AMD Offers Mantle For OpenGL-Next, Pushes Mantle To Workstations

[rvdboom]

they only did that in order to retain as much R&D and GCN compatibility as possible. opengl next is more or less last nail in the mantle coffin and amd knows that. they are just trying to salvage as much as possible here. it was completely closed until this last nail

In my opinion, AMD got exactly what they wanted with Mantle : the next DX12 is basically a rip of Mantle and the next OpenGL has now a very good chance to be mostly based on it too.
DX and OpenGL have roughly been stagnating for years, none of them addressing the main problem AMD had with them, which is an overhead that limited the performance they could get from their APUs compared to Intel in games, as far as I could read.
By releasing mantle, they actually managed to force both Microsoft and Khronos to move in the direction they wanted. They're making sure that both API, as well as all the game consoles, are appropriate for their graphics chips and provide the necessary functionality. And there will be mantle2, mantle3 for this same purpose : to keep providing insights to MS and Khronos for the next evolutions of their respective APIs.
So I don't see anything "to salvage here", but actually a very good strategy to become the one company to lead graphic API evolutions, displacing Nvidia in the process.

Almost everybody wins : game developpers have very similar APIs to develop their games on, whether on consoles, on Windows or on Linux ; Khronos and MS have a manufacturer that agrees to spearhead evolutions of the APIs and gives them without licence once they're finalized ; AMD is certain that their next chips will be used in the fullest by all the existing APIs, whether discrete chips or APUs, and so they should sell more chips, particularly more APUs.
I don't know if this strategy will work out well but for now, it does seem on tracks to me.

[justmy2cents]

On what, exactly?

if you didn't notice you contradicted to your self on few fronts

I'm not seeing your point. But supposing I did, everyone here knows to take what AMD says with a grain of salt. I like them and I my main rig is an all-AMD system, but if what they show isn't REAL data, I never take anything they say seriously anymore. Their description of Mantle's abilities and features were blown out of proportion.

blown out of proportion... couldn't care less about it. False open incentive... hell yeah, it bothers me

If the code is being used to contribute toward open source projects, why do you care? Not only would you not do anything about it but if OpenGL is allowed to take from it, what difference does it make in the end? At that point, all the important code anyone would care about has been extracted and open-sourced.

and if you check my first post here... or you can read this summary
amd offering mantle to gl next can only have neutral or positive impact. neutral if khronos ignores it, positive if this means gl next can skip some steps by using mantle playbook and as such shorten needed time to be released

does this means i'll suddenly love mantle? no, i can respect it though. but, the fact they swinged left and right with ?Open?/crossplatform definitely puts it in vaporware list in my books. tech seems good (i don't like GCN dependency though, even last claims it doesn't require it are at best vague), but amd pr did all in their power to destroy it

AMD/ATI has been surviving pretty long with crappy drivers without going bankrupt. That should be enough of an answer to your question.

i fail to see how this would help your point in any way... or contradict mine

No... then I would be saying "I am speaking on behalf of AMD". What I'm saying is my speculation, if I somehow haven't made that clear enough already.
...
Well when you dismiss my opinion then sure, this makes no sense. All I'm saying is Mantle was designed to act as a proof of concept and their investment in it ensured other APIs would follow their lead. You can't be convincing without a functional prototype; Mantle, to me, is just a prototype that has the potential to remain permanent. But if it doesn't, it won't be missed.

your speculation and amd claims are not even remotely similar. if you check your posts, you state it as worlds truth, never as singular oppinion

In my opinion, AMD got exactly what they wanted with Mantle : the next DX12 is basically a rip of Mantle and the next OpenGL has now a very good chance to be mostly based on it too.

ms denied that claim and jumping to mostly mantle based opengl is not realistic. consoles had api like that like... forever and don't use mantle at all. new apis were simple evolution of the sad truth cpu can't feed gpu anymore in the way as it did, not something caused by amd. you could say new apis evolved to what consoles had it since day 0.

[schmidtbag]

You've been offering a little more than "just 2 cents". You're trying way too hard to disagree with me. I never contradicted myself, and what I'm saying is a bit complicated to summarize in a single sentence starting with "I think that...." so it's not going to come across as a humble opinion. If it were a fact I would just simply tell you you're wrong and link you to proof.

Also, assuming AMD giving Khronos the information becomes a "neutral impact". So what? If they take nothing from it, that means there was nothing worth taking. Therefore, why does it matter? No matter what, the open source community will benefit in the end. You're making a big deal out of nothing.

As for everything else - I'm wasting my time, because you're reading things how you want to. rvdbloom seems to overall have the same opinion as me (just worded differently) so at least I know my thoughts aren't as ridiculous as you make them out to be.

I think the problem is you're taking this WAY too literally. Most of what me and rvdbloom has said is SPECULATION.

[M@yeulC]

so, while preaching you like oss, you also love tressfx? http://www.amd.com/en-gb/innovations...logies/tressfx which is DX11 only, not open, and only works on GCN as hw option? and you hate gameworks which works on dx+opengl and works on all nvidia cards. same for mantle, 1os and 1/10 of 1 vendor supported only.

i don't hate either, each vendor has its own place in my heart where i won't reaplce it with anything else: nvidia+blob for games and 3d, amd(oss only, catalyst is... performance cataclysm) for desktop and intel on servers. only thing i might replace is nvidia blob if mesa reaches 4.4 and in that case i'd be willing to spend 2x more on graphic card just to retain perf level.

I am not a specialist on the subject, but I though tressfx used only DirectCompute, thus making the implementation vendor independent (only on DX11, though, but I can't blame them for that yet). While gameworks uses shared objects or dynamic link libraries, that are in no way open, or hardware independent.
As far as I know (feel free to correct me), game developers are given access to tressfx source (more or less), thus a game could be compiled for, let's say PowerPC, and still use it (if DirectCompute is available on that platform, though). I don't think this is possible with gameworks. In the other hand, shared objects make updates easier.
Well, I dislike these proprietary technologies (I though tressfx was a lot more open than it turned out after a bit more research), but I find the nvidia one even less open. They have always gone the full proprietary/locked way, and that's what I dislike with them.

[gamerk2]

I am not a specialist on the subject, but I though tressfx used only DirectCompute, thus making the implementation vendor independent (only on DX11, though, but I can't blame them for that yet). While gameworks uses shared objects or dynamic link libraries, that are in no way open, or hardware independent.
As far as I know (feel free to correct me), game developers are given access to tressfx source (more or less), thus a game could be compiled for, let's say PowerPC, and still use it (if DirectCompute is available on that platform, though). I don't think this is possible with gameworks. In the other hand, shared objects make updates easier.
Well, I dislike these proprietary technologies (I though tressfx was a lot more open than it turned out after a bit more research), but I find the nvidia one even less open. They have always gone the full proprietary/locked way, and that's what I dislike with them.

The problem with TressFX was it was limited and slow. Lets face it, Lara's hair broke every mid-range card on the planet; even cards like the 570 GTX couldn't handle it, and that was hardly mid-range. Compare that to, say, PhysX, which was much more open in what you could do with it, and faster due to being tightly integrated with NVIDIA HW/SW.

That's always the problem: The more open you make it, the wider the range of HW it supports, the slower it gets. It certainly doesn't help that NVIDIA's developmental platforms are top notch, which itself is a major incentive to develop for NVIDIA (speaking as a dev, ease of coding/debugging is a major factor in what platforms i develop for).

Lets call it what it is: Mantel is AMD's answer to Gameworks. You want performance? Code with Mantel. Want ease of development and extra graphical goodies? Go with NVIDIA.

[Sdar]

and they changed the tune how many times since then? will be cross platform, won't be, will make opengl extensions, won't make those, will be, won't be... where are we now? o lost my count, they are swinging left and right daily

I put the slide as an irony.

[rvdboom]

ms denied that claim and jumping to mostly mantle based opengl is not realistic. consoles had api like that like... forever and don't use mantle at all. new apis were simple evolution of the sad truth cpu can't feed gpu anymore in the way as it did, not something caused by amd. you could say new apis evolved to what consoles had it since day 0.

MS may have denied but game developers and people who had access to inner information claim that mantle and DX12 are so close it's not even funny. Check semiaccurate.com articles and forums, all the info is there.
At this point, there's no way you can claim that a mantle-based OpenGL is not realistic. OpenGL Next is supposed to be a complete rewrite and it's probably easier for Khronos to base its work on a freely available API that already worked out most of the issues than going in a completely different way.
Consoles indeed have API like this for a long time but now, they're all running AMD chips and game developers went to AMD to have a similar API for PC. It's probably not exactly the same but close enough not to rewrite the whole engine, that's what the game developer wanted, a console-like API, and that's what AMD gave them.
There was no sign of DX12 or OpenGL complete rewrite before AMD presented mantle to public and both are still at least a year away. You seem to consider that this API evolution would have happened anyway. I don't think so and certainly not so quickly.

[M@yeulC]

mantle and DX12 are so close it's not even funny

I just hope OpenGL Next will not have the shitty DX coordinates system (no offense to DX developers, but all mathematics are done in a right-handed Cartesian system, with axis upwards). Every time I have to work with SDL (not sure about SDL2, though), I have first to flip the vertical axis, otherwise, I would have to rethink all my transformations...

[blackiwid]

you should respond to blackout then. i never even remotely claimed anything like that.

In my First Post I did not mention your name or referered to a statement of you, yes I could have responded explizitly to blackout, but when I dont do that I dont mean automaticly you

[profoundWHALE]

I couldn't follow the arguments. To me it went like this:
"That hair is black."
"It is black but it could also be grey."
"It's not grey! It could be, but it's not, it's black."


And then it gets even worse from there.