🚀 go-pugleaf

I reckon the shining point of the GeForce 3 is the memory architecture,
seeing as this is what the bottleneck is on current NVidia cards.


kenco.

"tnaw" <tnaw@MailAndNews.com> wrote in message
news:3AA2633E@MailAndNews.com...
> The antialiasing performance, both Quincunx AA and
> supersampling AA, is the most shining point of GeForce3.
> Benefit by both of the Quincunx AA mode and the high
> profitability memory bandwidth interface (Table 1.),
> GeForce3 run today games with Quincunx antialiasing
> enabled almost twice as fast as the GeForce2 Ultra with
> 4xFSAA(Supersampling) under the same situation (Table 2.).
> It is interesting to note that the new memory bandwidth
> interface to be extremely in favor of antialiasing fulfillment
> of GeForce3. The known benchmarking results (Table 2.) show
> that there is about +10% performance boost of GeForce3 over
> GeForce2 Ultra under the normal conditions, while we get a
> performance boost as high as over +50% with antialiasing
> enabled.
>
> http://www.geocities.com/tnaw_xtennis/NV20-7.htm
>

Yeah, if I got one at all it would be for the AA performance, at last nvidia
produce a feature that works out of the box!

On Fri, 2 Mar 2001 19:39:50 -0500, tnaw <tnaw@MailAndNews.com> wrote:

>The antialiasing performance, both Quincunx AA and
>supersampling AA, is the most shining point of GeForce3.

I don't mean to be a nitpicker, but GF3 uses multisampling, not
supersampling.

>Benefit by both of the Quincunx AA mode and the high
>profitability memory bandwidth interface (Table 1.),
>GeForce3 run today games with Quincunx antialiasing
>enabled almost twice as fast as the GeForce2 Ultra with
>4xFSAA(Supersampling) under the same situation (Table 2.).
>It is interesting to note that the new memory bandwidth
>interface to be extremely in favor of antialiasing fulfillment
>of GeForce3. The known benchmarking results (Table 2.) show
>that there is about +10% performance boost of GeForce3 over
>GeForce2 Ultra under the normal conditions, while we get a
>performance boost as high as over +50% with antialiasing
>enabled.

For out of the box features, I totally agree.  Long term, the vertex
processing and expanded pixel processing features are far and away the
most important features.  Not so much because they make the card
attractive to purchase now - because it isn't - but because the
existence of the features will push the developer to make better
looking games.  

>
>http://www.geocities.com/tnaw_xtennis/NV20-7.htm
>

"david woolfe" <dwoolfe@home.com> wrote in message
news:3aa15318.10067520@news...
> On Fri, 2 Mar 2001 19:39:50 -0500, tnaw <tnaw@MailAndNews.com> wrote:
>
> >The antialiasing performance, both Quincunx AA and
> >supersampling AA, is the most shining point of GeForce3.
>
> I don't mean to be a nitpicker, but GF3 uses multisampling, not
> supersampling.

Isnt Quincunx the multisampling AA (supersample + multisample actually) and
the rest of the modes are still the same supersampling method they use now?

Well, this article strongly implies that all GF3 FSAA will use
multisampling, with three options: 2x, 4x, and "quincunx."  But then,
even after editing, this guy writes like crap, so your guess is as
good as mine.

http://WWW.TOMSHARDWARE.COM/graphic/01q1/010227/geforce3-26.html

Over at Anands, here is another statement not quite definitve but
implies that it uses multisampling for all AA:

"The combination of the GeForce3's hardware based multisampling HRAA
and its use of Quincunx patterns results in quality identical to what
you would expect from a conventional 4-sample FSAA solution yet with
the performance hit of what you would expect from a 2-sample FSAA
solution."

By referring to the "combination" of multisampling and quincunx he
suggests that they are not synonymous, i.e. that quincunx is a further
innovation on top of multisampling.  Guess we'll know when the board
is released. 

On Sun, 04 Mar 2001 05:25:16 GMT, "fREoN" <none@nothanks.com> wrote:

>"david woolfe" <dwoolfe@home.com> wrote in message
>news:3aa15318.10067520@news...
>> On Fri, 2 Mar 2001 19:39:50 -0500, tnaw <tnaw@MailAndNews.com> wrote:
>>
>> >The antialiasing performance, both Quincunx AA and
>> >supersampling AA, is the most shining point of GeForce3.
>>
>> I don't mean to be a nitpicker, but GF3 uses multisampling, not
>> supersampling.
>
>Isnt Quincunx the multisampling AA (supersample + multisample actually) and
>the rest of the modes are still the same supersampling method they use now?
>
>
>

Here is what I saw at Sharkyextreme:
"Using a new technique dubbed HRAA, NVIDIA claims that they can achieve
better performance than super-sampling with comparable quality. A
"reconstruction filter" uses data from neighboring pixels to compute the
final output of any given pixel. The sampling pattern used by the
reconstruction filter is called Quincunx and, according to NVIDA, offers
quality rivaling traditional four-sample super-sampling while incurring the
performance hit of two-sample super-sampling."
http://www.sharkyextreme.com/hardware/articles/nvidia_geforce3_preview/6.sht
ml

To me, this says Quincunx is the HRAA, which is the MSAA.  After seeing
that, I checked Nvidia's site and downloaded the HRAA brief (PDF).  It backs
up the statement made at Shary's by having this:

"GeForce3 has new technology for graphics hardware antialiasing in the form
of multisampling using a new
sampling pattern. The locations of the sample points for multisampling
enable the use of a "reconstruction
filter" that uses data from neighboring pixels to compute the final pixel
color. This increases the total
amount of work, and therefore slows down the frame rate marginally, but it
delivers much higher quality as
a reward. This new sampling pattern is called Quincunx, after the name of
the pattern of the five dots on
the "5" sided of a 6-sided die."
http://www.nvidia.com/Pages.nsf/Lookup/HRAA/$file/HRAA.pdf

"david woolfe" <dwoolfe@home.com> wrote in message
news:3aa30361.27887413@news...
> Well, this article strongly implies that all GF3 FSAA will use
> multisampling, with three options: 2x, 4x, and "quincunx."  But then,
> even after editing, this guy writes like crap, so your guess is as
> good as mine.
>
> http://WWW.TOMSHARDWARE.COM/graphic/01q1/010227/geforce3-26.html
>
> Over at Anands, here is another statement not quite definitve but
> implies that it uses multisampling for all AA:
>
> "The combination of the GeForce3's hardware based multisampling HRAA
> and its use of Quincunx patterns results in quality identical to what
> you would expect from a conventional 4-sample FSAA solution yet with
> the performance hit of what you would expect from a 2-sample FSAA
> solution."
>
> By referring to the "combination" of multisampling and quincunx he
> suggests that they are not synonymous, i.e. that quincunx is a further
> innovation on top of multisampling.  Guess we'll know when the board
> is released.
>
> On Sun, 04 Mar 2001 05:25:16 GMT, "fREoN" <none@nothanks.com> wrote:
>
> >"david woolfe" <dwoolfe@home.com> wrote in message
> >news:3aa15318.10067520@news...
> >> On Fri, 2 Mar 2001 19:39:50 -0500, tnaw <tnaw@MailAndNews.com> wrote:
> >>
> >> >The antialiasing performance, both Quincunx AA and
> >> >supersampling AA, is the most shining point of GeForce3.
> >>
> >> I don't mean to be a nitpicker, but GF3 uses multisampling, not
> >> supersampling.
> >
> >Isnt Quincunx the multisampling AA (supersample + multisample actually)
and
> >the rest of the modes are still the same supersampling method they use
now?
> >
> >
> >
>

Interesting.  Assuming that we can interpret this as suggesting that
quincunx is the only multisampling mode, I wonder if the board will
have the two traditional 2x and 4x supersampling modes as well.  Like
I said, I guess we'll know in about two weeks.

On Mon, 05 Mar 2001 03:45:48 GMT, "Matt" <none@nothanks.com> wrote:

>Here is what I saw at Sharkyextreme:
>"Using a new technique dubbed HRAA, NVIDIA claims that they can achieve
>better performance than super-sampling with comparable quality. A
>"reconstruction filter" uses data from neighboring pixels to compute the
>final output of any given pixel. The sampling pattern used by the
>reconstruction filter is called Quincunx and, according to NVIDA, offers
>quality rivaling traditional four-sample super-sampling while incurring the
>performance hit of two-sample super-sampling."
>http://www.sharkyextreme.com/hardware/articles/nvidia_geforce3_preview/6.sht
>ml
>
>To me, this says Quincunx is the HRAA, which is the MSAA.  After seeing
>that, I checked Nvidia's site and downloaded the HRAA brief (PDF).  It backs
>up the statement made at Shary's by having this:
>
>"GeForce3 has new technology for graphics hardware antialiasing in the form
>of multisampling using a new
>sampling pattern. The locations of the sample points for multisampling
>enable the use of a "reconstruction
>filter" that uses data from neighboring pixels to compute the final pixel
>color. This increases the total
>amount of work, and therefore slows down the frame rate marginally, but it
>delivers much higher quality as
>a reward. This new sampling pattern is called Quincunx, after the name of
>the pattern of the five dots on
>the "5" sided of a 6-sided die."
>http://www.nvidia.com/Pages.nsf/Lookup/HRAA/$file/HRAA.pdf
>
>"david woolfe" <dwoolfe@home.com> wrote in message
>news:3aa30361.27887413@news...
>> Well, this article strongly implies that all GF3 FSAA will use
>> multisampling, with three options: 2x, 4x, and "quincunx."  But then,
>> even after editing, this guy writes like crap, so your guess is as
>> good as mine.
>>
>> http://WWW.TOMSHARDWARE.COM/graphic/01q1/010227/geforce3-26.html
>>
>> Over at Anands, here is another statement not quite definitve but
>> implies that it uses multisampling for all AA:
>>
>> "The combination of the GeForce3's hardware based multisampling HRAA
>> and its use of Quincunx patterns results in quality identical to what
>> you would expect from a conventional 4-sample FSAA solution yet with
>> the performance hit of what you would expect from a 2-sample FSAA
>> solution."
>>
>> By referring to the "combination" of multisampling and quincunx he
>> suggests that they are not synonymous, i.e. that quincunx is a further
>> innovation on top of multisampling.  Guess we'll know when the board
>> is released.
>>
>> On Sun, 04 Mar 2001 05:25:16 GMT, "fREoN" <none@nothanks.com> wrote:
>>
>> >"david woolfe" <dwoolfe@home.com> wrote in message
>> >news:3aa15318.10067520@news...
>> >> On Fri, 2 Mar 2001 19:39:50 -0500, tnaw <tnaw@MailAndNews.com> wrote:
>> >>
>> >> >The antialiasing performance, both Quincunx AA and
>> >> >supersampling AA, is the most shining point of GeForce3.
>> >>
>> >> I don't mean to be a nitpicker, but GF3 uses multisampling, not
>> >> supersampling.
>> >
>> >Isnt Quincunx the multisampling AA (supersample + multisample actually)
>and
>> >the rest of the modes are still the same supersampling method they use
>now?
>> >
>> >
>> >
>>
>
>

"Matt" <none@nothanks.com> wrote:

> To me, this says Quincunx is the HRAA, which is the MSAA.  After seeing
> that, I checked Nvidia's site and downloaded the HRAA brief (PDF).  It
backs
> up the statement made at Shary's by having this:

> "GeForce3 has new technology for graphics hardware antialiasing in the
form
> of multisampling using a new
> sampling pattern. The locations of the sample points for multisampling
> enable the use of a "reconstruction
> filter" that uses data from neighboring pixels to compute the final pixel
> color.

This sounds a lot like plain 'ol video interpolation. IE, you might reduce
jaggies and texture aliasing ("crawling ants") with this method, but it will
not be able to reduce "polygon flicker" (when small details like lamp-posts
far away in a driving game seem to flicker in and out of existence), as only
a true super/multisampling AA will deal with that. ...Provided this Quincunx
method is interpolating rather than supersampling, of course... I'm not sure
about that, but as I said, I would think it probable. Considering the claim
is 4x performance at 2x performance penalties, I would think it's some kind
of interpolation, perhaps using 2x super/multisampling as source material
since that's the kind of performance hit the system takes when enabling
it...


 Bye!
/HB.

"Hans Bergengren" <faar@swipnet.se> wrote in message
news:BFXo6.22015$AH6.3081839@newsc.telia.net...
> This sounds a lot like plain 'ol video interpolation. IE, you might reduce
> jaggies and texture aliasing ("crawling ants") with this method, but it
will
> not be able to reduce "polygon flicker" (when small details like
lamp-posts
> far away in a driving game seem to flicker in and out of existence), as
only
> a true super/multisampling AA will deal with that. ...Provided this
Quincunx
> method is interpolating rather than supersampling, of course... I'm not
sure
> about that, but as I said, I would think it probable. Considering the
claim
> is 4x performance at 2x performance penalties, I would think it's some
kind
> of interpolation, perhaps using 2x super/multisampling as source material
> since that's the kind of performance hit the system takes when enabling
> it...

To me, it looks like Quicunx is a combination of both Super and
Multisampling.
It supers to 2x, then basically blurs the samples to a near equivalent of
4x.
Of course, the only way to find out what it actually does for the image is
to buy one yourself and try it cause we all know the websites are anything
but reliable when it comes to reporting image quality, especially with FSAA.