News - AMD Mantle API embraced by Cloud Imperium, Eidos & Oxide

[wickedn]

If Star Citizen will use Mantle,that means Crytek must be onboard too as the the game uses CryENGINE3.

So,that means DICE,Activision and Crytek will be using Mantle.

Star Citizen will be using CryEngine (4th Generation)
along with Homefront 2 and Ryse: Son of Rome. Also Bethesda Softworks will use it.

[wickedn]

I can't see it being a huge success if games makers need to use both mantle and direct x. I'm really excited by the prospects for it but AMD need to let other GPU manufacturers get in on the action or it will be a failure. And if AMD are hoping to destroy Nvidia with Mantle they had better think again, Nvidia are too successful at present to go down because of a new API.
I mean even if AMD limited other brands to 5x more draw calls and licensed it for free, then keep the magic 9X to themselves for now - that would alleviate the problem.

games makers shouldn't have to do much as AMD stated that they can pretty much use the same coding from console and use it on mantle.

[Noxvayl]

Mantle sounds a lot better than I thought reading through the comments. It will be interesting to see what the performance is like on BF4 when Mantle is enabled.

[DanceswithUnix]

Makes me smile though since the availability of Direct X has long been touted as a key advantage of Windows over Linux. I'm quite happy to use either.

Only touted by Microsoft, for years it was a vastly inferior knock off of OpenGL backed up by a policy of Microsoft keeping the Windows OpenGL implementation bad and slow. These days the likes of John Carmack say the two APIs are about equivalent, so it all ended with a big "Meh" rather than something better.

[Biscuit]

Once we start seeing partner boards for new AMD cards and some results, i might be tempted to get a couple 290s

[MustardCutter]

I wonder if this will primarily be a gimmic like physx or actually prove worthwhile with nice boosts in performance, though I think it's bad if it onlly works on one manufacturers cards; we don't want monopoly or nvidia doing their own alternative and I doubt they will change the design of their cards to use mantle and pay AMD to license it or whatever.

[Dooms]

I wonder if this will primarily be a gimmic like physx or actually prove worthwhile with nice boosts in performance, though I think it's bad if it onlly works on one manufacturers cards; we don't want monopoly or nvidia doing their own alternative and I doubt they will change the design of their cards to use mantle and pay AMD to license it or whatever.

On paper we are looking at a 20-30% increase, we will actually know the numbers next month when the Battlefield 4 Mantle patch comes along.

As we talked about above there is nothing stopping nVidia or Intel from using Mantle, its going to be completely open... the main questions are going to be are they going to? can they afford not to? time will tell

[MustardCutter]

On paper we are looking at a 20-30% increase, we will actually know the numbers next month when the Battlefield 4 Mantle patch comes along.

As we talked about above there is nothing stopping nVidia or Intel from using Mantle, its going to be completely open... the main questions are going to be are they going to? can they afford not to? time will tell

I'm not so sure, if it requires changes to GPU design for Nvidia to adopt it I doubt they will. I've heard it will only work on GCN, which would exclude nvidia, including all nvidia cards out in the wild. I also have a feeling that the mantle patch will cause a lot of bugs and take ages to fix before we really see what mantle can do.

[kalniel]

I'm not so sure, if it requires changes to GPU design for Nvidia to adopt it I doubt they will. I've heard it will only work on GCN, which would exclude nvidia, including all nvidia cards out in the wild.

At the moment it only works on GCN because AMD aren't going to write a driver for nVidia's architecture are they? But GCN is present in XBone, PS4 and AMD GPUs, so it's a fairly big target. nVidia can always write a driver for their architecture I suppose, presuming it's flexible enough which it probably is, or could be made so without too much effort.

I also have a feeling that the mantle patch will cause a lot of bugs and take ages to fix before we really see what mantle can do.

Depends how different it is to consoles. The hardware is the same, only the OS stack is different, but we'll see. PCs always have bugs (just look at normal graphics drivers from nVidia and AMD), I don't think mantle will be any worse than normal.

[MustardCutter]

At the moment it only works on GCN because AMD aren't going to write a driver for nVidia's architecture are they? nVidia can always write a driver for their architecture I suppose, presuming it's flexible enough which it probably is, or could be made so without too much effort.

I'm not so sure it's something that can be solved with a simple driver, I get the impression the benefits of mantle is that it 'talks' more directly to GCN hardware (amd only) than the more generalised directX/openGL that is high level / bloated enough to talk to any of the graphics manufacturer's hardware. i.e. nvidia might need to change their architecture to something like GCN for it to work.

[3dcandy]

It's down to how many developers jump on board really isn't it. If games come out and instantly perform 20% on an AMD card than an NV card they'll have to get a driver coded up. If the games STILL perform 10% better on AMD v NV running Mantle then NV will have a real hard time selling their cards. Time will tell on that score but this is potentially a game changer for AMD, and they know it with their aggressive pricing too!

[Noli]

Is everyone else misunderstanding Mantle or just me?

I'm not wishing to sound rude but from some of the above comments my understanding differs greatly. It appeared very clear to me in my read ups that the whole point of Mantle is that it is a *low* level API designed to run just above AMD's proprietary graphics GCN architecture. It's got nothing to do with AMD not being open or others being able to license it.

It allows programming close to the AMD's GCN architecture and so *by definition* couldn't be used by nvidia or intel hardware. The pay off is higher efficiency (read frame rates).

If nvidia or intel want they can develop their own low level APIs - no one's stopping them. With GCN in the xbox one and PS4 though, they know it won't fly as well as Mantle. Programmers can still provide all round compatibility with high level Direct X API but for those games offering the Mantle pathway as well, AMD card owners will benefit from a boost. That's it.

I welcome any corrections but that was my understanding of how it works.

[kalniel]

That's right - and nVidia could write their own that translated their hardware into the equivalent calls that Mantle makes, assuming their hardware carries out (or could be made to carry out) the same operations.

[scaryjim]

... I get the impression the benefits of mantle is that it 'talks' more directly to GCN hardware (amd only) than the more generalised directX/openGL that is high level / bloated enough to talk to any of the graphics manufacturer's hardware. ...

I think perhaps there's some confusion about how DirectX and OpenGL work here, and what an API is. Neither DX nor OGL run directly on graphics hardware. They are both translated by graphics drivers that then make calls to the hardware. The issue with that is the DX and OGL stacks are intercepted by the OS at a fairly high level and the OS plays a large part in how the DX/OGL instructions are handed off to the graphics. But all DX and OGL do is define a set of methods that programmers can use to perform certain tasks that can be handle by graphics hardware. AFAIK you don't even need special graphics hardware to handle DX/OGL calls: you could write a graphics "card" entirely in software and translate the DX/OGL API calls to be run on the host CPU (but it'd run pretty damn slowly). But there's nothing special about DX or OGL that allows them to be run on every graphics card, and without an appropriate driver a graphics card can't run DX/OGL.

Mantle is no different. It defines a set of methods that programmers can use to perform certain tasks that can be handled by graphics hardware. But, instead of going through the OS and DX/OGL stack, those calls are handled by a different driver that has more direct access to the graphics hardware. Since AMD are going to publish the specification for what calls Mantle will use, anyone could theoretically build a graphics card that could run Mantle - they don't have to be the same as GCN, they just need to be able to perform the functions that Mantle specifies. Again, I strongly suspect that you could actually write a virtual graphics card that supported Mantle (again, it would be *very* slow), without any graphics hardware.

The fact that the initial implementation of Mantle will only run on GCN doesn't mean it will only ever be able to run on GCN. OTOH, it's always likely to have a competitive advantage on GCN as it's been specified to take advantage of GCN. That may mean nvidia do some redesigning of their cards in the future to better serve the Mantle API. But that'll be no different from tweaking a card to support new features in a DX update. And if nvidia get on board and the standard remains open, then nvidia will be able to make recommendations for the enhancement of the standard. If they get into Mantle the way they've previously got into developer relations, in the long term nvidia could actually benefit from Mantle. In the mean time, it's the first major sea change in graphics on Windows PCs since DX was introduced, and I can't see how that's a bad thing....

[3dcandy]

Remember old games that had software renderer's for an idea of what is posted above...
But yes, Mantle possibly will run a lot better on GCN hardware - thus my post about NV either becoming cheaper to make up a performance deficit or something else coming along. They can't however just ignore Mantle as the user base will be massive very quickly

[DanceswithUnix]

When people say Intel or Nvidia can implement this open standard, I have to ask if that is possible then why can none of the VLIW4 or VLIW5 cards/APUs I have at home run this? None of this is old hardware, well maybe the llano is getting on a bit, but this isn't just AMD only, this is GCN only. I suspect that this will work nicely for GCN, and when AMD move on it will all die out.

If AMD add the ability to drive more than just GCN, then it becomes a high level API, and worse than that a new and therefore buggy high level API without the years of tweaking and optimisation of OpenGL/DirectX.