AMD Will Release Mantle Programming Guide, API Reference This Month

[ravyne]

Really? I don't see all 3 consoles being a failure. More to the point, Mantle will most certainly be adopted by vendors developing across PS4/XBox and Nintendo WII replacement.

Next Gen OpenGL won't be mature for another 24-36 months. Directx 12 will take another 12-18 months. Meanwhile, vendors are wise to use Mantle.

Any one who thinks R9 300 series GPGPUs won't fly off the shelves knowing vendors will be leveraging the crap out of them are very short-sighted in their reasoning.

PC Gaming and Mantle will only grow, right along-side Consoles. Nvidia's dominance will dwindle as more games get far more performance out of AMD GPGPUs and switch their backends to leverage it. This means more used Nvidia cards on EBay.

That's not really how consoles work -- I have it on good word that the PS4 low-level API is an awful lot like Mantle. Xbox Developers have D3D.11x, which has capabilities beyond D3D11.2 on PC to expose platform specific capabilities and which has also been trimmed of the overhead needed to abstract the open PC ecosystem. No idea what the next Nintendo offering will look like API-wise, but Nintendo's APIs have historically resembled OpenGL. But you only really get what the platform vendors offer you -- you can't just choose to use Mantle, or OpenGL, or Direct3D -- effectively your choices are A) engines and middleware like Unity or Unreal Engine, or B) whatever low-level API the vendor exposes.

Given that best evidence points to the next-generation openGL borrowing liberally, if not entirely, from Mantle, I think your timeline is off though -- Developers don't really need to ramp up with these new APIs since they're very close to what they've been doing on consoles for years (there have been reports of shops porting their console render paths to Mantle in a handful of days, some of them just hours) -- AMD will easily be able to pivot their Mantle driver code to Vulkan, and nVidia and Intel will have sufficient motivation to not lag far behind -- plus, they all have D3D12 drivers already, which won't themselves be a far cry from Vulkan, in all likelihood. I say 12 months before *most* PC releases support one or both Vulkan/D3D12, and just 6 months for high-profile games, engines, and middleware.

I don't this will have any significant impact on nVidia's business, other than the minor Halo effect of AMD's early (and likely more-stable, to begin with) support for these new APIs. I am looking forward to upgrading to a tasty new 390x though.

[carewolf]

RIP Mantle, useless single vendor effort. 3dnow was a failure, but at least x86-64 was widely adopted.

SSE replaced 3dnow!, just like AVX replaces SSE. I wouldn't call SSE a failure, and neither 3dnow! it was the first kick in the ass AMD send Intel's way before the big Athlon rump.

[bridgman]

@bridgman

If AMD wanted to support Mantle for LONG time then

A) Full specs would have be provided at release
B) Linux binaries would exist (since the beginning)

It was 100 % designed as technolgy with obsolescence.

OK, so I'm not directly involved with Mantle (other than sitting beside the guy who was driving the project) but my impression was that we wanted to change the direction of APIs and make them closer to (a) what the game developers wanted and (b) what the HW vendors wanted. I don't think Mantle on its own was necessarily the end goal, but "something in the direction of Mantle" definitely was.

I can believe you that it wasn't specific to AMD GPUs, but hard fact it is/was only available for AMD GCN GPUs... and thus that is same said to me .

Yeah, from a practical perspective that's true, but from the time we offered it to Khronos I think that was a significant step towards making it cross-vendor. We're not big into vendor-specific stuff, in case that's not obvious.

that was statement in second last announcement, last one specified "most of functionality does not require GCN" which would say at least something is GCN reliant

I'm not aware of anything in Mantle that is dependent on GCN, just on "recent GPUs". Problem is that without actually porting Mantle to competitor's HW it's hard to be absolutely sure so there is a bit of fuzziness in the statements.

[justmy2cents]

I'm not aware of anything in Mantle that is dependent on GCN, just on "recent GPUs". Problem is that without actually porting Mantle to competitor's HW it's hard to be absolutely sure so there is a bit of fuzziness in the statements.

makes sense, thanks

FYI, i read these statements on AMD presentations, but can't find link now

[dungeon]

Yeah, from a practical perspective that's true, but from the time we offered it to Khronos I think that was a significant step towards making it cross-vendor.

That is fine, but making something cross vendor arise a question which is not a question ... is there still same adventage, probably not all

We're not big into vendor-specific stuff, in case that's not obvious.

Yup obvious to me, i know what means "should work".

[grndzro]

That is fine, but making something cross vendor arise a question which is not a question ... is there still same adventage, probably not all
.

Mantle was never about giving AMD an advantage. It was about leveling the playing field for both it's processors and GPU's. The main gaming bottleneck was having a single thread responsible for the majority of processing. Mantle eliminates that bottleneck, and so will OGL-N/DX12. AMD gets it's Mantle and access to Nvidia's OGL goodies. Nvidia gets access to Mantle without the PR nightmare of actually supporting Mantle. Khronos gets practically instant AMD compatibility with OGL-N, and a huge code boost from using what they want from Mantle.

Nvidia already had a leg up on AMD driver wise. So it would still be a win for AMD on graphics.

[johnc]

Nvidia gets access to Mantle without the PR nightmare of actually supporting Mantle.

Prediction: AMD will have the worst OpenGL Next drivers.

[Kano]

Well Nvidia had at least a special multithreaded rendering mode since driver 310. Made for source engine based games first. AMD created some profiles but the csgo binary still has no special treatment. Lets see if they could reuse some Mantle code for Vulkan to gain speed. OpenGL was never AMD's favorite toy...

[smitty3268]

Prediction: AMD will have the worst OpenGL Next drivers.

All that matters is if they're good enough that no one cares. Nobody worries whether AMD or NVidia have the worst DirectX drivers, and that's what we need to hope for from Vulkan.

[justmy2cents]

All that matters is if they're good enough that no one cares. Nobody worries whether AMD or NVidia have the worst DirectX drivers, and that's what we need to hope for from Vulkan.

where are the mod points, i want to mod up this post by +34745047