Gaming industry and its nosed browned by nVidia.

[fallen_StorM]

It seems that even before 3dfx went down, that most game developers had their noses wedged tightly in nvidias a$$. Most new games as you know support new nvidia features (Example: Hardware T&L), but not new 3dfx features (Example: Motion Blur). The gaming industry decided 3dfxs' fate, not 3dfx. Lucky for 3dfx FSAA worked without having to have game support for it! But of course nVidia had a far crappier version of it, so it would have been supported anyway.

-StorM

[pacman]

Nvidia made there FSAA after 3DFX made theres, thats why Nvidias is just a software hack and doesnt look so good (wait what am I saying nothing on a Nvidia card looks good)
But the point of my message is without 3DFX Nvidia wouldnt have put FSAA in there drivers!

[rhink]

Their FSAA is not software, and it was present in the Geforce and Geforce 2 before the V5 was out the door.

3dfx killed themselves. Did you ever ask why those features did not show up? Well, nvidia has done a good job evangalizing their products to game dev's.... and 3dfx, well, towards the end especially, did not.

[Santiago Rex]

Bleh. Motion blur is shit, was shit, and always will be shit. I didn't see one good demo of motion blur. Blame 3dfx for not properly touting their features.

nVidia has all sorts of demos out, 3dmark and such *( :p ) that show off its features. 3dfx didn't have any. Except the crappy web page 'demos'

[John Reynolds]

Originally posted by rhink:
Their FSAA is not software, and it was present in the Geforce and Geforce 2 before the V5 was out the door.

I would still argue that to an extent. Its compatibility was very much dependent on driver version throughout the summer, which doesn't smack of being too hardware-based, and there was no real consideration given to it during the GF chip's design. Most sites royally screwed up their RGSS vs. OGSS comparisons, and even fewer noted how much better RGSS cleans up texture aliasing (they were too busy conducting edge AA/screenshot comparisons instead of full-scene AA comparisons).

And you make it sound like Nvidia evangelized AA with the GF cards. While it really doesn't matter now, you have to remember that Nvidia was downplaying FSAA in the extreme fall of '99. . .their CEO during a FS interview stated something like higher resolutions removed the need for it. A blatant lie in a double sense, because such a statement is patently untrue and the GF1 was hardly capable of ultra high resolutions in most games.

But water under the bridge, as the saying goes.

[Cop]

In short: nVidia sux once again...

[opus512]

Originally posted by rhink:
Their FSAA is not software, and it was present in the Geforce and Geforce 2 before the V5 was out the door.

So then T&L on the V5 is hardware, right?

[GrimFaceOfReality]

I thought that it was pretty well-known that Nvidia's FSAA is hardware.

[Ailuros]

while I'm pretty much tired of all that NVIDIA bashing......

Grim,
I tend to believe it's rather an issue on definitions. Clarify first if you can enable FSAA in older games on GeForces and why it is so. I'm not sure about that one, so consider it more a question.

If however FSAA shouldn't work at dx6 and older games, than it should have something to do at least with the drivers. Under those conditions I would say that FSAA is performed in hardware yet with strong driver support.....

[GrimFaceOfReality]

Well, here is what Kristof and Dave had to say (http://www.beyond3d.com/articles/nvidia_fsaa/index3.php) about it:

Now the reason that super-sampling is commonly referred to as being done through software is because of the multiplication of the vertex positions, as this is generally done by the CPU. However, with NVIDIA, they are actually able to do this through hardware. As we mentioned, the T&L unit is used for this multiplication, so the operations never have to be sent to the CPU. If the T&L unit was not present, this could be considered a software method, but because of it, the anti-aliasing is actually a hardware feature.

[Ailuros]

Concerning my question then, is that the reason why it's supposedly not working on older games? (given the fact that it is so, I'm still in the dark with that one). :confused:

[Madd Frog]

As far as that goes, stock info on 3dfx was common knowledge, all a game developer had to do was take a quick glance at 3dfx to determine what the future held for them, don't get me wrong, I loved all my voodoo's, but on a financial level they were dead a long time ago. If I were a game developer then I would have went for nVidia's tech as well, or sold it all and got a small coffee shop?

[GrimFaceOfReality]

Well the most issues were caused by using certain 2D elements in a 3D scene and ware D3D-only. For example, several flighsims used a 2D cockpit (through Direct Draw or a custom 2D engine) and had the rest of the frame rendered in 3D. When FSAA was enabled, the card would not render the 2D portion of the screen, causing image corruption. Several older games that used DirectDraw/Custom 2D engine to draw GUI suffered from this problem and had to be fixed on game-by-game bases.

[Ailuros]

Ok let's assume that Kristof's take is accurate, would that mean that a tiler uses software FSAA while performing OGSS? Or it is hardware just because the sampling is performed in the tile buffer?

(I hate to come back to tilers every time, but I don't have any other example, except that the Radeon AFAIK doesn't do any FSAA in 1x7 4 sample and above....)

edit: Grim,
replace the tiler with kyro1/2. It's a bad habit I know hehehehe

[ 04-28-2001: Message edited by: ailuros21 ]

[opus512]

nVidia's FSAA, at least when they added it, was *not* hardware.

This is the concept that nVidia is trying to pass off as reality: Since the drivers tell the hardware what to do, then it's done in hardware.

Wrong.

Otherwise, everyhting is hardware, everything, there is no emulation, it is all hardware, because hardware performs the driver function. (ie: no Glide wrapper, it's supported in hardware. Again, wrong).

But the common point we keep coming back to is *software*. FSAA on the nVidia is/was software dependent. It did not work in all games, it had/has to have D3D. Theredore, it is software dependent, therefore it is done by the drivers.

It's a very fine, blurry line nVidia is trying to cross, but if they get to cross it, then so does 3dfx, and they have Hardware T&L. You can't argue for one and against the other.

But the fact is, is that the Hardware T&L on the V5 is a trick, a driver hack, with the GA, and so is nVidia's FSAA. Or at least it was, I don't know anything about this Cuincunx <?) bullshit.

So here, to me, is the "Is it hardware or a driver hack?" questions answer, in the form of another question, "Is it software dependent?"

[ 04-28-2001: Message edited by: opus512 ]

[GrimFaceOfReality]

I don't think you can directly apply what Kristof said to the tilers due to the architectural differences, but I suggest you ask Kristof about that :). IMHO, as long as it works, who cares if its hardware or software? BTW, if you look at compare the GTS and KyroII results in 800x600 with 4x FSAA and 1600x1200, you'll see that while GTS scores in two cases will be identical, KyroII will be faster in 800x600 4x FSAA, because in KyroII the scene is downsamples before being passed to the framebuffer (you only need 800x600 framebuffer), while GTS downsamples in the framebuffer (you need 1600x1200 framebuffer).

[GrimFaceOfReality]

Originally posted by opus512:
It did not work in all games, it had/has to have D3D.

Huh? It works just fine in OGL.

If you want to argue with B3D about software/hardware thing, be my guest (http://www.beyond3d.com/forum/cgi/forumdisplay.cgi?action=topics&forum=3D+Hardware+talk&number=1&DaysPrune=1000&LastLogin=).

[Ailuros]

should have said that "who cares" when software FSAA was mentioned hehehe.

Actually I do care because if I get a card I want to know it's weaknesses more, than it's strenghts.....Guess I'll have to ask Kristof about it..... :D

[GrimFaceOfReality]

But if it performs exactly the same, what is it: a weakness or a strength? ;)

[opus512]

Originally posted by GrimFaceOfReality:
If you want to argue with B3D about software/hardware thing, be my guest (http://www.beyond3d.com/forum/cgi/forumdisplay.cgi?action=topics&forum=3D+Hardware+talk&number=1&DaysPrune=1000&LastLogin=).

Hell, I don't want to go over there, they tear Mult and ailuros21 up I ain't stepping into it :p

I argued this point ad nauseum before, along with the whole 3DM2k thing, I got no taste for it anymore. I guess I'll just be content with the Hardware T&L on my V5 :D

[Particle]

3dfx wasn´t the best graphic card company out there, ATI has always been THE BEST company outthere.

It has always had products amongst the top.

3dfx, did especially crappy with their voodoo 3500tv. It was shit. I had one.

:o

[multigl]

LOL @ opus. i hadnt posted at B3D for a few months until i put up the rampage pics JR provided to us.

[Ailuros]

Originally posted by GrimFaceOfReality:
But if it performs exactly the same, what is it: a weakness or a strength? ;)


Well Grim if it should be software based FSAA on kyro2 than it's bad news to be honest. Especially for some older games. Think about it. I do replay several titles from my library from time to time.

I wouldn't call that one a feature....*ahem*
:D

[multigl]

a software based FSAA would scale a lot better when you upgrade your CPU to something faster, but its incompatibilities with some games su><ors.

[Ailuros]

oh and opus I wasn't attacked by the regulars. You know my "behaviour" as a poster by now. It where just a bunch of trolls probably deriving from some shut down POS message board.

The situation over at B3D became unbearable at one point with the board flooded with morons. It's getting better though lately....hehehe

Particle,

I refuse to discuss video cards with a guy who still thinks his cpu is on par with t-birds.....ouch

hehehehehe j/k ;)

[rhink]

John:

I understand what you're saying... you're right that FSAA was not emphasized for the GF series, and that it really was not a design consideration for the card. However, I do not see how that affects the hardware vs software argument. I won't argue who's implementation was better (ie, 3dfx's or NVIDIA's), I've seen both in action, and RGSS does look significantly better (although I don't think the differences are as dramatic as some here would have you believe). I still say that high res comes first... it's not worth sacrificing the detail given by higher resolutions for FSAA, regardless of how good that AA may be... FSAA without sacrificing res would be ideal.

Opus: 3dfx's T&L is always performed on the CPU. NVIDIA's FSAA is performed on the graphics chip itself. That is the difference. To some extent you are correct... there is some software control necessary in D3D games to tell the card when to allow writes to the framebuffer, and when to downsample the framebuffer... this is caused by difficulties mixing 2D with 3D(which is also why it doesn't function properly in some games). However, I believe it is very misleading to just pass it off as "software" FSAA, b/c the CPU isn't what is doing the FSAA... the card is. In OpenGL, there's none of that mixing 2D and 3D, either, so there's no extra software control there (I'm also not aware of any OpenGL apps that have difficulties with FSAA).

>>So here, to me, is the "Is it hardware or a driver hack?" questions answer, in the form of another question, "Is it software dependent?"<<

Silly question. Most hardware is dependant on software in some manner or another. How well does your V5 work without drivers? Silly question? Yup.... just as silly as "Is it software dependant?" 3dfx's FSAA is also software dependant.... how well does the V5's FSAA work in UT or Q3A under Linux? What's that you say? It doesn't work? Why's that? I thought it wasn't software dependant?

[fallen_StorM]

Originally posted by opus512:
So then T&L on the V5 is hardware, right?

Uh, what T&L on the v5. The v5 has NONE, software or hardware. v5 = no t&l, that simple.

-StorM

[fallen_StorM]

Originally posted by Particle Man:
3dfx wasn´t the best graphic card company out there, ATI has always been THE BEST company outthere.

It has always had products amongst the top.

3dfx, did especially crappy with their voodoo 3500tv. It was shit. I had one.

:o


UH, no... the 3500 was the best thing out at the time, I had one... I still have one... I still use it. This card was awesome in it day.

-StorM

[Ailuros]

Originally posted by StorM:
Uh, what T&L on the v5. The v5 has NONE, software or hardware. v5 = no t&l, that simple.

-StorM

I think it's pretty obvious that he was being sarcastic.

:rolleyes:

[GrimFaceOfReality]

Originally posted by opus512:
Hell, I don't want to go over there, they tear Mult and ailuros21 up I ain't stepping into it :p

I argued this point ad nauseum before, along with the whole 3DM2k thing, I got no taste for it anymore. I guess I'll just be content with the Hardware T&L on my V5 :D

B3D is a great place for discussion (even despite the recent troll invasion).

As for Hardware T&L on V5, I don't think even 3dfx went as far as to make such a claim.

[Quackass]

"Gaming industry and its nosed browned by nVidia"

thats what happens when you toss a salad

[JF_Aidan_Pryde]

nVidia has the best engineering team in the industry. (I wonder why)
The rest is corporate slime.

[opus512]

Originally posted by GrimFaceOfReality:
B3D is a great place for discussion (even despite the recent troll invasion).

As for Hardware T&L on V5, I don't think even 3dfx went as far as to make such a claim.

B3D is a great place, but I don't claim to hold that level of technical knowledge :D

As for Hardware T&L, no, not even 3dfx claimed it, but to me it's the same concept as nVidia claiming thier FSAA is done in hardware. I just can't accept that.

No offence to you intended :)

[opus512]

Originally posted by rhink:
Opus: 3dfx's T&L is always performed on the CPU. NVIDIA's FSAA is performed on the graphics chip itself. That is the difference. To some extent you are correct... there is some software control necessary in D3D games to tell the card when to allow writes to the framebuffer, and when to downsample the framebuffer... this is caused by difficulties mixing 2D with 3D(which is also why it doesn't function properly in some games). However, I believe it is very misleading to just pass it off as "software" FSAA, b/c the CPU isn't what is doing the FSAA... the card is. In OpenGL, there's none of that mixing 2D and 3D, either, so there's no extra software control there (I'm also not aware of any OpenGL apps that have difficulties with FSAA).

>>So here, to me, is the "Is it hardware or a driver hack?" questions answer, in the form of another question, "Is it software dependent?"<<

Silly question. Most hardware is dependant on software in some manner or another. How well does your V5 work without drivers? Silly question? Yup.... just as silly as "Is it software dependant?" 3dfx's FSAA is also software dependant.... how well does the V5's FSAA work in UT or Q3A under Linux? What's that you say? It doesn't work? Why's that? I thought it wasn't software dependant?

This is largely my point, it's too complicated to pass off simply as one or the other, and yet nVidia tries to pass it off in that way.

But you can only go so far with the last part of your arguement. In that vien, I can get a Creative Labs nVidia card that supports Glide in hardware, because they have a wrapper that tells the hardware what to do. See my fine, blurry line point? That's all, I just meant it's not as clear cut as they would like to present it as.

[edit] And your last point, about Linux, uhm, no offence, but that's not really applicable, as that is OS envrionment dependent. Not quite the same as driver :)

[ 04-29-2001: Message edited by: opus512 ]

[ 04-29-2001: Message edited by: opus512 ]

[John Reynolds]

Originally posted by rhink:
[B]John:

I understand what you're saying... you're right that FSAA was not emphasized for the GF series, and that it really was not a design consideration for the card. However, I do not see how that affects the hardware vs software argument. I won't argue who's implementation was better (ie, 3dfx's or NVIDIA's), I've seen both in action, and RGSS does look significantly better (although I don't think the differences are as dramatic as some here would have you believe). I still say that high res comes first... it's not worth sacrificing the detail given by higher resolutions for FSAA, regardless of how good that AA may be... FSAA without sacrificing res would be ideal.

Well, my point was sort of that earlier cards could do various forms of AA. For example, anyone remember Tomb Raider 1's support of the Voodoo1's edge AA? Nasty performance hit, though, but I guess that's no different from today. ;)

I just think if it wasn't for 3dfx, Nvidia probably wouldn't have assigned Scott Cutler to get the OGSS working in their drivers as quickly as they did. They might never have bothered with it at all. Who knows. And lord knows how I love my FSAA. hehe

As far as your higher resolutions vs. FSAA comments, I somewhat agree. Certain levels in Giants were too much for my V5 to handle 10x7x16 with 4x FSAA, so I'd switch to 16x12x16 with no AA to make it more playable while not losing too much IQ. But when it comes to texture aliasing, there's no replacing RGSS these days, IMO. Even high anisotropic filtering isn't as good, though it's probably a more balanced IQ/performance option. But for pure eyecandy, it's hard to beat.

And I agree with Ailuros. . .there's too much mindless and/or unjustified bashing of Nvidia lately.