subjective evaluations, since IQ might look good on one game, but bad on another, depending on what shortcuts and hacks the vendor took to get performance.


In a perfect world, there would be a standard "test scene" rendered by a software rasterizer to some standard level of IQ agreed upon by everyone. Hardware would be measured by how far their result is from the standard image. I think we're along way from something like that.

And I doubt everyone will agree, since every company is willing to trade off some accuracy for performance. They just differ on where they want to do the tradeoffs.

 

[demalion]

That's why a good comparison review is more than just a collection of graphs. Whether it is hard or not, difficult or not, prone to disagreement or not, it is what is worth doing. IMO, atleast.

 

[DemoCoder]

Doomtrooper, I said there was no stark difference in each incremental AA level like THERE IS (a difference) with no-AF vs AF. That means THERE IS A STARK DIFFERENCE WITH no-AF vs AF. I threw in the example of point sampling vs bilinear too.

I was making the point that no-AF vs AF has a "wow factor" that is immediately, and uncontrovertably obvious to anyone. The finer gradations of AA levels are so sublime that it takes hardcore enthusiasts to the point of taking zoomed screenshots to try to show the difference. And plenty of the so-called experts on AA IQ on these forums have gotten Rev's and others various Polls and Tests of mystery AA modes wrong in the past, clearly showing that it is in fact, not blantantly obvious which modes are better.

 

[Reverend]

As said before, you are a competent reviewer and my only problem with you as a technology reviewer was the constant one IHV favortism for whatever reason (lack of contacts...whatever).

Please substitute "favortism" [sic] with "reports". I review what I am given. If VisionTek continuously gives me their stuff without my asking, I review their stuff. If I am not given hardware by non-NVIDIA companies even though I have asked many times, there's nothing I for me to write about.

Look not for who I report on but what I report on and whether I report these honestly or not. That should be what matters most.

You should have problems with me for writing dishonest reviews/etc but you should not have problems with me for writing (to the best of my abilities) honest reviews/etc of only NVIDIA and/or its products.

 

[DemoCoder]

]

Spin that however you wish as a pro-NVidia statement

There is no need to spin your statements at all IMO. You have done a fine job making the case yourself. Suddenly a select group of people here are all about the superiority of 2x AA, and how we cant compare 4x FSSA between products anymore..Imagine that....

Don't you realize that Rev's position (and the one I am agreeing with) benefit the R300 and not the NV30? Rev isn't saying that you can't compare them because it is unfair to NVidia, he is saying you can't compare them because it is unfair to ATI.

Rev is basically saying that in the benchmarks, you should compare the framerate numbers of ATI's 4x mode with some higher NV30 mode like 6xS or 8X to do a fair comparison.

To directly compare the framerate of NV30 @ 4X vs ATI @ 4X, since NV30 will be faster, is unfair to ATI because ATI's AA might look as good as NVidia's 8X.

I happen to agree with the idea that identical (or as close to identical) IQ frames should be compared, and then once the settings are calibrated, then you compare the framerates.

You seem to imagine that every word that comes out of my mouth must somehow be a pro-NVidia statement. If's Rev's advice is to be taken, it will make the NV30 look bad.

That said, like I said in other messages, unless you can convince every reviewer to use this standard, it won't matter a hill of beans to the consumer, because the consumer just looks at the standard benchmarks people run, and looks at the price, and does not spend hours pouring over zoomed screenshots.

Rev's advice, if everyone followed it, would make it easier for ATI to market IQ to the masses.

(of course, how would be synchronize the AF implementations, so that any artifacts from ATI's shortcuts don't get them off the hook either)

 

[Reverend]

As for the AGE comment..I'm older than you.

I wasn't talking about age but since you brought it up, I thought I'd mention how childish I find my father (who is older than me, of course )at times.