[H]ardocp Review for the Geforce FX

[KimB]

I play all of my games at 1600x1200 w/ 4x AA and 8 or 16x AF on my 9700 Pro. I think I'll take high res with more AA, thanks for asking.

And what games am I playing? Hitman 2, SoF 2, Warcraft 3, Undying, and Counterstrike. None of these games have problems with the settings I gave above. In fact, I'm still CPU limited in many cases because of my Athlon XP1900+.

I was mostly referring to more fillrate-limited games that you can't play acceptably at that resolution, such as Unreal Tournament 2003. And you can't say that the 9700 Pro would be better for Counterstrike, at least (I don't know about the fillrate limitations of the others...), as 8xS should be very playable in that game.

 

[OpenGL guy]

And you can't say that the 9700 Pro would be better for Counterstrike, at least (I don't know about the fillrate limitations of the others...), as 8xS should be very playable in that game.

Except that 8xS uses an ordered grid which may not help Counterstrike much given the relatively primitive geometry.

 

[demalion]

Some people simply don't agree with you that the GF FX aniso is better than that on the 9700. Your entire argument is predicated on that being indisputable, not just a personal opinion (not that I consider your well-established opinion unbiased, but I consider it possible that someone not demonstrating the bias you have might share it).

Based on the GeForce FX's anisotropic being no worse than the GeForce4's (which is probably a very good assumption), the FX's anisotropic is absolutely better than the Radeon 9700's at the same level of anisotropy.

I repeat, "some people simply don't agree with you that the GF FX aniso is better than that on the 9700. Your entire argument is predicated on that being indisputable, not just a personal opinion (not that I consider your well-established opinion unbiased, but I consider it possible that someone not demonstrating the bias you have might share it."

Some people are crazy and when comparing:
GF FX at 8x
and
9700 8x

are daft enough not to conclude that the GF FX is "absolutely better".

Some people, when they combine that with the 9700's AA, might actually come away thinking the 9700 had clearly better image quality, and might have some faint ability to compete with the GF FX in high resolution settings with aniso and FSAA (low or high). In fact, they might look at you strangely when you term the decision between the two cards as a choice between having "Higher resolution, or more FSAA?" when they can have both.

Even ignoring comparing 9700 8x to GF FX 8x, and focusing on comparing only 16x on the 9700 which you can atleast admit is "debatable", let's look at some example performance differences between 8x and 16x on the game that demonstrates the largest hit that I know of for the 9700: UT 2k3.

I don't think 84.18 compare to 87.34, 59.94 compared to 62.90, or 45.15 compare to 47.64 (as resolution increases) disqualifies comparing 16x ATI aniso to 8x GF FX aniso, do you?

 

[Reverend]

Good read, Brent.

 

[BenSkywalker]

Except that 8xS uses an ordered grid which may not help Counterstrike much given the relatively primitive geometry.

CS also uses a lot of alpha textures, I'd take 4x SS over 16x MS in CS when looking at AA.

Do any of the reviews mention anything about aliasing issues in motion? I see lots of comments that have everything to do with a more agressive LOD bias, but nothing about texture aliasing.

 

[KimB]

And you can't say that the 9700 Pro would be better for Counterstrike, at least (I don't know about the fillrate limitations of the others...), as 8xS should be very playable in that game.

Except that 8xS uses an ordered grid which may not help Counterstrike much given the relatively primitive geometry.

Update: I guess 8xS wouldn't be usable, given a 128MB FX at 16x12. 4xS should be very playable, though, and that mode is very close in edge quality to ATI's 4x. 6xS is pushing it at that resolution.

 

[KimB]

Some people are crazy and when comparing:
GF FX at 8x
and
9700 8x

are daft enough not to conclude that the GF FX is "absolutely better".

None of these websites have adequately-examined the image quality of the FX, but the algorithms for the Radeon 9700 Pro and GeForce4 are well-known. I'm assuming that the anisotropic algorithm did not worsen from the GF4 to the FX, which seems a reasonable assumption. If true, then the Radeon 9700 Pro is fundamentally inferior to the GeForce FX when using the same degree of anisotropy. This is for two primary, objective reasons:

1. Angle problems. While improved significantly from the 8500, the Radeon 9700 Pro still has problems with off-angle surfaces.
2. MIP map level selction. The GeForce line's is more accurate, meaning greater texture clarity/aliasing ratio.

And the third, subjective reason:
3. The Radeon 9700 Pro displays more texture aliasing.

If you are basing your assumptions that the 8-degree anisotropic of the 9700 Pro is not any worse from those images, then you are far from looking at a comprehensive view of the texture filtering methods of the respective video cards. Those images include low-contrast textures, and no apparent off-angle surfaces, making the comparison rather pointless.

In fact, they might look at you strangely when you term the decision between the two cards as a choice between having "Higher resolution, or more FSAA?" when they can have both.

And on games in which you can have both on the Radeon 9700 Pro, you can also have both on the GeForce FX. But the main thing here is newer games that are more fillrate-limited, such as Unreal Tournament 2003, and, likely to a greater extent, the upcoming DOOM3. In these games there will be a tradeoff. Games where either can play at high resolutions with FSAA are pointless to consider.

 

[SvP]

I'm assuming that the anisotropic algorithm did not worsen from the GF4 to the FX, which seems a reasonable assumption.

Why do you assume that? I remember NVidia stating "faster anisotropic filtering" numerous times. Don't you? Therefore it's more correct to assume that the algorithm did change, isn't it?

None of these websites have adequately-examined the image quality of the FX

Could you tell me why this page didn't qualify as "adequately examining" for you:
http://www.anandtech.com/video/showdoc.html?i=1779&p=8

If true, then the Radeon 9700 Pro is fundamentally inferior to the GeForce FX when using the same degree of anisotropy

But what if not true? Look at this page: http://www.anandtech.com/video/showdoc.html?i=1779&p=9
It may be fundamentally inferior to you, but it certainly looks superior to me.

 

[Althornin]

But once you add in supersampling to the mix, the texture quality on the FX will become better. Certainly it's going to be better at 8xS FSAA, which supersamples in two directions, for the equivalent of 16x FSAA plus better filtering on closer textures. For 4xS and 6xS, it will at least look better for surfaces closer to the horizontal.

GAH!
where do you get this stuff?
8xS is 8 SAMPLES! NOT 16!!!!!!!
Its 4xOGSS + 2xRGMS.

I'm assuming that the anisotropic algorithm did not worsen from the GF4 to the FX, which seems a reasonable assumption

Nice, but flawed.
What do you think the agressive setting shows?
RE screenshots in antoehr thread, there is no real difference when using agressive on a GF4 - there is on a GFFX. Ergo, the algorithm HAS TO HAVE changed, at least for that setting. When you realize it ahs changed, at least in one place, it is folly to ASSUME it ahsnt changed in another.

 

[demalion]

Some people are crazy and when comparing:
GF FX at 8x
and
9700 8x

are daft enough not to conclude that the GF FX is "absolutely better".

None of these websites have adequately-examined the image quality of the FX, but the algorithms for the Radeon 9700 Pro and GeForce4 are well-known. I'm assuming that the anisotropic algorithm did not worsen from the GF4 to the FX, which seems a reasonable assumption. If true, then the Radeon 9700 Pro is fundamentally inferior to the GeForce FX when using the same degree of anisotropy.

If the supposition about the Balanced setting is true, for all the comparisons the algorithm has indeed changed.

This is for two primary, objective reasons:

You tend to abuse the word objective, Chalnoth.

1. Angle problems. While improved significantly from the 8500, the Radeon 9700 Pro still has problems with off-angle surfaces.

The existence of these problems is objective. Your evaluation of them making image quality on the 9700 inferior to the GF FX is not.
The last images illustrating these problem angles I saw showed, in my evaluation, a level of detail at these problem angles for 16x that looked to me comparable to 8x on the GF 4, and I don't even know if the displayed level of detail in current GF drivers are the same or not, let alone for the GF FX balanced mode. You, on the other hand, have no problem making subjective assumptions that fit your established viewpoint about superiority for the GF.

2. MIP map level selction. The GeForce line's is more accurate, meaning greater texture clarity/aliasing ratio.

What is the criteria for "more accurate"? I'm guessing it has to do with the way you think colored mip map level gradients should look. So, where has the connection between this and your concept of "greater texture clarity/aliasing ratio" phrase been established, and then shown to be to such a degree that a user's evaluation is that the GF aniso is "absolutely better"?

And the third, subjective reason:

Yes, I agree, this makes the third subjective reason.

3. The Radeon 9700 Pro displays more texture aliasing.

Well, I don't have a 9700...or a GF FX. So since you must have a GF FX, and you have had a 9700 and have never demonstrated unreasonable bias regarding it I suppose I'll take your word.

If you are basing your assumptions that the 8-degree anisotropic of the 9700 Pro is not any worse from those images, then you are far from looking at a comprehensive view of the texture filtering methods of the respective video cards.

The problem here Chalnoth is that you think you are.

Those images include low-contrast textures, and no apparent off-angle surfaces, making the comparison rather pointless.

I don't recall seeing an image where the 9700 didn't compare well to the GF in anisotropic filtering. Since the GF 4 image quality is "absolutely better", I'm sure you'll have no problem atleast providing sufficient examples demonstrating this clearly in contrast to those "rather pointless" images? Once you do that, we can cover the discussion about what "Balanced" aniso means, I guess.

In fact, they might look at you strangely when you term the decision between the two cards as a choice between having "Higher resolution, or more FSAA?" when they can have both.

And on games in which you can have both on the Radeon 9700 Pro, you can also have both on the GeForce FX.

But the 9700 looks better with AA, is the point.

Nice switch from "The FX's primary purpose will be to run at very high resolutions (1600x1200+) with lower FSAA, whereas the 9700 Pro would be better served by running at lower resolutions". Strange how the 9700 framerate compares well with the GF FX in such conditions, even for your stipulated fill-rate limited UT2k3, has superior AA, and the GF FX is the card suitable for such cases.

But the main thing here is newer games that are more fillrate-limited, such as Unreal Tournament 2003, and, likely to a greater extent, the upcoming DOOM3. In these games there will be a tradeoff.

The tradeoff where it is impossible for the user to hold an opinion contrary to yours concerning the relative quality of aniso implementations between the cards...yes, I know.

Games where either can play at high resolutions with FSAA are pointless to consider.

Hmm...so now games where you can play at high resolutions with FSAA are "pointless to consider"? I'm sure the game developers and users appreciate you laying down the law that the only games that count will be games performing at or below UT 2k3's average framerates in the range of 50 fps and above in those conditions.
If you've decided on the GF FX regardless of any other options, then yes, at this point it certainly looks like such games are "pointless to consider" by your reasoning, since the GF FX seems to have a weak point in the area of FSAA. You seem to be trying to pass your argument off as something other than a long excuse to justify making just that decision, and that is the only "objective" validity you seem to have demonstrated so far.

 

[KimB]

But once you add in supersampling to the mix, the texture quality on the FX will become better. Certainly it's going to be better at 8xS FSAA, which supersamples in two directions, for the equivalent of 16x FSAA plus better filtering on closer textures. For 4xS and 6xS, it will at least look better for surfaces closer to the horizontal.

GAH!
where do you get this stuff?
8xS is 8 SAMPLES! NOT 16!!!!!!!
Its 4xOGSS + 2xRGMS.

Typo. Get over it.

I'm assuming that the anisotropic algorithm did not worsen from the GF4 to the FX, which seems a reasonable assumption

Nice, but flawed.
What do you think the agressive setting shows?
RE screenshots in antoehr thread, there is no real difference when using agressive on a GF4 - there is on a GFFX. Ergo, the algorithm HAS TO HAVE changed, at least for that setting. When you realize it ahs changed, at least in one place, it is folly to ASSUME it ahsnt changed in another.

I didn't assume it hadn't changed. I assumed it hadn't worsened. There's a difference there, and it will require a much more in-depth look.

And yes, not one of the previous out right now have adequately examined the anisotropic image quality. In particular, all that should be necessary are some colored MIP map shots comparing balanced to aggressive.

 

[SvP]

I didn't assume it hadn't changed. I assumed it hadn't worsened. There's a difference there

And why exactly did you assume that, please?

 

[KimB]

You tend to abuse the word objective, Chalnoth.

1. Angle problems. While improved significantly from the 8500, the Radeon 9700 Pro still has problems with off-angle surfaces.

The existence of these problems is objective. Your evaluation of them making image quality on the 9700 inferior to the GF FX is not.

But the 9700 is certainly inferior when comparing the same degree of anisotropic. If you want to talk about 16-degree vs. 8-degree, than I have stated many times that I felt that was subjective, but I side on the side of the GeForce4 (and presumably the FX).

What is the criteria for "more accurate"? I'm guessing it has to do with the way you think colored mip map level gradients should look. So, where has the connection between this and your concept of "greater texture clarity/aliasing ratio" phrase been established, and then shown to be to such a degree that a user's evaluation is that the GF aniso is "absolutely better"?

No, not gradients, border shapes. The correct selection gives a sort of semi-circular pattern, which the GeForce series displays. The Radeon 9700 Pro does not. But, this is a relatively minor point, and will almost always not be noticeable.

I don't recall seeing an image where the 9700 didn't compare well to the GF in anisotropic filtering. Since the GF 4 image quality is "absolutely better", I'm sure you'll have no problem atleast providing sufficient examples demonstrating this clearly in contrast to those "rather pointless" images? Once you do that, we can cover the discussion about what "Balanced" aniso means, I guess.

Try looking at some shots in the Serious Sam: SE technology test sometime. Notice how the black brick borders on the ground look more jagged on the 9700 shots. This is one example. More noticeable for me was in Neverwinter Nights on a similar texture. This effect is far more noticeable in motion than it is in screenshots.

But the 9700 looks better with AA, is the point.

Nice switch from "The FX's primary purpose will be to run at very high resolutions (1600x1200+) with lower FSAA, whereas the 9700 Pro would be better served by running at lower resolutions".

The AA is identical at 2x, and neither is playable at 4x FSAA at 1600x1200x32.

Hmm...so now games where you can play at high resolutions with FSAA are "pointless to consider"?

Yes, performance-wise, because it's going to be higher than it needs to be for good playability on either card. It only makes sense to consider games (in performance comparisons) that actually stress the video cards.

 

[KimB]

I didn't assume it hadn't changed. I assumed it hadn't worsened. There's a difference there

And why exactly did you assume that, please?

When have you seen a successor video card that has done that?

But seriously, the main reason is that it would require more work to worsen the implementation than keep it essentially the same.

 

[LeStoffer]

But seriously, the main reason is that it would require more work to worsen the implementation than keep it essentially the same.

Let's hope so, but I won't bet the farm on it yet. I remember a program shown here that showed how AF is implemented on different cards and ATI does some tricks that nVidia in the past wouldn't accept as the correct way to do it. But they had to cut some corners to get where ATI are today, so the jury will be out on that one for another month.

 

[christoph]

as it was mentioned earlier, heres a pic of the passive cooled sapphiretech 9700 pro

http://www.sapphiretech.com/VGA/9700proult.asp

 

[Snyder]

What is the criteria for "more accurate"? I'm guessing it has to do with the way you think colored mip map level gradients should look. So, where has the connection between this and your concept of "greater texture clarity/aliasing ratio" phrase been established, and then shown to be to such a degree that a user's evaluation is that the GF aniso is "absolutely better"?

No, not gradients, border shapes. The correct selection gives a sort of semi-circular pattern, which the GeForce series displays. The Radeon 9700 Pro does not.

Hmm. Well, what do you make of this, then?
http://www.beyond3d.com/forum/viewtopic.php?t=4051&start=20
(screenshot on that page)
Seems to me that the AF algorithm the GFFX uses has quite departed from the "semi-circular" pattern you propose as being the most correct one.

EDIT: Oh, and I hope you were kidding when you said 2xAA quality is identical.

 

[Mize]

Hmm. Well, what do you make of this, then?
http://www.beyond3d.com/forum/viewtopic.php?t=4051&start=20
(screenshot on that page)

So Nvidia went and hacked up a pseudo anisotropic that's considerably worse than the 9700's - very interesting. I guess that was how the managed to speed up their filtering.

Comments Chalnoth?

Mize

 

[Joe DeFuria]

I am still very puzzled by the FX "options."

Is it possible to force the "highest quality" anisotropic filtering via the FX control panel? The way they have it set up, it appears that you can only force the "balanced" and "aggressive" options. The only way to get the "high quality" mode would be to set to application preference...assuming the application has a setting for anisotropic filtering.

 

[jb]

Correct me if I am wrong but isnt the xS modes of AA only an option for D3D games? If so anybody that has run CS knows its runs so much better in OpenGL.