Ideal 9700 review

[pelly]

If someone here at nV News were to get ahold of a 9700, what specific areas would you like to see focused on. More than likely, the card would be tested against a Ti4600 ( standard and overclocked ). Essentially, one would be looking for strengths and weaknesses....

What would make the ideal review for us???

:confused:

[Bigus Dickus]

Personally, I've always thought it would be nice to conduct a review (or at least part of a review) as follows:

(1) Choose a representation of games to benchmark on (um... like always)
(2) Determine what the "desired framerate" level is for smooth gaming, like 60fps or whatever is deemed appropriate
(3) Change resolutions and IQ settings to find perhaps two or three combinations for each card in each game that yields approximately the desired performance level
(4) Present screenshots for comparison of what each card "buys" you interms of IQ at the desired gaming performance

A review done in this manner would thus focus less on the absolute speed of each card, and instead show the gamer just what image quality differences there are between the cards at comparable performance levels.

For example, suppose that both cards can run QIII @ 1600 x 1200 with IQ maxxed, FSAA maxxed, and AF maxxed at the desired performance level (I'm not saying they can...). Screenshots then reveal that the AA and AF IQ differences are fairly small (they may not be, but this is hypothetical). In that case, the two cards would more or less "tie" in this benchmark, since they both deliver at the desired performance level with maximum image quality. If you are a gamer concerned only about QIII, then you would know that for you, the extra money is not buying you anything (which might not be so obvious looking at graphs showing one card scoring double what the other one is... what difference does 200 fps vs. 100 fp make?).

OK, now most games (and probably QIII) are not like that. At some point the framerate will drop below the "threshold." Screenshots will then show what each card offers to the gamer at the "smooth gaming" performance level. If you as a gamer don't see much difference (as in, if you aren't particularly bothered by aliased textures or geometry) then you might decide that the impressive looking graphs aren't all that impressive after all. I mean, if the reviewer is telling you that a $125 video card will run your favorite game just as fast as that $400 video card, and you don't see that much difference in IQ, then why should you spend the extra cash?

And, if you really do care about image quality, then the review will show you just what you are most interested in: how much better will things look for the extra hundred or two bucks? You might take one look and the screenshots and think "Holy crap, that's beautiful," in which case you have your answer... for $400 that picture can be yours at smooth performance levels.


Well... that's just my opinion anyway. I've never really seen a review approached in that manner. They always simply show framerates at what is supposed to be equalivalent image quality (though, it hardly ever is, and they hardly ever show screenshots at the settings they are testing for comparisons). Approaching it from an "equal performance" instead of "equal quality" angle would more closely mimic how the actual gamer uses the card: fiddle with resolutions and settings until they find a combination that gives them maximum image quality without dropping below their preferred "threshold" of performance.

Why shouldn't a review do the same?

[SnakeEyes]

I agree wholeheartedly with bd on this one. Too much is either made of fps or of IQ, but usually only basing reviews wholly on maximum framerates or framerates with best IQ. It's rare to see reviews show each piece of compared hardware using the best settings for each game tested for each piece of hardware, and compare screenshots, as well as allowing the readers to make their own conclusions by allowing them to look at those same screenshots.

[The Baron]

BD is a GENIUS!

Quick, somebody send him a card :p

[-=DVS=-]

Well that is simple
dumb all benchmarks and use real games at same conditions you play every day :rolleyes: no need brains to figure this out

lets say moust common 1024x768 with max Aniso and 2xAA and 4xAA then test it at other popular res 1152x864 with max aniso and 2x,4x AA with max in game quality also :rolleyes:

offcourse higher resolution is good also if we have enough horse power :p 1280x1024 and 1600x1200 :rolleyes:

Games to test
SOF2 :D
Quake 3 :rolleyes: ehh who plays it anyways
DAOC , Red Faction , HalfLife :eek: , RTCW , and whatever other games :rolleyes: WC3

[Gator]

i'd like to see some other games added to the list of testing: Alien vs Preditor 2 & Unreal Tournament (old but still very popular) in both opengl and direct3d modes

[Spectral]

Yes, UT

I have played a few maps under normal conditions and Ti4600 performance in UT is far superior compared to my 9700..

Avg 100fps 9700Pro
Avg 140fps Ti4600

Dont ask me how..

[pelly]

Strange....is the Ti4600 overclocked?

:confused:

[saturnotaku]

Spectral, you using OpenGL or D3D for UT?

If it's OpenGL, NVIDIA has always been strong with that API. I wouldn't be suprised if ATI addressed anything like this in a future driver release.

[malachi1313]

I would really like to see image comparison shots while using Cyberlinks Power DVD.Also would like to see detailed image comparisons of just Anistrophic and a seperate one for just Maxed out AA. Everytime you see a screen shot lately it is with both enabled. I would like to see them seperated. Thanx!!!

[DadGT]

The game I currently put the most stock in when looking at reviews is Serious Sam: The Second Encounter. This game can really look spectacular. Especially with the special "extreme quality" setting. And it will output an "instaneous" frame rate to look for min frame rates and other strange performance trends. And it still taxes current hardware.

[Spectral]

Cyberlink PowerDVD does not work anymore since I got the 9700 Pro... Its just a garbled mess and then it locks up.

Not sure if its something on my end.. but I had just watched a DVD a few hours prior to installing the card and then it no longer worked.

As for UT, yes.. it was overclocked 312/730 ...

The 9700 Pro hits a lot higher framerates than the Ti4600 ever did, but they are not consistant at all.. going from 250 to 70 fps in half a second is normal...

The Ti4600 was always a consistant framerate and never ever dropped below 100 fps in the vast majority of maps I played.

[xamfear]

In addition to the standard testing methods. Do double blind tests with several people and let them rate the various cards for quality, smoothness, etc.

[pelly]

In addition to the standard testing methods. Do double blind tests with several people and let them rate the various cards for quality, smoothness, etc.

Great idea!!! For my next review, I'll pull my fiance into the computer room ( kicking and screaming ) and ask her which screen looks prettier! :D

Honestly though, that could be a good idea......we'll see if we can pull it off....

Any more suggestions?......there is a point to all this.............

:)

[legion88]

Originally posted by pelly


Great idea!!! For my next review, I'll pull my fiance into the computer room ( kicking and screaming ) and ask her which screen looks prettier! :D

Honestly though, that could be a good idea......we'll see if we can pull it off....

Any more suggestions?......there is a point to all this.............

:)

It could be difficult in getting enough people to participate. A "few" isn't worth it.

This is what I suggested to RivaExtreme (and others) regarding how to do a real FSAA comparison screenshots a few years ago.

Show pictures of FSAA shots from various cards but DO NOT INFORM the readers which cards were used to generate the pictures. Just ask the readers to vote on which shots look better. After the poll is finished, then you can inform the readers the winning card.

Back in the days when FSAA comparison shots were the rage, bogus polls were put up where fanboys picked their favorite card rather than pick the screenshot with the best FSAA. A true blind poll would have negated much (not all) of the fanboyism. Naturally, fanboys hated the true blind poll.

You can add blind polls at the end of the review for FSAA comparisons, aniso comparisons, and FSAA with aniso comparisons. How many polls do you see at the end of reviews? At the very least, the review--if you plan to do one--would be different.

And for quality-control, make sure you actually have pairs of IDENTICAL screenshots (i.e. the two screenshots are actually the same screenshot with different name).

In theory, when groups of people are asked to pick which looks better, the two shots will receive a roughly similar scores. But to those looking to cheat at the poll, they wouldn't know that. Scoring huge on one of the identical shots but not the other. So basically, each screenshot that you put up for comparison comes equipped with an identical twin.

Naturally, some pairs are more easily identified than others but that is to be expected. It is especially true if the screenshot contained some graphical anomalies. You should avoid anomalies as much as possible.

So as an example, suppose you have four video cards: Radeon 9700, GeForce4 4600, Radeon 8500, Matrox Parhelia.

You can four screenshots with aniso enabled, one shot per video card. Let's call them shot 1, shot 2, shot 3, and shot 4. For each screenshot, make a copy so now you have shot 1b, shot 2b, shot 3b, and shot 4b in addition the previous four.

So in the poll, the readers would have 8 choices to pick from. And they can only pick one. You simply ask the readers which screenshot looks better.

When the poll is finished, shots 4 and shots 4b would have roughly similar score. Shots 3 and 3b should have roughly similar scores. Etc., Etc.

You could say in the review that the aniso shot of the radeon 9700 looks superb. Best of all the cards you've seen so far. The poll results then can be used to see what others think. And since the readers do not know which video card generated those shots, they will have a harder time playing favorites. The readers in the end can agree with your assessment on the quality or not. Update the review with the results.

[Bigus Dickus]

I think that would be a nice addition to a review, but I wouldn't substitute that "blind screenshot" method for a straightforward "this is what card X looks like, this is what card Y looks like" approach.

I still think baseline performance comparisons instead of baseline IQ comparisons (which are difficult to achieve) would be very useful.

[StealthHawk]

i'm not so sure a blind poll would be a good idea in a review. the whole purpose of a review is to find out how new cards stack up to each other. as such, i would want to know immediately what card x looks like compared to card y.

the blind screenshots however are promising, but the answers should be somewhere in the review for those who don't want to wait for the answers to be revealed.

i also believe that most people wouldn't care one way or another what other people think looks better. it's your money: whatever you think looks best is the only thing that matters.

[legion88]

Originally posted by Bigus Dickus
I think that would be a nice addition to a review, but I wouldn't substitute that "blind screenshot" method for a straightforward "this is what card X looks like, this is what card Y looks like" approach.

I still think baseline performance comparisons instead of baseline IQ comparisons (which are difficult to achieve) would be very useful.

A blind poll would be something at the end of the review, not somewhere in the middle. This means that screenshots of how the card(s) render(s) images would be available prior to the poll.

These screenshots, however, must not be the same shots in the poll.

Where did you get this "substitution" idea from? In fact, I clearly stated:


You can add blind polls at the end of the review...

[Bigus Dickus]

Originally posted by legion88


A blind poll would be something at the end of the review, not somewhere in the middle. This means that screenshots of how the card(s) render(s) images would be available prior to the poll.

These screenshots, however, must not be the same shots in the poll.Yes, that sounds like a good idea. I was just clarifying my opinion on the matter, becuase it was a bit vague whether you meant to only use the blind SS or use them as a tag-on.

Where did you get this "substitution" idea from? Probably from this statement:

"Show pictures of FSAA shots from various cards but DO NOT INFORM the readers which cards were used to generate the pictures."

Even though you said they could be at the end of the poll, you didn't mention whether there would be normal labeled comparison shots or not. Since many reviews have no screenshots, and most of the ones that do tag them on at the end of the review, I just wasn't sure what you meant.

Thanks for clarifying, I agree, would be a nice addition.

[legion88]

A few years ago, I did a "poll" of sorts over at 3DGPU in the forums. The four cards that I had were the GF2, Radeon 256, G400Max, and the Voodoo3. I used q3 and took in game shots.

The "poll" asked which image look better, and the amount of FSAA used (e.g. 2x, 4x). I listed the video cards used but I did not mention which video card created which image.

It was not a real poll as participants had to write a post to make their choice. So this leads to discussion, which in turns affects other people's decisions...leading to more discussion. Anyway....

People quickly picked the Voodoo3 as the worst and the one without FSAA. Obviously, the card's inability to support higher resolution textures was a dead giveaway.

Would it surprise anyone that one of the so-called 'experts' on FSAA and a 3dfx fan thought that the G400Max screenshot had 2x FSAA enabled? The guy used to run one of the hardware sites and was a hardware reviewer. The site was alleged to be a non-biased site when the reality was the opposite.

In the end the GF2 had the lead but there was no actual winner. Not enough participants to really say who won.

Hardware sites are, generally, shying away from true objective analysis of hardware. Objectivity tends to hurt their abilities to get "freebies" and most importantly, the ability to attract a loyal following of fans. Even benchmark results are fishy (see Tom's Hardware as an example). So I'm not expecting too many blind polls--if any.