AMD Reportedly Plans To Bring Mantle To Linux, Calls Mantle An Open-Source API

[dimko]

Well I woldn't care about them, the other GPU developers like Qualcom, Imagination, ARM and whoever I forgot to mantion, would have an bigger impact if exposed with Android which involves oss linux support in the first place.

1) Android != Linux

I wonder, what has so far been backported to Linux from Android.(i acknowledge Google support, however)

2) Hypothetically, ARM processor vendors suddenly support Mantle, what makes you believe that Google will accept mantle when there is already OpenGL for mobile devices? I don't see mantle happening with ARM producers anyway.
People stick to stuff that works and is multi platform, to something that is already widely acceptable. Big businesses are rarely non conservative.

[/QUOTE]

[dimko]

Would be super nice

I'm an OpenGL advocate first, but I realise that there are non-gaming companies such as Autodesk, that keep OpenGL from being as lean as it can be, and maybe they even have a right to. OpenGL didn't start out as a gaming API anyway.
If OpenGL 5.0 doesn't solve its issues, I hope AMD hand over Mantle to Khronos or something, and NVidia and Intel get onboard. On Windows, I've seen games having up to a 50% increase in framerate on Mantle compared to DirectX. Mantle would need to play nice with SDL, but I believe that's in the SDL devs' court.

But even now, Nvidia is a Maverrick, is only big GPU company that has both performance AND good driver, albeit, binary, on Linux.
When I see pigs flying in my future house in Spain, i will know, that this has happened.

[scottishduck]

AMD is failing to sell it's cards to gamers thank to cryptocurrency miners. They have a minority share in the graphics market. There's no chance for them to push mantle really.

[dungeon]

AMD, lets release Mantle for Linux at Civilization: Beyond Earth launch date and all GCN users will be happy .

[Kraut]

I'm an OpenGL advocate first, but I realise that there are non-gaming companies such as Autodesk, that keep OpenGL from being as lean as it can be, and maybe they even have a right to. OpenGL didn't start out as a gaming API anyway.

Gaming or nongaming should not make a difference. You want a performant, low latency, easy to mutlithread, good to debug API in all cases.

If OpenGL 5.0 doesn't solve its issues, I hope AMD hand over Mantle to Khronos or something, and NVidia and Intel get onboard.

I think they created mantle in the first place because of the political issues and lack of progress within Khronos.

Then explain to me why Johan Andersson of DICE who was one of the main guys behind Mantle and its integration into Battlefiled 4, says that he isn't even sure if Kepler would work with Mantle?

http://www.heise.de/newsticker/meldung/APU13-Der-Battlefield-Schoepfer-ueber-AMDs-3D-Schnittstelle-Mantle-2045398.html

Not that hard to translate. You have to look at every GPU architecture seperately to see if it's Mantle capable.

Johan Andersson just says he does not know exactly how it would work because he is lacking the low level knowledge about Kepler. But he thinks it is possible.
Basicly you just quote somebody who says he doesen't know...

AMD is failing to sell it's cards to gamers thank to cryptocurrency miners. They have a minority share in the graphics market. There's no chance for them to push mantle really.

"miners" already start to sell there GPU stacks. So in the long run they won't lower AMD marked share in the gaming community.
Also there are hopes that this flood of cheaper used GPUs will force both Nvidia/AMD to drope there GPU prices.

[jakubo]

Not that hard to translate. You have to look at every GPU architecture seperately to see if it's Mantle capable.

Yet you misunderstood. It is no bidirectional logic.

He said its independent from DX11 compatibility (like comparing apples and pears) and that it definitely WOULD work with Kepler. He just doesnt know how.

[Yorgos]

would be pretty funny if they open source it and it performs better on nvidia.

:epic_facepalm:
sure, they don't have proper performance in openCL but there is a chance to support a rival's product and do it better.

Well, I bet it will be 10% behind AMD's performance and it will cost 5 times more than AMD's gpus

[mdias]

now. story changed as it seems. they actually said that MOST of mantle is not gcn specific.

Without knowing the API we can imagine a number of situations where that's irrelevant.

example 1) imagine Mantle is OpenGL but people don't know exacly what OpenGL is yet; we could say most of OpenGL is IHV independent (ARB extensions), however some of it is specific to certain hardware (NV_ or AMD_ extensions).
example 2) maybe you can query the graphics device for a specific interface (that's part of the API) that only GCN implements.

Who knows?... We'll have to wait to comment further on this topic.

still, until nvidia and intel join, mantle is just short time vaporware

Agreed.
I'm skeptical it will make a significant impact if it goes on delaying and releasing vague statements without any real substance (even if only docs for now).

[chithanh]

Get Mantle into MESA, *then* call it open source.

Why? Is only mesa open source? If the vendor chooses to implement things outside mesa (like Intel beignet), does that make it not-open-source?

I've no idea about mantle API. But from Apple's low-level API metal I think Mantle will be very similar to gallium API.

People who have compared the whitepapers about Metal and Mantle say that the two APIs are based on similar concepts, with Metal being somewhat more limited:

both are very similar one have queues dispatched to a command buffer
the other have commands send into a command buffer. and more similarities
still metal seems a little bit more constrained at least that is what i think, it will be no surprise if when mantle spec is out will almost have one to one api calls

http://www.semiaccurate.com/forums/showthread.php?p=214279#post214279

Sadly Mantle will not bring any games to Linux. Who wants to cater to 4% of 2% of gamers, where only 10% of those would actually buy the game?

And no. Mantle doesn't simply need DX11 hardware. Even the DICE Mantle guy said that he doesn't know if it would work on Kepler.

Then explain to me why Johan Andersson of DICE who was one of the main guys behind Mantle and its integration into Battlefiled 4, says that he isn't even sure if Kepler would work with Mantle?

http://www.heise.de/newsticker/meldung/APU13-Der-Battlefield-Schoepfer-ueber-AMDs-3D-Schnittstelle-Mantle-2045398.html

Not that hard to translate. You have to look at every GPU architecture seperately to see if it's Mantle capable.

That Mantle needs more than DX11 hardware is correct, but the DICE developer did not say what you claimed. He said (emphasis mine):

Die Architektur der GPU muss davon unabh?ngig ganz bestimmte Anforderungen erf?llen. Man muss jede Architektur einzeln anschauen. Nvidias Kepler ist auch eine ziemlich moderne GPU. Ich weiss nicht genau, wie sie mit Mantle funktionieren w?rde, aber klappen w?rde es bestimmt.

The interview was likely done in English and then translated to German, so translating it back may lose some precision. But this is how I would translate it:

The GPU architecture must meet certain requirements independently [of DX11 support]. You have to look at every architecture individually. Nvidia's Kepler is a fairly modern GPU, too. I do not know exactly how it would perform with Mantle, but it would certainly work.

The original might be slightly different, but there is no way he said that he was unsure whether Mantle would work on Kepler.

actually, that was told on one of the first mantle presentation. if i remember correctly it was representative for nvidia that asked it. and answer was "you have to license gcn"

now. story changed as it seems. they actually said that MOST of mantle is not gcn specific. still, until nvidia and intel join, mantle is just short time vaporware

It was said in the very beginning from the original Mantle reveal by DICE at AMD APU13 conference that the API can basically work with any modern GPU.

Source: TechReport

1) Android != Linux

I wonder, what has so far been backported to Linux from Android.(i acknowledge Google support, however)

2) Hypothetically, ARM processor vendors suddenly support Mantle, what makes you believe that Google will accept mantle when there is already OpenGL for mobile devices? I don't see mantle happening with ARM producers anyway.
People stick to stuff that works and is multi platform, to something that is already widely acceptable. Big businesses are rarely non conservative.

Even if they got Mantle support to mesa/gallium3d, it would still need support from official NVidia, nouveau is not really a choice for gaming so far and legacy applications depend on OpenGL.

How so? Whether NVidia chooses to support Mantle is not important. Games will need an OpenGL renderer anyway for other vendors and pre-GCN AMD cards, so they will just fall back to that. OpenGL and Mantle will co-exist on Linux, just like DirectX and Mantle co-exist on Windows.

The situation is fundamentally different than back when games were coded to one API. Now the market is dominated by a few game engines which output to a number of APIs already (e.g. DirectX 11, PS4, Xbox One, OpenGL, OpenGL ES). And now they will add support for Mantle and Metal which is a matter of weeks if AMD is to be believed.

[xeekei]

I think they created mantle in the first place because of the political issues and lack of progress within Khronos.

Then maybe they should form a new consortium between themselves, NVidia, and Intel. Possibly include game developers. I don't want an API that is controlled by one entity.