Nvidia/ATI Feature philosophy

[trinibwoy]

This quote from an article by Brent @ HardOCP got me thinking....

Right now with Shader Model 3.0 it is all about getting developers involved and working with the code for future games. NVIDIA told us at their Editors Day that they actually brought in developers into the architectural design phase of NV40 to get their feedback and suggestions. They also shipped the first 100 boards straight to developers. NVIDIA is very serious about working with the developers and getting their hardware out there so that developers can start making games with the new technology.

Of course there will be those of you out there saying, “But there aren’t any PS/VS 3.0 games now, so what does it matter?â€￾ The answer to that of course is if the hardware was never put out there first, then those games using those new features would definitely never be made. Developers need to have an actual piece of hardware they can run their games on with those features. The features will always come out on the hardware first before we have the software that can use them. With the next DirectX version a very long ways off and games evolving at a fast pace, we need a video card that is built for the future.

I believe ATI did just this with their excellent DX9 support with R300 but why would they drop the ball on R420. I may be missing something but feature wise does R420 bring anything to the table in the context of the above quote?

And on another note - let us assume that there is no NV40, what hardware would be used during development of games to be released in a year or two requiring advanced features? Are these things done in software or is there always professional hardware that leads the consumer market?

Appreciate your thoughts on this.

 

[speng]

I take it you haven't been reading the threads.

Search for SM 3.0 and you'll find your answers in multiple threads.

You can achieve nearly the same things as SM 3.0 with SM 2.0 if you have enough performance to cover it.

Speng

 

[trinibwoy]

I take it you haven't been reading the threads.

Search for SM 3.0 and you'll find your answers in multiple threads.

You can achieve nearly the same things as SM 3.0 with SM 2.0 if you have enough performance to cover it.

Speng

Jesus man!!! I expected this. Useless reply.

Please limit responses to the context of my original post. I would like some feedback on the developer/IHV relationship when it comes to programming for future hardware/features.

 

[digitalwanderer]

I think what it comes down to is: will pure PS3.0 support be required to take advantage of the features Shader 3.0 will bring to DX9.0c?

Gads I don't want to get into that debate yet since it is WAY to early too know though!

That is a big question right now, and methinks we're all gonna have to wait until we at least see what the R420 supports and in what way the developers decide to use the features of PS 3.0 in games.

EDITED BITS: "too" for "to". Sorry, I am a space cadet today!

 

[jolle]

As ive been saying in many threads (which by the way isnt based on anything but my own impressions, guesses and speculations, and im not by far what you call "in the know" hehe)

Initially, since ATI doesnt seem to support it, the implementations we will se in games will mainly be "doing the same thing as PS2.0" but slightly faster (compared to the same card in PS2.0)..
In some cases devs might choose to do something that would cost to much performance using PS2.0 then they would sacrifice on "a wall" or whathave you..
This type of implementation is fairly cheap and Farcrys "SM3.0" which looks like it would be this type, took only 3 weeks to make...

For it to be used widly and to make a real difference in the final image, you would prolly have to invest alot of time, and since the market for it will be rather small with only Nvidia offering support, it will likly not happen alot until
the next gen...

A parallell i like to use is 64bit computing..
AMD took the jump and if you bought a Athlon64 now, you dont have any real software support, but its on the way and being worked on...
Intel said they will also use these 64bit extentions, but not until there is support for them, such as WinXP64bit on the market..
That support wouldnt be on the way if AMD wouldnt have taken the first step tho..

So Nvidia is paving the way for SM3.0, software will get there, and ATI will jump on wagon if it gets widly adopted..
chances are that there will be DX next with a new Shadermodel before that happens tho..
in which case SM3.0 will never had come to real use...

 

[trinibwoy]

I think what it comes down to is: will pure PS3.0 support be required to take advantage of the features Shader 3.0 will bring to DX9.0c?

Gads I don't want to get into that debate yet since it is WAY to early to know though!

That is a big question right now, and methinks we're all gonna have to wait until we at least see what the R420 supports and in what way the developers decide to use the features of PS 3.0 in games.

Digi, sorry I used SM3.0 as an example. The gist of my question was meant to be broadbased. Please ignore the current SM2.0 can do 3.0 stuff too issue.

Thanks

 

[trinibwoy]

So Nvidia is paving the way for SM3.0, software will get there, and ATI will jump on wagon if it gets widly adopted..
chances are that there will be DX next with a new Shadermodel before that happens tho..
in which case SM3.0 will never had come to real use...

Aaaah!! Now that's what I was getting at. There is a lot of talk of 'X will never be used to its full potential until architecture Y comes out' But until Y comes out what will developers use if generation Y-1 doesn't introduce it? What are people's thoughts on this dynamic?

 

[Miksu]

This quote from an article by Brent @ HardOCP got me thinking....

Right now with Shader Model 3.0 it is all about getting developers involved and working with the code for future games. NVIDIA told us at their Editors Day that they actually brought in developers into the architectural design phase of NV40 to get their feedback and suggestions...

Appreciate your thoughts on this.

Isn't the ShaderModel 3.0 courtesy of Microsoft? One could read this so that NVIDIA based the SM 3.0 on developers' wishes which isn't the case.

edit: Though I know this didn't answer your question. Just pointed out the "feeling" I got from that quote.

 

[speng]

If you phrase your question in that context that's how people will interpret it.

If you are looking for a philosophy, right now for both companies it's this:

Gain market share and increase profit - as with all companies.

To do that - ATI wants to keep the market performance lead.

Nvidia - wants to regain market performance lead.

ATI would rather sacrifice on some features this generation to keep that lead, especially when those features may not be necessary.

Speng.

 

[digitalwanderer]

Digi, sorry I used SM3.0 as an example. The gist of my question was meant to be broadbased. Please ignore the current SM2.0 can do 3.0 stuff too issue.

Ok, then I don't get your question too well. Are you asking for what PS 3.0 can do that PS 2.0 can't and how will THAT affect the upcoming competition?

If so, it really is a less-than-moot point at this time. The only 3.0-only features that will be utilized on the nV40 will be seen in tech demos and PR releases only for the useful life of the card. ("useful life" being the time the nV40/R420 are the top dog cards)

 

[trinibwoy]

This quote from an article by Brent @ HardOCP got me thinking....

Right now with Shader Model 3.0 it is all about getting developers involved and working with the code for future games. NVIDIA told us at their Editors Day that they actually brought in developers into the architectural design phase of NV40 to get their feedback and suggestions...

Appreciate your thoughts on this.

Isn't the ShaderModel 3.0 courtesy of Microsoft? One could read this so that NVIDIA based the SM 3.0 on developers' wishes which isn't the case.

Yes Miksu, but as mentioned before it takes API, IHV and application to make a feature a reality.

The application (game) comes last and my question revolves around the ability to develop for future hardware. The many posts about 'so and so isn't going to be used till 2006' are kinda illogical since you have to develop on something and development takes time so it makes sense to have the hardware out a while before. The games we will be playing on R500/NV50 are in development now or soon will be.

 

[jolle]

for DirectX there is a dev group from what i understand..
MS has to be involved i guess since its their child, ATI and Nvidia is also part of that group i think, since they are the players on the market..
I guess you got other groups involved aswell..

So SM3.0 might have been "lobbied" forward by Nvidia, looking at NV3x, SM3.0 isnt a HUGE leap, but it would be for the R3xx core since its more PS2.0 then NV3x was..
But it would prolly have had to be approved by MS and others..

From what i hear NV3x failure in DX9 was due to Nvidia assuming FX would be adopted as standard, do to their "greatness in the market" at that time (pre DX9 gen)
and they left the DX dev group for now reason, only to return and find out they moved the specs along, and not according to theif "FX shaders", ATI who was there all the time built their R300 along the lines of the specs drawn up..
Might not be true, but that is what ive read on that...
doesnt sound very incredible...

 

[trinibwoy]

If you phrase your question in that context that's how people will interpret it.

If you are looking for a philosophy, right now for both companies it's this:
Gain market share and increase profit - as with all companies.
To do that - ATI wants to keep the market performance lead.
Nvidia - wants to regain market performance lead.
ATI would rather sacrifice on some features this generation to keep that lead, especially when those features may not be necessary.

Speng.

Whew!!! Thanks for the really obvious stuff. If I can get feedback from a developer and get their take that would be cool. How do you guy's approach development given the available hardware on the market, expected future hardware and the features that you would like to support in the game.

 

[trinibwoy]

Digi, sorry I used SM3.0 as an example. The gist of my question was meant to be broadbased. Please ignore the current SM2.0 can do 3.0 stuff too issue.

Ok, then I don't get your question too well. Are you asking for what PS 3.0 can do that PS 2.0 can't and how will THAT affect the upcoming competition?

If so, it really is a less-than-moot point at this time. The only 3.0-only features that will be utilized on the nV40 will be seen in tech demos and PR releases only for the useful life of the card. ("useful life" being the time the nV40/R420 are the top dog cards)

No I'm not. Apparently people have been blinded by the barrage of SM2.0/3.0 arguments to see the meat of my question. Just pretend that we are back in 99 and we're talking about the first T&L cards

 

[akira888]

I take it the question you want answered is: "What does the R420 have in the way of new features?" That's impossible to answer since we (except Dio, OpenGL Guy, Sireric who won't say anything) know next to nothing about R420. Even the clockspeed is still up in the air.

 

[Joe DeFuria]

No I'm not. Apparently people have been blinded by the barrage of SM2.0/3.0 arguments to see the meat of my question. Just pretend that we are back in 99 and we're talking about the first T&L cards

Well, you can't completely separate the SM 2.0/3.0 "debate" and your question. I'll do the best I can at addressing your questions....I'll just ramble a bit.

* Developers DO for the most part need "real hardware" before they start playing with advanced features. The sooner they have advanced features in hardware, the sooner they can start evaluating the usefulness of it, and plan on how to integrate the technology into new and existing projects.

* "Professional hardware" at this point, actually tends to lag the consumer market in terms of gaming features. Complete software emulation of advanced features is possible from a standpoint of looking at visual results, but practically useless for games developers who need to be assured of a certain performance level in addition to the visual effect.

* What ATI may be bringing to the table in the context of that quote...is performance. If the R420 can perform certain techniques at real-time frame-rates that R3xx can't, then this offers developers an opportunity to utilize PS 2.0 in a way that they couldn't a generation ago.

 

[trinibwoy]

I take it the question you want answered is: "What does the R420 have in the way of new features?" That's impossible to answer since we (except Dio, OpenGL Guy, Sireric who won't say anything) know next to nothing about R420. Even the clockspeed is still up in the air.

LOL I never knew that I could pose a question that could be so cryptic to so many.

Akira, you're right in that it is part of my question. But let's take a look into the future. Should we expect the first hardware to support DXNext to only support those new features IF they are able to do so with good performance? I think it's a good idea to have the features regardless of performance so that developers have something to work with. I thought that was clear from my original post...but maybe I'm wrong

 

[ben6]

SM3.0 will be used by all of the IHVs eventually (witness PowerVR's talk on PS 3.0 in the ShaderX book (forget the name right now), witness ATI's hints at GDC about R500 arch in Richard Huddy's notes), and XGI's roadmap. As to how useful it'll be in the lifetime of 6800U/R420 etc. I dunno.

Note, I'm not talking about features or lack thereof on R420, I'm simply talking about SM 3.0 in general

 

[Joe DeFuria]

I think it's a good idea to have the features regardless of performance so that developers have something to work with.

It's good for developers...but that doesn't mean the products themselves are good for consumers.

Many of the "disagreements" on this board arise because we get viewpoints mainly from two different perspectives: developer and consumer.

 

[trinibwoy]

*Developers DO for the most part need "real hardware" before they start playing with advanced features. The sooner they have advanced features in hardware, the sooner they can start evaluating the usefulness of it, and plan on how to integrate the technology into new and existing projects.

Thanks. This is the heart of it and it's exactly why I don't understand comments that aim to devalue the worth of new features in hardware that won't be supported in games for some time.

* What ATI may be bringing to the table in the context of that quote...is performance. If the R420 can perform certain techniques at real-time frame-rates that R3xx can't, then this offers developers an opportunity to utilize PS 2.0 in a way that they couldn't a generation ago.

Thanks. A valid point that I overlooked. I was more focused on doing new things with new features than new things with existing features