Innovation: Rewarded. . .and not so much.

[Geo]

Sireric's comments on innovation in the FP24 thread lead me to wonder if, since the rise of the non-vendor specific API (D3D and OGL), has "innovation" (i.e. "we got it and they don't") requiring ISV support *ever* been rewarded to any signficant degree?

Certainly some of the transparent innovations have been, like Z optimizations, but have the Tru-Forms and the like ever really been embraced? Any cases at all?

 

[BRiT]

Way back in the day, innovation was rewarded. Have a look at how long 3dfx enjoyed success with glide. That's the list time I recall anything being innovative outside of base D3D/OGL functionality.

However, in this day and age, I don't see anything that would point to being rewarded for being innovative. Perhaps the biggest innovations are rewarded via improved performance or visual quality.

 

[Geo]

So the lesson to be learned is: if you don't own the API or can't make it completely transparent, you're wasting your time and resources?

Sad.

S3TC, maybe? The lesson being you have to get it implemented into the API for everyone's use pretty quick? Then, of course, you've lost your competitive edge, so why bother?

 

[rwolf]

How about pixel shaders?

 

[nelg]

Sometimes the innovation comes in the form of execution not just features. Also conveyed in the other thread was the message of how the API develops with the input of all interested parties. Look at 3Dc for example, it was developed by ATI but IIRC MS may included it in future DX (or WGF) versions.

 

[Dio]

Yes, I'd agree S3TC was an example. Of course, part of the reason for its acceptance is that Microsoft selected it from the various methods offered to them for the DirectX standard as offering the best implementation cost, size and image quality tradeoff.

 

[Geo]

Yes, I'd agree S3TC was an example. Of course, part of the reason for its acceptance is that Microsoft selected it from the various methods offered to them for the DirectX standard as offering the best implementation cost, size and image quality tradeoff.

So it worked out well for the *industry* --was S3 themselves rewarded? That's not a rhetorical question either --the idea being that if the innovator isn't rewarded you'll likely see much less innovation, the world being the way it is. . .

 

[WaltC]

Sireric's comments on innovation in the FP24 thread lead me to wonder if, since the rise of the non-vendor specific API (D3D and OGL), has "innovation" (i.e. "we got it and they don't") requiring ISV support *ever* been rewarded to any signficant degree?

Certainly some of the transparent innovations have been, like Z optimizations, but have the Tru-Forms and the like ever really been embraced? Any cases at all?

First, I think it appropriate to ask why innovations like truform and other custom IHV-specific ideas haven't been adopted formally into APIs for universal support. Might it not be because such innovations were never anything more than stop-gap solutions on the road to much better things? I think so, and truform is a good example--as it came at a time when gpus couldn't handle enough polys quickly enough to get rounder, more realistic renders. As such, the power of recent and current gpus to push polys far exceeds the power of the gpus available at the introduction of truform, and like those gpus the time for truform has come and gone already. (ATi has said this often for the last couple of years.)

I think there's a difference between worthwhile innovation (like shaders) and stop-gaps like truform, and I think the good ideas are indeed absorbed into the APIs for enduring support while the temporary measures fall off when no longer of practical value. It's the difference between true innovation and marketing gimmickry, to me, and as such the two terms should not be confused. IE, everything that today portends to be extra-API, custom "innovation" isn't innovation but actually is but a stop-gap, instead, on the road to something better which will find a home in an API and endure.

I think there's plenty of R&D and innovation going on, and that much of it is obvious. I don't think it's a good idea to define "innovation" as something which has to fall outside of the API to merit the label.

 

[Dio]

So it worked out well for the *industry* --was S3 themselves rewarded? That's not a rhetorical question either --the idea being that if the innovator isn't rewarded you'll likely see much less innovation, the world being the way it is. . .

That's a very good point. S3's reward was certainly limited in the PC space due to the need to issue a royalty-free licence for its use in DirectX.

It did however get a relatively large amount of early support because it was clearly a massive benefit for everyone. There's no doubt the Unreal Tournament-based techdemos must have shipped a few extra chips.

I imagine they also made a reasonably large amount of money licencing it for non-DirectX applications, but I've no idea where or when it was used. I'd certainly say that Sony made a mistake in NOT licencing it for PS2.

 

[Reverend]

Perhaps we best identify what is really meant by "innovation".

"Truform" has been mentioned here. Is that a ATI innovation?

 

[Geo]

I think there's plenty of R&D and innovation going on, and that much of it is obvious. I don't think it's a good idea to define "innovation" as something which has to fall outside of the API to merit the label.

Possibly a poorly titled thread then, as I would agree with you in the largest context. As noted in the intro, it was inspired by Sireric's comments in the FP24 thread that were stating his personal preference that the API not be so definitive and controlling as to not leave room for innovation outside the API, or in the way the API is implemented by a given vendor. . .which got me to wondering how much that particular kind of innovation has been successful, and what its limits seem to have been historically in recent years. Tentatively, my conclusion is "good for transparent techniques; not impossible, but pretty rocky, for non-transparent techniques."

 

[Reverend]

I don't think DX is too restrictive in one aspect -- you can exceed certain "minimum specs" for a certain feature (number of instructions, for example). Dunno if we can call this "innovation" though

 

[Ichneumon]

Perhaps we best identify what is really meant by "innovation".

"Truform" has been mentioned here. Is that a ATI innovation?

Absolutely "Trueform" was an ATI innovation. The concept itself was not a revolution, but the implementation ATI came up with (through the mind of Alex Vlachos) and implemented in their hardware made it entirely useable in a time where other concepts of the idea were not at all feasibly useful.

 

[DegustatoR]

Absolutely "Trueform" was an ATI innovation. The concept itself was not a revolution, but the implementation ATI came up with (through the mind of Alex Vlachos) and implemented in their hardware made it entirely useable in a time where other concepts of the idea were not at all feasibly useful.

Npatches was ATI's innovation? Wow.

 

[Ichneumon]

Absolutely "Trueform" was an ATI innovation. The concept itself was not a revolution, but the implementation ATI came up with (through the mind of Alex Vlachos) and implemented in their hardware made it entirely useable in a time where other concepts of the idea were not at all feasibly useful.

Npatches was ATI's innovation? Wow.

Its implementation was. which was my point if you read what I wrote.
ATI did not invent Npatches, but it was certainly their innovation to implement it in hardware the way they did, and implement it in such a way that it was useable by developers, and could be experienced by end-users in its time. That's more than can be said for other implementations of HOS of the time. Therefore it was most definately an innovation... you may just disagree on how much benefit it provided.

 

[dksuiko]

Innovation is second only to "free" in the book of a marketing director. It is best used when attemping to sell something. To label any of these features as "innovative" is beside the point of the improvement/feature itself. Improvements are what they are, nothing more. Blah.

 

[sireric]

Some come to mind:

ATI: Npatches, instancing, 3dc, F-buffer, barycentric-sampling, gamma-corrected AA, Early Z, Hierarchical Z, etc...

NV: DST/PCF, Their Hierarchical Z, Destination-Z culling, some of the NV30 shader instructions (sin/cos), etc...


Both IHVs take some chances. Some pay-off, some do not (and some take a huge amount of time to get driver support for).

 

[Ostsol]

*pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! (etc. . .)

 

[sireric]

*pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! *pokes sireric* F-Buffer please! (etc. . .)

Yes, the driver work was done recently, but I'm not sure on availability. There are things in the way.

 

[tEd]

barycentric-sampling

explanation please