From the little we know about DX12 and Mantle, it seems to me that Mantle will have its one or two years in the sun and will then sail into the sunset, or evolve into some side API that works next to DX12 exposing AMD-specific functionality only.
AMD Mantle API [updating]
- Thread starter MarkoIt
- Start date
Since Mantle already runs with HLSL as a development path, and DX12 is so similar to Mantle, a Mantle 1.x version could be updated to overlap DX12 and present some kind of architecture-specific extensions.
Two years is a bit of a wait for game engines waiting for DX12.
I'm wondering if an AMD vs Nvidia side-taking is going on, given Epic's full embrace of Nvidia and DX12 versus EA and AMD.
Two years is a bit of a wait for game engines waiting for DX12.
I'm wondering if an AMD vs Nvidia side-taking is going on, given Epic's full embrace of Nvidia and DX12 versus EA and AMD.
It seems quite clear to me that Mantle has been very influential in the development of DX12.
Nvidia development can be traced back 4 years when they started talking to Microsoft about lower level API's but that in no way means that Microsoft was listening at that time or had plans to develop a new DX version based on NV's wishes.
In fact AMD has more or less said they were having the same conversations with Microsoft, saw them going no-where and decided to create an API of their own.
Nvidia says actual development of DX12 started about a year ago - which fits in pretty nicely with when they likely starting hearing about Mantle - or at least when they started taking it seriously.
And now we have repi stating that the design is pretty much based on Mantle.
If I had to bet on it I'd say if there had been no Mantle, there would be no DX12.
So Mantle has about 18 months of plain sailing but once DX12 lands I expect it to become more or less obsolete. However that may well have been the plan/hope of AMD and especially DICE all along. Mantle seems to have done it's job, AMD can now just sit back and enjoy the marketing advantages it will bring in over the next 18 months.
The weird thing is that Mantle provides DX12-type capability to Windows 7, but it's not clear if DX12 is going to do that. It might not matter as much in 2 years, but that's a pool of DX12-ready hardware that won't be on DX12.
If DICE's wish list for non-Windows support for Mantle comes true, then things could get even weirder.
If DICE's wish list for non-Windows support for Mantle comes true, then things could get even weirder.
Well we may need Mantle again to push for DX13 (or 14 if they don't like that number). It's not easy to push a reaction-based 800 pounds with only words, graphic tech also doesn't stop at DX12. Standard is nice, but it isn't as agile as a vendor-lock ApiWorks. Linux and Android can also be AMD's playground for Mantle.
I must admit, I'm still happily rocking Windows 7 but if Microsoft ties DX12 to Windows 9 then I'll be first in line to upgrade.
The problem with Windows 8 was that it seemed like a kick in the face to PC gaming (and just enthusiast computing in general) but if it comes tied to DX12 then Windows 9 will be much more attractive as a "gamers" OS.
The problem with Windows 8 was that it seemed like a kick in the face to PC gaming (and just enthusiast computing in general) but if it comes tied to DX12 then Windows 9 will be much more attractive as a "gamers" OS.
why ? windows 8 and 8.1 are much faster than 7. The $30 bucks ms was charging for it was a steal.
In that context, it's worth asking if ARB will bring something like dx12 to opengl? Or was this presentation a way to make devs live with what they have?
Yes, they look *very* similar in most concepts. dx12 looks to be a little more bindless (one of the things that stroke me in the mantle talks yesterday was that it's not just using hlsl syntax for the shaders - it's actually the same shadermodel / can use the same shaders as dx11 version, because you can make the same binds, just in another way from the host), and some mantle concepts like the seperate render qeues (compute+gfx at the same time) weren't mentioned. I could also suspect that mantle got some gcn-level hardware requirements that dx12 won't make.
It was quite strange in the last dx12 talk today, held by amd - it was really a pure dx12 talk with no mantle mentioned anywhere (ofcourse ms/nv won't make the reference, even when they are blatantly clear, but in an amd held session, where many would have seen the mantle stuff from yesterday). Even when asked about it, they refused to make the comparison, saying that that there had been plenty of mantle talks yesterday and this was about dx12.
By definition I guess discussions about DX12 started the day DX11 was out of the door...
I think microsoft said something about they started the work last fall (he said that in the first part, just before the forza demo).
But in that last talk they also emphasized the more iterative nature of dx12 development, compared to prior versions. Also implying that they were about done with the very first iteration of the api and there will be several more iterations (with drivers, engine integrations etc) before release.
Last edited by a moderator:
By saying, hey look, here is DX12, it's basically the same thing as Mantle, this is the same and this is the same etc. they would be sending the wrong message. I think they wanna keep Mantle as separate from DX12 as possible.
Its like Nvidias Nsync. We all know something free and non vendor specific is gonna come out and we all know Nvidia will still tout Nsync as the best thing out there.
D
Deleted member 13524
Guest
So Mantle takes a 2-year lead over Direct3D 12.
If AMD manages to quickly push Mantle as a single API to develop for both consoles and PC, Epic could be put in a complicated situation by choosing to take nVidia's "usual paycheck" again and not supporting Mantle with UE4.
The install base of GPU's that will support the D3D12 API (including Fermi, Kepler, Maxwell, Haswell, GCN, future GPU's from the three major PC IHV's, and future GPU's from ultra mobile vendors such as Qualcomm, ImgTech, ARM, etc.) will absolutely dwarf the install base of GPU's that support the GCN-only Mantle API. Obviously it is much quicker for one IHV to push out an IHV-specific and GPU-specific API (which happens to be in the beta stages too) compared to an industry standard and non-IHV specific API that will shape the future of gaming on all Windows platforms.
FWIW, Xbox One will use D3D12 as well. Unreal Engine 4 will work great for PC, console, and upcoming mobile hardware too.
Last edited by a moderator:
UE4 also incorporates NV Game Works. In concert with DX12 I'd be expecting something very close to console levels of efficiency on UE4 based games with NV hardware from 2015 onwards.
Incidentally, since Mantle clearly has a limited shelf life, I think AMD should look to start developing it in the direction of Game Works so that it works hand in hand with DX12 (and 11 for that matter) to eek the most out of the GCN architecture without having to rely on a separate API.
I will slap my mother on the ass and call her Susie the day Mantle is available on non-GCN hardware. Future AMD architectures notwithstanding.
Priyadarshi
Newcomer
Looking at the Mantle.dll I see a generic OpenCL like graphics API and nothing Nvidia/Intel can't support.
Have a look at its functions : http://pastie.org/8956471.
Have a look at its functions : http://pastie.org/8956471.