NVidia Cg...now we know where the "Glide" rumors c
- Thread starter Joe DeFuria
- Start date
Matt Burris
Newcomer
Mfa:
The embargo is lifted on Thursday, anything before that is NDA territory. Since IGN got the information straight from NVIDIA, it's safe to say that they were put under NDA.
The embargo is lifted on Thursday, anything before that is NDA territory. Since IGN got the information straight from NVIDIA, it's safe to say that they were put under NDA.
DemoCoder said:
Hang on, why assume I view this as a threat or an either/or situation? I’m asking questions. And I’m asking questions because I know nothing about this (which I am slightly miffed about) whereas I’ve heard a lot about OpenGL 2 -- and right now its seems to me that there’s a lot of duplication between what’s trying to be achieved with OpenGL 2 (with a large group of 3D vendors) and Cg (which we don’t yet know whether the other vendors are even aware of).
I am thinking that the immediate benefit will be to expose hardware functionality that goes beyond current DX specification, which does suggest that NV30 has a greater shader functionality that’s going to be in DX9, which we’ve heard suggested before.
As for OpenGL 2 Prototype compiler code is already available and by Siggraph they hope to have the shader language ratified by the ARB which I believe is the final element.
http://www.beyond3d.com//articles/p10tech/index.php?page=page5.inc
To me this sounds like NVIDIA lost control of the high level language that will be part of OpenGL2.0 (heck NVIDIA seems to have had a lot less to say about OpenGL2.0 than they would have liked, most previous versions of OpenGL were driven by NVIDIA, now with OpenGL2.0 others took over and seemingly side-stepped NVIDIA - mainly due to the IP problems NVIDIA kept making) and created their own thing.
What worries me is where is this Cg going to sit for NVIDIA and where is it going to sit for NVIDIAs competitors ? NVIDIA might claim IP on the high level language and decline competitors to handle the high level language directly, possibly competitors will be forced through pre-compilers (of unknown quality - that output DX8 or DX9 as input for competitors drivers/compilers) that NVIDIA offers, while NVIDIA takes the direct path from Cg directly to their own low level GPU instruction set.
On the other hand they might completely opensource it (but why would NVIDIA do that, to get support behind it ? But then why not license it to MS for a lot of money - that is assuming its any good?). Even if its Open Sourced it will be "yet another compiler" and "yet another language" to support.
Most likely, IMHO, is that Cg will be NVIDIA property, only NVIDIA will be able to take the direct path from Cg to its own low level hardware. They will make it sound very nice by saying that they have created some very nice compilers so competitors can still run Cg, these compilers will scale Cg down to the DX standards. The big issue obviously is the quality and efficiency and level of optimisation of these "compilers".
The whole article also says nothing about this being free for competing hardware vendors, it talks a lot about things being free for developers. The article also does not really talk about compatibility with other hardware, it only mentions different platforms and talks a lot about "NVIDIA" hardware. It does mention DX and OpenGL shaders compatibility which sounds like pre-compilers...
So questions to ask yourself :
- Why would NVIDIA do this ?
- Will NVIDIA do all the support, SDK, compiler, development work and then offer it "all for free" to Competing Hardware Vendors ?
- Why would Developers use this ?
- This "standard" was created by NVIDIA (for NVIDIA ?) without any influence of competitors (?) can this ever be a "good" and "fair" standard ?
- Do we want ATI, Matrox, PowerVR, SIS, etc to come up with their own high level API ?
- Should we not just stick with a true standard high level API as will be part of DX and OpenGL2.0 where a base compiler is offered for free and every hardware vendor can write its own optimised version and where the language was created by a forum where all vendors had their input ?
All sounds way to proprietary to me...
G~
What worries me is where is this Cg going to sit for NVIDIA and where is it going to sit for NVIDIAs competitors ? NVIDIA might claim IP on the high level language and decline competitors to handle the high level language directly, possibly competitors will be forced through pre-compilers (of unknown quality - that output DX8 or DX9 as input for competitors drivers/compilers) that NVIDIA offers, while NVIDIA takes the direct path from Cg directly to their own low level GPU instruction set.
On the other hand they might completely opensource it (but why would NVIDIA do that, to get support behind it ? But then why not license it to MS for a lot of money - that is assuming its any good?). Even if its Open Sourced it will be "yet another compiler" and "yet another language" to support.
Most likely, IMHO, is that Cg will be NVIDIA property, only NVIDIA will be able to take the direct path from Cg to its own low level hardware. They will make it sound very nice by saying that they have created some very nice compilers so competitors can still run Cg, these compilers will scale Cg down to the DX standards. The big issue obviously is the quality and efficiency and level of optimisation of these "compilers".
The whole article also says nothing about this being free for competing hardware vendors, it talks a lot about things being free for developers. The article also does not really talk about compatibility with other hardware, it only mentions different platforms and talks a lot about "NVIDIA" hardware. It does mention DX and OpenGL shaders compatibility which sounds like pre-compilers...
So questions to ask yourself :
- Why would NVIDIA do this ?
- Will NVIDIA do all the support, SDK, compiler, development work and then offer it "all for free" to Competing Hardware Vendors ?
- Why would Developers use this ?
- This "standard" was created by NVIDIA (for NVIDIA ?) without any influence of competitors (?) can this ever be a "good" and "fair" standard ?
- Do we want ATI, Matrox, PowerVR, SIS, etc to come up with their own high level API ?
- Should we not just stick with a true standard high level API as will be part of DX and OpenGL2.0 where a base compiler is offered for free and every hardware vendor can write its own optimised version and where the language was created by a forum where all vendors had their input ?
All sounds way to proprietary to me...
G~
fresh said:
I fully agree.
btw, does anyone remember anything that nVidia would have allowed used by other companies?
soo... how realistic it really sounds that we are going to see PS 1.4 compiler for this one?
And now Nappe's "The Way Off Thinking(tm.)" (thanks to dsuiko inventing that term in some Bitboys thread.
)so, here we go...
IMO it's not at all realistic. IMHO it is going to be most likely that nVidia protects this with huge copyrights, patents and with Division of lawyers. so where this takes us then? well, what ATi does? they release their own "Higher Level Shader language compiler" for their cards and of course it's syntax isn't compatible with Cg. So, then we have Two Higher level languages with uncompatible syntax and developers have to write programs twice again... so, what actually changes?? maybe games get more complex and more detailed, but the main problem is still there.
Matt Burris
Newcomer
You guys are so pessimistic when it comes to NVIDIA.
Matt Burris said:You guys are so pessimistic when it comes to NVIDIA.
Pessimistic or Realistic ?
Think MONEY, its all about making MONEY, by handing stuff free to your competitors you do not MAKE MONEY, you LOSE MONEY. In the end NVIDIA wants to sell as many boards as possible, ways to do that is exclusive features or extra performance and both are possible by a well shielded and protected proprietary high level language. Trick will be convincing the developers to use it, they might initially fall for it until they find out of performance (or other) issues with a large % of their target audience (think money again).
G~
Matt Burris
Newcomer
Deleted.
Matt Burris said:
By trying to force your own designed standard you are not going to "advance the 3d graphics industry", your trying to force it into your direction - which might not "be" the "direction" of the "3d graphics industry". Remember that "NVIDIA" does not equal the "3D Industry".
Upto now NVIDIA has been using OpenGL to showcase their next things through extensions that could become standard, (un)fortunatly NVIDIA lost control of the heading of OpenGL so now they -logically- do their own thing... something they will ALWAYS control and rule over.
Ah well we'll have to wait some days to have access to all the small print.
Also even if it ends up OpenSource for all, the competitors will only get the specs when NVIDIA is already done and launched, meaning there will be huge delay times still giving NVIDIA months of exclusive use (in both demos and possibly games if they pump enough money into the game companies).
Soo much doubt... soo many questions... still some days to wait
Read the PR Marketing talk very carefully...
G~
Matt Burris
Newcomer
Deleted.
Whether proprietary or not, if it does really bring something new to the table, I'd say it is good for the industry. One should understand that it is very hard to push something new in a open standard.
Even if NVIDIA keeps its CG proprietary, the industry can still benefit from it. At least we'll know whether such idea will work or not. If it works well, there will always be an alternative way to do something similar. If it does not work at all, at least we don't need to try that again.
Even if NVIDIA keeps its CG proprietary, the industry can still benefit from it. At least we'll know whether such idea will work or not. If it works well, there will always be an alternative way to do something similar. If it does not work at all, at least we don't need to try that again.
I get the impression ign have cottoned on that this was still NDA:
http://pc.ign.com/articles/361/361823p1.html
http://pc.ign.com/articles/361/361823p1.html
Matt Burris
Newcomer
Deleted by the NDA police.
I wonder if CG includes a scene-graph API as well? Hmm..
If NVidia was smart, they would release their CG compiler with source included, that way it could be quickly updated by any developer with a patch to support their favorite API or graphics hardware.
If they are semi-smart, they will have an API to plug in for other hardware, good docs, and a repository on their site where you can download all the plugins.
If they are a little bit smarter than that, they will even make a super-good implementation for the ATI 8500, and support PS1.4 code generation.
If they are dumb, they will make it support NVidia hardware only, charge money for its use, and it will be an utter failure.
If NVidia was smart, they would release their CG compiler with source included, that way it could be quickly updated by any developer with a patch to support their favorite API or graphics hardware.
If they are semi-smart, they will have an API to plug in for other hardware, good docs, and a repository on their site where you can download all the plugins.
If they are a little bit smarter than that, they will even make a super-good implementation for the ATI 8500, and support PS1.4 code generation.
If they are dumb, they will make it support NVidia hardware only, charge money for its use, and it will be an utter failure.
MfA said:
If your hardware doen't support that level of functionality doesn't mean you have to have a level of abstraction there?
The abstraction comes in because the API, unlike DX8/8.1, isn't defined at the register level, and its up to the 3D vendors compilers to determine what is actually supported by their hardware:
Regardless of what is exposed I don't see how this significantly differs from Cg?
Matt Burris said:Told you it was under NDA.
You've also confirmed to everyone that this is true so essentially you broke NDA as well
