AMD: No Such Thing As 'Full Support' For DX12 Today

[Icecube26]

Problem is DX12 is windows 10 exclusive, and DX11 will support more windows OS's.

There are more traction this time, compared to previous iterations of DX.

you forget that consoles exist... and that developers are making consoles versions first... so there is very high change that consoles will also use modified versions of DX12 and vulkan to get that magic 60 FPS... its going to be easier for developers to port games to pc when it is going to using same API as consoles.... the only chance for people to play new games on older OS will be Vulkan ... since it is going to be multiplatform as well..if i were u i would upgrade to win 10 because not every game will run on vulkan...

[Tomsen]

you forget that consoles exist... and that developers are making consoles versions first... so there is very high change that consoles will also use modified versions of DX12 and vulkan to get that magic 60 FPS... its going to be easier for developers to port games to pc when it is going to using same API as consoles.... the only chance for people to play new games on older OS will be Vulkan ... since it is going to be multiplatform as well..if i were u i would upgrade to win 10 because not every game will run on vulkan...

Consoles won't gain anything off DX12 really, not in performance.

So your point is moot.

[Majestic]

Consoles already have API's with low overhead. It's why consoleports are shit right now, the console API is actually better than DX11. Funnily enough.

So DX12 will only make the gap more apparent between the PC and console platform. 

[GoodBytes]

No. They have lower level APIs, yes. They can afford this because they know that the hardware doesn't change (beside drive space).

WiiU and PS4 uses OpenGL. Usually a modified version of OpenGL with these lower level APIs, so you can't just easily port a game from Console to PC by just recompiling it for Windows.

The XBox One supports DirectX. We don't know if it has special lower level APIs inside (probably it does). DirectX12 might benefit a bit the XBox One, able to achieve a bit better graphics, bringing it closer to the PS4, is a possibility on well coded games.

[ACatWithThumbs]

Consoles won't gain anything off DX12 really, not in performance.

So your point is moot.

The XboxOne still uses it and DX12 has more similar code to the consoles than DX11 does which makes porting over easier.

And easier port = faster = less work = more $$$

It also allows for better performance on slower hardware which means it widens the consumer base in general.

And it allows for better graphics and shiny graphics sell better especially if it runs good.

[Briggsy]

Although the consoles already use low level API's, its only now (afaik) that the asynchronous Compute Engines are starting to really get used. Naughty Dog are making considerable use of the ACE's with Uncharted 4 so they can crank out more frames and effects. This trend probably will continue as it makes no sense not to.

[Tomsen]

The XboxOne still uses it and DX12 has more similar code to the consoles than DX11 does which makes porting over easier.

And easier port = faster = less work = more $$$

It also allows for better performance on slower hardware which means it widens the consumer base in general.

And it allows for better graphics and shiny graphics sell better especially if it runs good.

Xbox one most likely use a modified version of DX. That is tied to the hardware.

Not if you limit your userbase like that.

[Briggsy]

Xbox one most likely use a modified version of DX. That is tied to the hardware.

Not if you limit your userbase like that.

 

Xbone still using 11.3 I think

[zMeul]

No. They have lower level APIs, yes. They can afford this because they know that the hardware doesn't change (beside drive space).

WiiU and PS4 uses OpenGL. Usually a modified version of OpenGL with these lower level APIs, so you can't just easily port a game from Console to PC by just recompiling it for Windows.

no, PS4 doesn't use OpenGL at all, not even a heavily modified version of it

PS4 uses 2 APIs - one low level called GNM and a 2nd one, GNMX (not low level, kinda' like DX11)

[Briggsy]

no, PS4 doesn't use OpenGL at all, not even a heavily modified version of it

PS4 uses 2 APIs - one low level called GNM and a 2nd one, GNMX (not low level, kinda' like DX11)

 

pretty much

 

http://gamingbolt.com/ps4-should-support-vulkan-ps4s-api-not-completely-native-for-current-gen-yet-brad-wardell

[Tomsen]

Xbone still using 11.3 I think

They use DX 11.X

Which is a superset of DX 11.2 (do feature some DX12 features).

[GoodBytes]

no, PS4 doesn't use OpenGL at all, not even a heavily modified version of it

PS4 uses 2 APIs - one low level called GNM and a 2nd one, GNMX (not low level, kinda' like DX11)

Yes. I was mistaken.

Thanks for the correction

[Briggsy]

They use DX 11.X

Which is a superset of DX 11.2 (do feature some DX12 features).

 

You are indeed correct. 11.3 launched with 12 apparently. http://www.anandtech.com/show/8544/microsoft-details-direct3d-113-12-new-features

 

I dunno why i though 11.3 was a thing before now...

[alamox]

i hope vulkan, and steamOS takes off, because i hated microsoft before and their OS/DX, with Win10, and generaly where the firm is going, i really wish for a reasonable alternative to microsoft's OS.

[Linus_torvalds]

Maxwell supports async compute, this isn't even the question

No it does not!

Asynchronous computing is AMD only tech Nvidia does not have anything like that!

[zMeul]

No it does not!

Asynchronous computing is AMD only tech Nvidia does not have anything like that!

maybe I'm wrong but isn't this it?

http://devblogs.nvidia.com/parallelforall/how-overlap-data-transfers-cuda-cc/

also, this: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#asynchronous-concurrent-execution

[Briggsy]

maybe I'm wrong but isn't this it?

 

that appears to be the task scheduling AMD pointed out in their (now famous /s) async compute youtube video, which incurs (or can incur) latency penalties.

 

 

From what I've gathered so far:

 

Even though AMD can queue 64 compute tasks, they are actually queuing up to 8 compute tasks per compute engine, with 8 compute engines working in parallel and also in parallel with the graphics lane, so all 9 "lanes" are working together down to the millisecond.

 

Nvidia have a single engine that can queue up to 31 compute tasks, and 1 graphic task. the engine (or warp scheduler thingy) cannot queue up compute tasks while there is a graphic task in the pipeline. In theory that would mean that Nvidia would have to heavily prioritize and/or limit compute tasks from piling up in a real world game engine that is using compute at random, and having to optimize drivers for each game to ensure that workloads are prioritized, with or without async compute enabled.

[Sintezza]

the 370X = 270X = 270 = 7870Ghz = 7870.

They are all fully unlocked pitcairn chips with 1280 stream processors.

[Notional]

maybe I'm wrong but isn't this it?

http://devblogs.nvidia.com/parallelforall/how-overlap-data-transfers-cuda-cc/

also, this: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#asynchronous-concurrent-execution

 

That is compute, not compute + graphics. That is the problem. No one questions nvidia's ability to do async compute on things like Tesla. The problem is in games with compute + graphics.

[Prysin]

you are all wrong guys.

 

the 370X = 270X = 270 = 7870Ghz = 7870.

They are all fully unlocked pitcairn chips with 1280 stream processors.

and WE WERE NOT TALKING OF THE R7 370 x

 

we were talking about the R7 370 -> WITHOUT A X

[Sintezza]

and WE WERE NOT TALKING OF THE R7 370 x

 

we were talking about the R7 370 -> WITHOUT A X

 

ah i guess im blind sorry lol.

the R7-370 is indeed 7850 1024 stream processors.

 

Still im a bit dissapointed with the  rebranding of this old pitcairn chip tbh.

R9-370X would have made more sense to me, if they had simply used a gimped down version of Tonga.

This way, the 370X would atleast support true audio and freesync.

[Prysin]

ah i guess im blind sorry lol.

the R7-370 is indeed 7850 1024 stream processors.

 

Still im a bit dissapointed with the  rebranding of this old pitcairn chip tbh.

R9-370X would have made more sense to me, if they had simply used a gimped down version of Tonga.

This way, the 370X would atleast support true audio and freesync.

at this end of the spectrum, you arent going to have so much use of freesync just yet... we need freesync monitors to reach its lowest VESA standard refresh rating first, that is 9 Frames per seconds...

 

once we get a cheap (sub 150 USD) monitor with 9-75 FPS adaptive sync range, THEN freesync makes sense at this price range..

True audio is a great concept, but too few game developers are using it...

 

I would expect DICE to use it, perhaps for Battlefront. As it would be epic for FPS games given that it does produce more precise sound placement.

[Linus_torvalds]

maybe I'm wrong but isn't this it?

http://devblogs.nvidia.com/parallelforall/how-overlap-data-transfers-cuda-cc/

also, this: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#asynchronous-concurrent-execution

That is NOT for gaming

[zMeul]

That is NOT for gaming

who said it's for gaming? Async Compute isn't for gaming, Async Shader is

[xAcid9]

who said it's for gaming? Async Compute isn't for gaming, Async Shader is

Got link to back that up?