G80 indepth IQ reiew

[Gstanfor]

Originally posted by: BFG10K

ppl, this is actually true. NV3x does have the best AF dispite its lackluster performance.

This is false. Both the R5xx and G80 have better AF quality not to mention it's free, unlike the huge performance drop exprienced by the NV3x.

Rubbish! First of all I never claimed nv30 was superior to G80 - I said it was the best quality DX9 GPU class AF up until G80.

As for ATi r5xx being superior to nv3x AF, here's a side by side image to demonstrate exactly how inferior R5xx is (note especially the red outer mip region which is all but square). It may be faster AF, but it certainly isn't superior quality wise.

 

[GundamSonicZeroX]

Tee hee, this will make all the people on these and other forums that go "Nvidea nevr cared bout img qualty! EVR!" (misspelled intended for mockery) shut the fvck up and as now they have proof for now. For some reason if it's proven in the present than in a past article it has more of an effect. I tell those people that the GeForce 6 series had superior image quality compared to the Radeon X series.

 

[BFG10K]

First of all I never claimed nv30 was superior to G80

I never claimed you did.

As for ATi r5xx being superior to nv3x AF, here's a side by side image to demonstrate exactly how inferior R5xx is

Your own link shows it isn't superior except perhaps with the first red mip-map.

The rest of the mip-maps on the R520 are tighter, closer to being a circle and are back further than the NV30's.

Overall the R520 wins that test.

 

[Gstanfor]

What? The inner yellow mip ring is all but missing on the ATi AF pattern and this is supposed to somehow translate to better image quality? Likewise the spikes between the lobes (these translate to visible artifacts).

 

[BFG10K]

The inner yellow mip ring is all but missing on the ATi AF pattern and this is supposed to somehow thranslate to better inmage quality?

LOL, you really don't know much about the AF tester, do you? As usual you like to talk about things despite your sorely lacking understanding.

Here's a hint for you: the yellow ring is missing on the G80 too and this is a good thing.

The fewer rings on the wheel the stronger the card's AF is and the fewer mip-maps it needs to generate the required samples.

Likewise the spikes between the lobes (these translate to visible artifacts).

But the spikes on the R520 are much further back than on the NV30 so overall IQ is better there too, except on the first red one which I've already conceded.

Again according to that test the R520's AF is better than the NV3x's.

 

[Gstanfor]

No, the inner yellow mipband is present on G80 if you look, as a very thin ring. Just like the perfect AF image in the article has a thin yellow inner mip ring (that "perfect AF" image shimmers something awful for me at arms length viewing distance BTW).

The spikes on nv3x are extremely short where they do occur. On R5xx they are extremely pronounced and branch away from between the lobes.

 

[BFG10K]

No, the inner yellow mipband is present on G80 if you look, as a very thin ring.

There is a tiny element of yellow if you zoom in close and the same applies to the R520.

In any case you're shifting the goal-post now and comparing the R520 to the G80. This comparison is irrelevant because the NV30's yellow ring is fat which is why it's inferior to both cards.

I can't believe you're still debating this nonsensical tangent. I mean according to your reasoning the R420's AF is superior to the G80's because it has a red ring after the yellow one but the G80 doesn't. :roll:

The spikes on nv3x are extremely short where they do occur.

You must be blind if you think they're short. Furthermore the R520's mip levels are further back and both the green and light blue ones are closer to being a circle than the NV30's.

You obviously have no idea how to interpret the images so stop wasting my time.

 

[Gstanfor]

Originally posted by: BFG10K

No, the inner yellow mipband is present on G80 if you look, as a very thin ring.

There is a tiny element of yellow if you zoom in close and the same applies to the R520.

In any case you're shifting the goal-post now and comparing the R520 to the G80. This comparison is irrelevant because the NV30's yellow ring is fat which is why it's inferior to both cards.

I can't believe you're still debating this nonsensical tangent. I mean according to your reasoning the R420's AF is superior to the G80's because it has a red ring after the yellow one but the G80 doesn't. :roll:

The spikes on nv3x are extremely short where they do occur.

You must be blind if you think they're short. Furthermore the R520's mip levels are further back and both the green and light blue ones are closer to being a circle than the NV30's.

You obviously have no idea how to interpret the images so stop wasting my time.

I'm not talking about the outer spikes, i'm talking about the spikes that occur within the colored miprings. You can clearly see that R5xx breaks down badly on the yellow and blue rings especially. In fact the pattern closely resembles the moire patterns found in the outer extremeties of the B3D images.

Your claim that R5xx also has an inner yellow mipring is nonsense!

If you look at the nv30 image you will see there is a green mipring inside the inner yellow mipring - an amazing amount of AF detail that in some ways is still yet to be matched by anything.

I brought up G80 in relation to the inner mipring because you claimed

Here's a hint for you: the yellow ring is missing on the G80 too and this is a good thing.

I don'r consider R4xx or nv4x/g7x to do proper anisotropic filtering so I could care less about their quality issues.

In closing, someone here does not know what they are talking about and is clearly wasting peoples time - that person isn't me...

 

[BFG10K]

You can clearly see that R5xx breaks down badly on the yellow and blue rings especially.

The yellow ring isn't there like it is on the NV30 and that's why R520 is superior.

Again I'll ask whether you think the R420 is superior to the G80 because it has a red ring after the yellow but G80 doesn't?

Answer the question or stop parroting your nonsense about the yellow ring.

As for "breaking down" the R520's light blue ring is a rough circle while the NV30's is still a clover shape.

Your claim that R5xx also has an inner yellow mipring is nonsense!

The yellow makes the blue lighter because it's merged.

In closing,

Sadly I doubt that's the case. You'll keep coming back just like when you claimed you were getting HDR + AA in Far Cry on your 7900 GT SLI setup.

 

[BFG10K]

I don'r consider R4xx or nv4x/g7x to do proper anisotropic filtering so I could care less about their quality issues.

That doesn't answer the question, it's just your usual useless dodge using irrelevant rhetoric.

You claimed NV30 is better than R520 because NV30 has an extra mip-map ring (yellow).

Therefore using that ridiculous reasoning you would also have to claim the R420's AF is better than G80's because it too has an extra mip-map ring (red).

Likewise using your nonsensical reasoning disabled AF is the best because it has even more mip-map rings after the red one.

You clearly have no idea how AF works or how the AF tester works yet you continue to argue like some kind of expert on the subject.

This is your usual modus operandi and why so many people get irritated with your trolling.

 

[Gstanfor]

I have nothing to say about R4xx AF other than it isn't done correctly, and it's butt ugly, even by nv40 standards.

You probably missed my edit above, so here it is again.

image (wrong image before)

the nv3x image is displaying even more inner mipring detail than any of the other images. Note the thing green inner mipring. In some ways, I don't think nv3x's AF quality will ever be truly exceeded.

And once again, there in no yellow ring in the R5xx image. It's not "lighter in color" or something (that would be a fascinating optimization BTW...) it simply does not exist period! The pale blue mipring barely even exists!

 

[BFG10K]

I have nothing to say about R4xx AF other than it isn't done correctly, and it's butt ugly, even by nv40 standards.

Answer the question. Does a greater number of mip-rings equal better image quality?

ANSWER THE QUESTION.

ANSWER THE QUESTION.

ANSWER THE QUESTION.

ANSWER THE QUESTION.

ANSWER THE QUESTION.

Note the thing green inner mipring. In some ways, I don't think nv3x's AF quality will ever be truly exceeded.

Sure it can.

Clearly according to your reasoning disabled AF exceeds any form of AF because it not only has your touted green and yellow rings but it also has plenty of other pretty colours. :roll:

Unbelievable. I can't believe someone who knows so little can argue so much. The hallmark of a first class troll.

 

[Gstanfor]

G80 AF is very good indeed, don't get me wrong, but it has its roots in nv40 AF technology (developed mainly for consumers). nv30 gained its AF capabilities from nv20 (a product that while sold to consumers was *designed* for, not *adapted to* the professional market).

 

[Cookie Monster]

Originally posted by: Gstanfor
G80 AF is very good indeed, don't get me wrong, but it has its roots in nv40 AF technology (developed mainly for consumers). nv30 gained its AF capabilities from nv20 (a product that while sold to consumers was *designed* for, not *adapted to* the professional market).

Where are you getting this from?

G80 =/= anything NV4x and its variants e.g NV47 (G7x~)

G80 was in the works for almost 4 years, which means it was started during 2002 at the end of the FX disaster. This G80 project or NV50 is nothing like the NV40 or its future variants based on NV40.

You could say the G80s AF is a much more refined AF of the GF/3/4/FX series.

Gstanfor, BFG10K has proved you time after time that your wrong in your knowledge about the AF results. Its that simple.

-The smaller and rounder the circle the better.
-The less colour rings the better the quality.

Anyone with basic logic can work that out.

 

[thilanliyan]

lol I love GStanfor vs. BFG arguments.

 

[Gstanfor]

The more mipbands the better as the filter has more raw data at its disposal to work with.

The idea that the less color rings the better the quality is ridiculous. It only means that one particular mipmap is supplying almost all the information instead of the mipmaps being *filtered* like they are supposed to be!

 

[josh6079]

Originally posted by: Gstanfor
NV3x, despite any other percieved faults featured the absolute best AF of any DX9 class GPU until G80 showed up.

...I don't think nv3x's AF quality will ever be truly exceeded.

Gstanfor, BFG10K has proved you time after time that you're wrong in your knowledge about everything. Its that simple.

Fixed, and QFT. It never gets old either.

In closing, someone here does not know what they are talking about and is clearly wasting peoples time - that person is me...

Fixed again.

Gstanfor, you argued with BFG10K night and day about the G70/71's AF being good despite it having the worst AF out of all of them. It just seems that the only company that can ever make good enough AF for you is Nvidia.

What is your point/goal even if the NV30 had better AF than anything else? Should we all downgrade just to have a useless, well-filtered slide-show?

The R5xx series had better AF than the NV30. Wake up, move on, grab your G80/81 that we all know you're going to buy/recieve and get on with your life.

 

[Gstanfor]

The point was and remains, that nv30's AF is excellent (bested only by G80) and in no way deserved the slur ryszard's stupid image in the B3D IQ review handed it. It's certainly better quality than *anything* ATi has ever offered (or is ever likely to offer IMHO given what we have seen from R2/3/4/5xx).

 

[josh6079]

NV3x, despite any other percieved faults featured the absolute best AF of any DX9 class GPU until G80 showed up.

...I don't think nv3x's AF quality will ever be truly exceeded.

The point was and remains, that nv30's AF is excellent (bested only by G80)...

So which is it?

It's certainly better quality than *anything* ATi has ever offered...

How?

You stated that the R5xx AF was inferior to the NV30's because the NV30's had an extra yellow circle.

You then said that the G80's AF is even better than the NV30's even though it also doesn't have the yellow line you speak highly of.

If anything G80's AF is more like the R5xx series only it sports better transitioned colors and tighter meshes.

 

[BFG10K]

The more mipbands the better as the filter has more raw data at its disposal to work with.

LMAO!

So therefore in your little world bilinear filtering is better than than any form of AF since it has the most mip-bands?

ROFL!

Thanks for proving how utterly clueless you are.

The idea that the less color rings the better the quality is ridiculous.

To someone who doesn't understand AF.

It only means that one particular mipmap is supplying almost all the information

Exactly - that's the goal of AF.

Mip-mapping is only needed because no current form of AF is strong enough to go without it. If we had something much stronger like 256xAF the card could do all its filtering from the base texture without mip-mapping at all (so there would be no colour rings).

That would be perfect AF since all of the filtering data would be coming from the highest available detail.

The point was and remains, that nv30's AF is excellent (bested only by G80)

But you just told us:

I don't think nv3x's AF quality will ever be truly exceeded.

Are you now retracting that nonsensical claim?

It's certainly better quality than *anything* ATi has ever offered

Rubbish.

(or is ever likely to offer IMHO given what we have seen from R2/3/4/5xx).

You mean improvements in each generation such that R5xx is better than NV3x and only bested by G80?

 

[Gstanfor]

True ansiotropic filtering uses trilinear filtering and filters between perspective corrected mipmaps to determine the correct result. Less mipmaps means less possible choices. Less colors in the AF pattern means less mipmap filtering was performed (which means less AF was performed).

 

[BFG10K]

Less mipmaps means less possible choices. Less colors in the AF pattern means less mipmap filtering was performed (which means less AF was performed).

:roll:

So using that comical reasoning R420's AF is better than G80's since R420 has more mip-maps and more colours than G80?

Likewise you would also consider trilinear filtering (with AF disabled) to be better than any form of AF?

 

[Gstanfor]

There is nothing comical about the reasoning. That (True ansiotropic filtering uses trilinear filtering and filters between perspective corrected mipmaps to determine the correct result) is the definition of Anisotropic filtering.

http://en.wikipedia.org/wiki/Anisotropic_filtering

I couldn't give a rat's arse about R420's "AF".

Likewise you would also consider trilinear filtering (with AF disabled) to be better than any form of AF?

No, I would not. Those who like looking at mipmap transition boundaries may find it pleasant though. (which is precisely why the less colors = more IQ for the AF flower argument is so pitiful - we are supposed to be creating a smoothly filtered transition between mipmaps, not abruptly jumping from level to level).

 

[BFG10K]

Answer the questions:

So using that comical reasoning R420's AF is better than G80's since R420 has more mip-maps and more colours than G80?

Likewise you would also consider trilinear filtering (with AF disabled) to be better than any form of AF?

Answer the questions.

Answer the questions.

Answer the questions.


Not only that but your own link proves how wrong you are and how right I am:

If we were to explore an approximation to anisotropic filtering, mip mapping

True anisotropic filtering generates the anisotropic texture maps on the fly on a per-pixel basis.

Like I said, mip-mapping is only around because current AF methods are not strong enough to avoid it. If AF was strong enough all of the samples would be generated on the fly from the base texture and no mip-mapping would be required.

In such a case the AF colour wheel would have no colours on it and that would be perfect AF.

The better the AF the closer the AF wheel gets to that (in the case of the G80 it's the closest we've ever been in consumer space).

Your claim that the more colours and rings the better is beyond lunacy. In fact it's completely backwards to reality (a recurring trend in your posts I might add).

 

[BFG10K]

No, I would not.

Then by your own admission your claims are wrong:

Less mipmaps means less possible choices.

Using your reasoning G80's HQ 16xAF has less possible choices than trilinear with AF disabled since it has less mip-maps.

Less colors in the AF pattern means less mipmap filtering was performed (which means less AF was performed).

Using your reasoning G80's HQ 16xAF is performing less AF than trilinear with AF disabled since it has less colours.

which is precisely why the less colors = more IQ for the AF flower argument is so pitiful

Do you actually think before you type or does is just explode outwards like diarrhea in text form?