AMD Is Still Looking At The Feasibility Of Mantle On Linux

[Ericg]

I can't see the point in Mantle, it only seems to be making a difference in rigs with crappy CPUs (or rather APUs). It seems to be focused on use cases where the CPU is the bottleneck, which is hardly ever the case. If you've got a pimped out gaming rig then it doesn't seem to make much difference.

Not everyone HAS a pimped out gaming rig though. I gave myself a budget of $600 and a managed to build a $500 Kaveri gaming rig and a $600 R7 260X gaming rig (links available if anyone's curious). On the $500 one, Mantle would absolutely help system performance and probably even help on the 260X rig too.

[RealNC]

Interesting. AMD is now slowly backing away from a Linux version of Mantle? In the past, the did say that at some point it *will* be available. Now they're "looking at the feasibility" of it.

That is *really* interesting.

[blackiwid]

Some referened to directx here, 2 years, have you any idea how late that is. Customers will vote with their wallets in 2 years the war is over either opengl or mantle is the new mainstream gaming api.

I doubt it very hardly that a Windows 9+ only api can be successfull after 2 years waiting to get the same speed that you have today with opengl and mantle.

But, that said, to opengl, everybody says that the full mantle speed is in opengl today. Thats hardly much more than a urban myth. We did not see one game that proofes that, and some very basic demos proove nothing.

And there are very big hints that its just not true, yes opengl is faster than dx now no question about that, but why are they planning now as major theme to win the race over the fastest api if they would be the fastest today? Thats for me basicly a proof that even opengl 4.4 is slower than mantle today.

Even Nvidia on GDC 2014 says that opengl is not fast enought (not as fast as mantle):

http://scr3.golem.de/screenshots/1402/OpenGL_GDC2014/thumb620/2014-02-28_110614_cr.png

So yes I am skeptical if mantle is a good thing, but at least gamers even talk again about the potential of opengl and Windows-only-monopoly-stabilisation D3D gets under pressure from 2 sides and people see the face of this ugly slow as hell garbage api.

BTW even Microsoft try to spin it like DX 12 is a update of DX, basicly its a completly different api that has nothing to do with dx but the name.

SO Companies have to migrate to a complete new api anyway, maybe with that they use one api that also supports linux and playstation and android and everything.

[bridgman]

In the past, the did say that at some point it *will* be available.

Are you sure ? I don't remember seeing/hearing that...

[AnonymousCoward]

Even Nvidia on GDC 2014 says that opengl is not fast enought (not as fast as mantle):

http://scr3.golem.de/screenshots/1402/OpenGL_GDC2014/thumb620/2014-02-28_110614_cr.png

key word is Naive GL is faster. that mean clasic bind, set, draw approach. not moder DrawIndirect approach that were presnted on GDC.

[Kraut]

Even Nvidia on GDC 2014 says that opengl is not fast enought (not as fast as mantle):

http://scr3.golem.de/screenshots/140..._110614_cr.png

That slide looks like something where Nvidia talks about GL4.4 features or there own GL extension that deals with the draw call limit. (http://www.opengl.org/registry/specs...w_indirect.txt)

BTW even Microsoft try to spin it like DX 12 is a update of DX, basicly its a completly different api that has nothing to do with dx but the name.

SO Companies have to migrate to a complete new api anyway, maybe with that they use one api that also supports linux and playstation and android and everything.

Yes and the same goes for OpenGL. If they make all the changes that people want, it would be a different API. And then you may even give it a new name... like for example "mantle".

At this point I think it even would be healthier if AMD and Nvidia both publish there own very thin API. One thin layered API per big vendor still would be better then D3D or OpenGL. Developers could easily build a decent abstraction on top of that but prevent all the guess work and workaround with the current "APIs".

[tuke81]

key word is Naive GL is faster. that mean clasic bind, set, draw approach. not moder DrawIndirect approach that were presnted on GDC.

Yeah the slide is from steam dev days, just before they are talking about modern openGL extensions and how to reduce driver overhead. The picture of article is from gdc, where the difference of na?ve vs modern is very big.

[Sdar]

But, that said, to opengl, everybody says that the full mantle speed is in opengl today. Thats hardly much more than a urban myth. We did not see one game that proofes that, and some very basic demos proove nothing.

And there are very big hints that its just not true, yes opengl is faster than dx now no question about that, but why are they planning now as major theme to win the race over the fastest api if they would be the fastest today? Thats for me basicly a proof that even opengl 4.4 is slower than mantle today.

Even Nvidia on GDC 2014 says that opengl is not fast enought (not as fast as mantle):

http://scr3.golem.de/screenshots/1402/OpenGL_GDC2014/thumb620/2014-02-28_110614_cr.png

Why don't you have a look of all the presentation?

http://i.imgur.com/AhqsZQm.jpg

Opengl could be more than 18 times faster than Naive... proof that Opengl could be as fast or even faster than Mantle.

[johnc]

Not everyone HAS a pimped out gaming rig though. I gave myself a budget of $600 and a managed to build a $500 Kaveri gaming rig and a $600 R7 260X gaming rig (links available if anyone's curious). On the $500 one, Mantle would absolutely help system performance and probably even help on the 260X rig too.

Yeah but cheaper rigs are going to have a tendency to be GPU-limited, where Mantle isn't particularly helpful.

[Kraut]

Why don't you have a look of all the presentation?

http://i.imgur.com/AhqsZQm.jpg

Opengl could be more than 18 times faster than Naive... proof that Opengl could be as fast or even faster than Mantle.

18 times less CPU time for draw calls it not 18 times less CPU time for the overall driver work.

And how do you get the idea that OpenGL could be faster then a thinner layered API like mantle? (That many still suspect to be designed to much in favouritism of the GCN architecture and may not be viable for other GPUs)