V5 5500 vs R9800PRO AA quality

[Ruined]

Just wondering, has anyone done an A/B comparison of Voodoo5's T-Buffer AA against R9800PRO's 6x AA? Wondering if its worth it for Nvidia to use the T-Buffer instead of their current AA tech?

[ChrisRay]

IMO the Radeon looks better than the Voodoo5 4x mode. At 6x. But looks inferior at 4x.

So..


Voodoo 5 4x > Ati 4x
ATi 6x > Voodoo5 4x


Nvidia would wanna use mbuffer, As tbuffer is completely super sampling. the Mbuffer on the rampage chipset was supposed to do Multui Sampling Anti Aliasing at no performance loss

Where have I heard that before :rolleyes:


anyway. Tbuffer is not something Nvidia should be using. It would devestate the cards fill rate

[Hanners]

AA has moved on a long way since the Voodoo5. Give me MSAA with gamma correction over SSAA any day!

[ChrisRay]

Originally posted by Hanners
AA has moved on a long way since the Voodoo5. Give me MSAA with gamma correction over SSAA any day!


Super Sampling still has its place. It's worth having as a viable option to Multi Sampling in Some titles.

[druga runda]

I agree we should be given both options so we can shoose if we have older or less graphically intensive game to use SSAA

[Hanners]

Originally posted by ChrisRay
Super Sampling still has its place. It's worth having as a viable option to Multi Sampling in Some titles.

In older titles (particularly those with alpha textures), yes. Beyond that, it's just too slow to be superior across the board.

[PreservedSwine]

Talk about a misleading title...I thought this was the "Other" graphics card secion:condom:

[ChrisRay]

Originally posted by Hanners
In older titles (particularly those with alpha textures), yes. Beyond that, it's just too slow to be superior across the board.

You are making rather strong statements. People assume that Alpha textures are the only reason to want Super Sampling.

There are definate other scenerios which could benefit the user. Which may not be relevent to you. That is why flexiibility is important. It's in excusable to not have super sampling in the drivers.

I would like super sampling for Nin64, And Playstation Emulation. To me this is a viable reason to want super sampling. You probably could give a nack about it.

So no. It's not too slow to be "superior" across the board for things I want it for. There's nothing wrong with flexibility. There are many instances it's not too slow for me.

P.S. Forgive my sig. It's not directed at you. It's Directed at my annoyance with Catalyst Maker ;)

[walkndude]

It's always seem to me that ss went to the wayside for the fact of its such a bandwidth/fillrate hog...not because of IQ

Even with higher degrees of AF todays games suffer from too many texture artifacts(not including just alpha). Crawling, aliased, nasty looking moire patterns drive me nuts to this day and it sure seems not to be a priority with card makers.

Sure were seeing the best EDGE aliasing yet, but there is alot more to overall IQ than how clean the edges of polygons appear.
Actually it makes texture aliasing that much more noticeable at times because of the stark contrast between clean, well filtered polygons and textures and noisy, shimmering, crawling textures.

I think thats the biggest problem I have with people posting screen shots and purporting great image quality when a still image is going to tell only half the story... kinda like how you can take a geforce4 and crank the lod to -3 and take stunning looking screen caps. The pics look wonderful but what happens when you try to play that way :)

Personally I think nV's biggest mistake was not expoiting the hybrid aa modes such as 4xs as seen in the geforce 4 line by increasing the edge quality and doing what was needed to get performance in line with their multisample modes.

I'm not sure if its a driver or architecture difference, but 4xs on the 5900 series just doesnt look nearly as good it does on the geforce4 line... I suppose its a moot point because the performance is less than stellar on either.

[Darth Rancid]

I hardly ever used the FSAA of my old V5, because of the fuzzyness... to get rid of it you had to tweak the LOD bias quite a bit.. and FSAA and "good" LOD together totally murdered performance...

The FSAA of my R9500Pro is superior to that of the V5 in every way I'd say.. so nVidia needs something mych better than the T-Buffer if they want to beat ATi on FSAA...

[Hanners]

Originally posted by ChrisRay
You are making rather strong statements. People assume that Alpha textures are the only reason to want Super Sampling.

There are definate other scenerios which could benefit the user. Which may not be relevent to you. That is why flexiibility is important. It's in excusable to not have super sampling in the drivers.

I would like super sampling for Nin64, And Playstation Emulation. To me this is a viable reason to want super sampling. You probably could give a nack about it.

So no. It's not too slow to be "superior" across the board for things I want it for. There's nothing wrong with flexibility. There are many instances it's not too slow for me.

P.S. Forgive my sig. It's not directed at you. It's Directed at my annoyance with Catalyst Maker ;)

I can't argue with that, you can never have too much variety (Although I can understand why ATi aren't too keen on enabling SSAA, I can just imagine the field day sites would have benching a 9800 Pro at 4x SSAA against a 5900 Ultra with 4x MSAA).

However, the question here is should nVidia should use the Voodoo 5's anti-aliasing method, stick with what they have or do something else. In the context of that question, anyone who thinks that nVidia should go for a pure SSAA implementation a la 3dfx in their next generation card would be labelled a little bit crazy in my book. ;)

[digitalwanderer]

ATi's AA wins for me hands-down. The AA on me V5 is too blurry by comparsion. :)

[ChrisRay]

Originally posted by Hanners
I can't argue with that, you can never have too much variety (Although I can understand why ATi aren't too keen on enabling SSAA, I can just imagine the field day sites would have benching a 9800 Pro at 4x SSAA against a 5900 Ultra with 4x MSAA).

However, the question here is should nVidia should use the Voodoo 5's anti-aliasing method, stick with what they have or do something else. In the context of that question, anyone who thinks that nVidia should go for a pure SSAA implementation a la 3dfx in their next generation card would be labelled a little bit crazy in my book. ;)

Obviously not. :) I actually believe that hybrid modes have there place. Like 4xS, Being 2 sample Multi Sampling and 2 sample Super Sampling.


The problem is. Nvidia needs to improve on its MS a bit.

4x should be rotated. And 2x should be performhance hitless.

Then they could do 8xS and 4xS more logically. Since 8xS could be have 4 sample Multi Sampling.


I generally like the idea.


Nvidia needs to improve its 4x pattern and lower the performance hit on it. Then its hybrid modes will be acceptable :)

[Nv40]

Nvidia only need to improve the performance and tweak their AA Xs modes ..
Nv 4xS ~ 4xaa ATI when it comes to jaggies..
for texture quality 4xS wins easily here in IQ..
the cons.. it takes huge performance

6xS and 8XS > than ATI 4x/6x at vertical angles..
but in horizontal angles they lose by a great margin.

so what Nvidia needs to do is to Tweak their 6xS,8xS AA in horizontal angles ,to be as good as they are in vertical angles.. and to increase their performance. a new 4xMS mode is welcome too..

[5150 Joker]

Originally posted by Nv40
Nvidia only need to improve the performance and tweak their AA Xs modes ..
Nv 4xS ~ 4xaa ATI when it comes to jaggies..
for texture quality 4xS wins easily here in IQ..
the cons.. it takes huge performance

6xS and 8XS > than ATI 4x/6x at vertical angles..
but in horizontal angles they lose by a great margin.

so what Nvidia needs to do is to Tweak their 6xS,8xS AA in horizontal angles ,to be as good as they are in vertical angles.. and to increase their performance. a new 4xMS mode is welcome too..


4xS does *not* win in texture quality or any part of IQ. :rolleyes: Do you just like going around making **** up?

[reever2]

I wasn't aware AA did something to anything other than polygon edges.... :rolleyes:

[StealthHawk]

What does the poll have to do with the thread topic :confused:

[Ruined]

Originally posted by StealthHawk
What does the poll have to do with the thread topic :confused:

Pretty simple. Many would agree that Nvidia's current AA isn't the best compared to ATI. Nvidia also owns the rights to 3dfx's T-Buffer. My poll queries whether Nvidia should explore the T-Buffer, or move onto something completely different to match ATI's 6x AA mode. Being that I never owned a V5 5500 I was never able to witness the T-Buffer firsthand, I was wondering how it compared to today's AA solutions, and whether it could be of any interest to Nvidia in the future.

[AnteP]

Originally posted by reever2
I wasn't aware AA did something to anything other than polygon edges.... :rolleyes:

Super Sampling does, it samples the whole scene.

[AnteP]

Originally posted by Nv40
so what Nvidia needs to do is to Tweak their 6xS,8xS AA in horizontal angles ,to be as good as they are in vertical angles.. and to increase their performance. a new 4xMS mode is welcome too..

You don't just "increase performance" with a SSAA implementation. It's basically completely hardware limited.
nVidia could tweak all day but it would still eat up too much bandwidth and fillrate at the end of the day.

Personally I'm all for SSAA and MSAA+SSAA hybrids. I really wish ATi would support them. At least we're now seeing light in the end of the tunnel with pure SSAA modes (2x and 4x).

Limiting the powerfull NV3X with only 4x MSAA is a very foolish move in my opinion though. As is only supporting 4x OGMS etc.

I was really hoping for some improved AA/AF on the FX series, instead they just left us with the same old stuff they've been using since over two years now. (ie GF3 tech)

Instead of focusing on the hyper advanced mumbo jumbo beyond DX9 spec shaders, which will probably never be used since games that will use that kind of tech will probably be launched around NV50, they should have focused on stuff that has an impact on actual gaming today. And in fact: stuff that is universal:
I can't ask HL to use PS2.0+ shaders but I sure as hell can enable some AA.



For the next gen though I'm hoping that they've come up with some better methods than SSAA for clearing up alpha textures and texture aliasing.
Preferebly I'd see 8x "RG"MS with gamma correction etc. and 32x "real" trilinear anisotropic filtering.

[StealthHawk]

Originally posted by Ruined
Pretty simple. Many would agree that Nvidia's current AA isn't the best compared to ATI. Nvidia also owns the rights to 3dfx's T-Buffer. My poll queries whether Nvidia should explore the T-Buffer, or move onto something completely different to match ATI's 6x AA mode. Being that I never owned a V5 5500 I was never able to witness the T-Buffer firsthand, I was wondering how it compared to today's AA solutions, and whether it could be of any interest to Nvidia in the future.

Getting 3dfx quality is something NVIDIA should strive for. Should they use a T-buffer though? No. SSAA is much too costly, people are used to low hit 2x FSAA and now lower hit 4x FSAA on NVIDIA cards. Plus gamma correction is a must.

[ChrisRay]

Originally posted by AnteP
You don't just "increase performance" with a SSAA implementation. It's basically completely hardware limited.
nVidia could tweak all day but it would still eat up too much bandwidth and fillrate at the end of the day.

Personally I'm all for SSAA and MSAA+SSAA hybrids. I really wish ATi would support them. At least we're now seeing light in the end of the tunnel with pure SSAA modes (2x and 4x).

Limiting the powerfull NV3X with only 4x MSAA is a very foolish move in my opinion though. As is only supporting 4x OGMS etc.

I was really hoping for some improved AA/AF on the FX series, instead they just left us with the same old stuff they've been using since over two years now. (ie GF3 tech)

Instead of focusing on the hyper advanced mumbo jumbo beyond DX9 spec shaders, which will probably never be used since games that will use that kind of tech will probably be launched around NV50, they should have focused on stuff that has an impact on actual gaming today. And in fact: stuff that is universal:
I can't ask HL to use PS2.0+ shaders but I sure as hell can enable some AA.



For the next gen though I'm hoping that they've come up with some better methods than SSAA for clearing up alpha textures and texture aliasing.
Preferebly I'd see 8x "RG"MS with gamma correction etc. and 32x "real" trilinear anisotropic filtering.


If they could do 4x RGMS, It would be nice. Shame they cant. They could improve there hybrid modes atronomically with it

[StealthHawk]

Originally posted by ChrisRay
If they could do 4x RGMS, It would be nice. Shame they cant. They could improve there hybrid modes atronomically with it

I think the hybrid modes are using OGSS components and not RGSS too, in addition to using OGMS.

[ChrisRay]

Originally posted by StealthHawk
I think the hybrid modes are using OGSS components and not RGSS too, in addition to using OGMS.


They are. because there's no 4x RGMS.

Only 2x uses Rotated Grid Multi Sampling. Which is why 4xS is rotated :P

Assuming 4x could be Rotated Grid MUlti Sampling. Theoretically we would see amazing quality from an 8xS that was half 4x Multi Sampling.

Similar to 4xS.

Which is 2 part RGMS and 2 RGSS.

just assume if 8xS was 4 sample RGMS and 4 part RGSS ;) ohh. But unfortunately it has to be all ordered because 8xS = 4x OGMS and 4x OGSS :(


The Hidden 8xS mode. was true 8 sample Rotated Grid Super Sampling. But it never saw light in open driver use.

[AnteP]

Originally posted by ChrisRay
The Hidden 8xS mode. was true 8 sample Rotated Grid Super Sampling. But it never saw light in open driver use.

The hidden 8xS Mode is 2x RGMS + 2x2 OGSS. (2x2x2 = 8)

The normal 8x Mode is 4x OGMS + 2x1 OGSS. (4x2x1 = 8)

But from what I recall 8x in Open GL is actually 8xS, no?