AMD Mantle API [updating]

[pharma]

NVIDIA said at GDC, at some other point, that DX12 has been in development for about a year, not 4.

Based on the article you are correct ... a year ago is when Nvidia started their collaboration with MS

For the past year, NVIDIA has been working closely with the DirectX team to deliver a working design and implementation of DX12 at GDC

Time will tell, but seems Nvidia may only need a driver update to achieve similar performance gains in DX11.

http://www.guru3d.com/news_story/nvidias_tba_dx11_will_be_better_than_mantle.html

 

[Kaotik]

Based on the article you are correct ... a year ago is when Nvidia started their collaboration with MS

Time will tell, but seems Nvidia may only need a driver update to achieve similar performance gains in DX11.

http://www.guru3d.com/news_story/nvidias_tba_dx11_will_be_better_than_mantle.html

There's been a lot of talk about it, but it's not all that simple, and even the biggest "theoretical gains" rarely realize in big %'s in games.
Even in their own examples while 1 gains some 80-90% the other just manages 10% improvement. Worth noting is that the benches were also done on extremely fast 6-core CPU, which makes Mantle seem a lot worse in comparison.

Before there's actual 3rd party benches proving these, I'd take even those actual game numbers with a pinch of salt, 80-90% with driver update seems just impossible unless there was something fundamentally wrong with your driver before, which the current benchmarks don't support. Unless your drivers have pretty much the whole game rewritten for your card with something that could just as well be called "custom API" to begin with

 

[DSC]

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Talks-DX12-DX11-Efficiency-Improvements

NVIDIA is currently the only GPU vendor to have a DX12 capable driver in the hands of developers and the demo that Microsoft showed at GDC was running on a GeForce GTX TITAN BLACK card. (UPDATE: I was told that actually Intel has a DX12 capable driver available as well leaving AMD as the only major vendor without.)

DX12 caught AMD with their pants down, deer in headlights.

 

[3dilettante]

This seems unlikely for any reasonable interpretation of that characterization.
Industry standard APIs take too long and involve too much communication between major stakeholders to say one of the major IHVs did not see it coming.
It seems rather unlikely that AMD blundered along oblivious to DX12, or that the whole industry kept it a secret from them, while at the same time AMD provided an API running in commercial products that just so happened to match many (almost all?) major architectural elements of an API that won't be in products for almost two years.

To say AMD fell down on delivering an early version of the driver on time or did not prioritize it sounds more in-character.

 

[heli]

Anyone have an idea how Nvidia have increased their DX11 performance so much?

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Talks-DX12-DX11-Efficiency-Improvements

I think the hint is the CPU they used. Can they show such gain on an old Core 2 Quad or APU like Mantle does? NVIDIA driver is known to spawn many driver threads to combat D3D overheads. It relies on high-end CPU and sufficient cores to get the most out of their optimization. I think they are getting really aggressive on that front in each driver. I know they are proud of their software team and I think they do a fantastic job, but it doesn't change the fundamental problem.

NVIDIA's DX11 driver supports driver command list and it probably helps a lot with deferred context enabled in StarSwarm, and their close tie to Unreal Engine can also help (though the label on y-axis can be misleading). Oxide admits they have done more optimization on D3D11 path than Mantle when they released the demo.


The only way they started "working" on DX12 four years ago would be AMD, Intel and game developer like repi were intentionally left out in the dark, in other words, Microsoft and NVIDIA wanted to conspire against them

 

[AlNom]

To say AMD fell down on delivering an early version of the driver on time or did not prioritize it sounds more in-character.

Yeah, I could see them putting more time into Mantle given the differences in schedules (e.g. not relying on various IHV/ISVs, DX12 still has to be finalized + OS support).

 

[nutball]

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Talks-DX12-DX11-Efficiency-Improvements

DX12 caught AMD with their pants down, deer in headlights.

AMD have no DX12 drivers, and this affects how many end-users precisely at this point in time? Or to turn the question around, NVs DX12 support benefits who, right now?

 

[Freemantle]

AMD have no DX12 drivers, and this affects how many end-users precisely at this point in time? Or to turn the question around, NVs DX12 support benefits who, right now?

Game developers, and the future adoption of DX12 on the PC and XBone.

 

[NRP]

And tinfoil hatters.

 

[DmitryKo]

AMD have no DX12 drivers, and this affects how many end-users precisely at this point in time?

Exactly. We are at least one year from the RTM release of Windows 9 (aka NT 6.4), and the spec freeze probably won't even happen before Christmas.

I've heard the reasoning that Microsoft might have teamed with NVidia simply because they do not want Sony to have any preliminary information on the direction Direct3D 12 is going, which can be used to further optimize the PlayStation 4 before Microsoft. However if you take GMN/GMNX, Sony's proprietary APIs on the PS4, one of them is a higher-level API, probably a further development of OpenGL ES with Cg/HLSL, but the other is some lower-level API which might be already quite close to Mantle concepts since this is what their 5th and 6th gen consoles were all about in the first place.

 

[Dave Baumann]

Wow. This thread is going into the realms of fantasy.

 

[Malo]

Wow. This thread is going into the realms of fantasy.

The only way to stop tin-foil speculations and crazy fantasies is to provide better information. AMD, Nvidia and MS are not going to provide that so they're stuck with their fantasies.

 

[dobwal]

AMD have no DX12 drivers, and this affects how many end-users precisely at this point in time? Or to turn the question around, NVs DX12 support benefits who, right now?

Pretty large leap in logic to assume AMD has no DX12 drivers just because Nvidia is the only vendor to have them in developers' hands.

Given that, whats the point of passing out the drivers for an unreleased unfinished api? Don't devs need both in hand? Seems to me that statement made by the article is based on Turn 10 having nvidia based drivers but Turn 10 is a MS team. How many outside of MS have the DirectX 12 api?

Nevermind the XBOX DX12 API will be dependent on an AMD driver (which has a good chance of seeing light before the PC api gets released). Or the fact that plenty of devs who are developing for the XB1 already have AMD drivers for a low level based Direct X api.

I highly doubt AMD is playing catch up in this regard.

 

[Dave Baumann]

Or the fact that plenty of devs who are developing for the XB1 already have AMD drivers for a low level based Direct X api.

This also explains the demo hardware. If you want to promote a cross vendor API would demoing Forza on AMD hardware have been a surprise to any developer?

 

[imaxx]

Nevermind the XBOX DX12 API will be dependent on an AMD driver

While it's true AMD skeleton's driver must come from AMD, I would be dubious about any thing more.
MS is the producer, MS tunes up their custom OS.
It is likely that MS got the sources of a 'working, minimal driver with all the features and bootup' and then customized it... heavily.
Which would make sense, given their crazy (read: drunk) architect's VM design.

 

[3dilettante]

The original quote also included an additional claim as to the availability of an Intel DX12 driver and the lack of something similar from AMD.
This goes beyond the context of the Xbox One to Nvidia demonstration and goes to a more general question of readiness.

One article's account isn't much data to justify the idea that AMD has somehow missed the DX12 train, but one would ask whether the article is misinformed or, if not, why the IHV with all the appearances of a massive head start appears to be finding ways to be behind.

Does the platform holder want to give other stakeholders a way to not lose face?

 

[firstminion]

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Talks-DX12-DX11-Efficiency-Improvements



DX12 caught AMD with their pants down, deer in headlights.

Excuse me, did they ask AMD? They didn't ask Intel before releasing that based on their update.

What I don't understand is how people are supposing that AMD doesn't have any knowledge of DX12 before GDC, that's preposterous.

 

[Freemantle]

What I don't understand is how people are supposing that AMD doesn't have any knowledge of DX12 before GDC, that's preposterous.

Well it seems no one told Roy Taylor.

http://www.heise.de/newsticker/meld...h-Es-wird-kein-DirectX-12-kommen-1835338.html

 

[lanek]

Well it seems no one told Roy Taylor.

http://www.heise.de/newsticker/meld...h-Es-wird-kein-DirectX-12-kommen-1835338.html

At the same time, when you look the date of the article.. nearly one year back.... During this time, AMD was allready working on DX12 with MS ( as have been told in the conferences of GDC by MS ).

 

[Freemantle]

At the same time, when you look the date of the article.. nearly one year back.... During this time, AMD was allready working on DX12 with MS ( as have been told in the conferences of GDC by MS ).

No doubt, but it was a slightly tongue in cheek reply to firstminion suggesting they didn't know before the GDC.

Although equally it is (apparently) around the time work on DX12 started in earnest (at least according to nVidia), so I would imagine at least plenty of initial discussions would/should have taken place.

Now that is ever complete bull or Microsoft/nVidia and the gang have managed to knock up a working DX12 demo in the time since the initial Mantle announcement around September?