DirectX, OpenGL May Soon Allow Low-Level Hardware Access

[genz]

I was pointing out that Android was geared for a different type of phone that has only emerged recently but if you want to twist things around then that's up to you, I'm done responding to you.

Good, then maybe you'll listen, because when the G1 was released, it was not a 'different type of phone' any more than Mantle is a 'different type of API'

 

[Mousemonkey]

I was pointing out that Android was geared for a different type of phone that has only emerged recently but if you want to twist things around then that's up to you, I'm done responding to you.

Good, then maybe you'll listen, because when the G1 was released, it was not a 'different type of phone' any more than Mantle is a 'different type of API'

I don't listen to text! :lol:

 

[genz]

I don't listen to text! :lol:

How about reason? Do you listen to that? 😀

 

[Mousemonkey]

I don't listen to text! :lol:

How about reason? Do you listen to that? 😀

Do you have any? Thought not.

 

[genz]

I give up, this is childish now. You win. Go collect your complimentary toilet brush.

 

[somebodyspecial]

Don't be silly Somebodyspecial, AMD will share Mantle as soon as they finish making it. Why? Because it serves them MUCH better if everyone uses it than if only they do. Remember that Mantle is an API, and Nvidia have no API, so Nvidia is not competition in that space. Assuming it does what it says no the tin, uptake will be fast, and naturally Nvidia will never see the speed boost that GCN will as Mantle is built specifically for that chip design... but it will still see significant boosts, forcing Nvidia to actually have to build a faster card than AMD just to keep up and allowing AMD bigger margins on similar performing models until Nvidia can redesign Mantle orientated chips instead of OGL/DX orientated chips.

This all naturally assumes that Mantle is reliable and hugely faster than DX/OGL. Fast enough that gaes are coming out that completely revolutionise gaming graphics. With the Carmack quotes and the 30,000 draw calls vs 3,000 DX draw calls, that may well be the case, but it's hard to separate HUMA and Mantle advantages so far.

What is the point? Mantle only works on GCN cards and Nvidia are not going to change their architecture any time soon (have you seen what Maxwell can do?) and as Mantle doesn't even work on older AMD cards why should anyone believe that Mantle will work on Nvidia cards?

No. Mantle only CURRENTLY works on GCN cards. Whilst we're talking about limitations that will change, Mantle only currently runs under 1 game so far.

As for why Nvidia will support Mantle, in case you didn't read the second part of my post, Mantle is proposing that it will be able to do 10x the rendering that DX could in the same time due to its much more direct approach. If it's only 2x the rendering in the end product that will still force Nvidia to adopt it as there will be a whole string of very shiny new games that will not run at max on anything but Mantle, suddenly taking something like a Titan and making it look like a 8600GT next to AMDs current generation.

Mantle claims to have 10x the draw call ability and hUMA. If it has either, Mantle only games or games that use Mantle only to generate the 'Ultra' graphics settings will start appearing. Nvidia needs the flexibility (to play Mantle only games) and the ability to keep up with AMD.


And @Mupples. That's called a HAL. DX already is a HAL. It is bypassable, but the whole issue with bypassing a HAL is that there's no such thing as a universal command structure beneath DX on a card as every generation of card speaks a different language even before we get to the differences between AMD and Nvidia (think of it being like they all have different instruction sets and DX is a Java virtual machine on top of them that allows DX applications to run no matter the stuff underneath).

Mantle is proposing the equivalent of a runtime 'C++' for Nvidia which is at the same time a runtime 'Assembly' for it's own cards. If Nvidia take the carrot, they can make their next generation of card as fast or faster than AMD's under Mantle.

GDC 2014 - OpenGL already has 10x just like mantle 😉 Nvidia doesn't need to re-create that which already exists in a form that runs everywhere and makes devs happy about supporting it due to that fact. DirectX looks to have the same thing soon too. But while Mantle is a beta, OpenGL isn't. Considering OpenGL runs on EVERYTHING on the planet (phones, tablets, iOS, OSX, Linux, PS3, PS4, Windows, and at some point MS will have to add it to xbox1 if directX fades), I have no fear of stepping out on a limb and saying Mantle is DEAD. I said it all along, it appears it's true even sooner than I think (in two weeks...ROFL). We done here? A new API that has a few cards that run it, backed by a company that is broke and can't even get out of phase3 drivers, or an established API that EVERYBODY uses? Which way would you go as a dev? OpenGL. PERIOD. DirectX if forced which at least covers ALL PC gpus and xbox1. Make a game for Mantle and it works on a SUBSET of a small SUBSET of the PC market only at least for the near future and I really don't see it running on anything but AMD for the long term. None of the other players would want it to live and no dev wants MORE API's.

I'd say just about the same for DirectX now that we have mobile adding so much to the sheer market presence of OpenGL devices and how easily you can port to everything if you use it (not to mention Valve pushing linux/opengl). MS won't add OpenGL willingly to Xbox1, but if enough devs tell them get it on that console or our games don't port to it, MS will add it or DIE vs. Sony who has already sold 2x more units. If you let games get massively made for PS4/Mobile/PC before being ported to DirectX last (smallest audience of all right now on xbox1) then you pretty much kill your sales. No software=No hardware sales. Chicken, meet egg 😉 Android has the complete opposite situation. A billion units are getting sold yearly, so here comes all the devs (GDC 2013, 60% of devs making mobile games, <15% for ALL consoles/handhelds). GDC will show even MORE devs going to mobile as sheer units to sell to forces you to consider doing a game there FIRST and port an amped up version to PC's etc. The fact that you have such a large audience allows a game to be sold for $10 and sell multiples more than a $60 title on consoles. Right now mobile has 120x the audience (1.2B+ units per YEAR 980mil phones/300mil tablets now), compared to next gen consoles (under 10mil sold so far). Where would you go? MOBILE. Devs are making the right moves.

 

[genz]

GDC 2014 - OpenGL already has 10x just like mantle 😉 Nvidia doesn't need to re-create that which already exists in a form that runs everywhere and makes devs happy about supporting it due to that fact. DirectX looks to have the same thing soon too. But while Mantle is a beta, OpenGL isn't. Considering OpenGL runs on EVERYTHING on the planet (phones, tablets, iOS, OSX, Linux, PS3, PS4, Windows, and at some point MS will have to add it to xbox1 if directX fades), I have no fear of stepping out on a limb and saying Mantle is DEAD. I said it all along, it appears it's true even sooner than I think (in two weeks...ROFL). We done here?

Not even close. You think supporting that many targets while maintaining a low footprint whilst managing your code so that there aren't kernel exceptions every time a game that the devs tested on AMD boards is used in an Nvidia one (the whole reason DX existed in the first place, and later OGL as an open source counterpart) is easy?

No. But it is doable.

Doable in less time that it took for AMD to bring it to market though.... I'm a skeptic.

ALSO...

These are extensions, whilst I am happy to say that yes they are faster, they are faster extensions, not a faster language. Unless you now make the entirety of the game using only these extensions then I don't think you will experience Mantle like speeds.... assuming that enough extensions are even there and possible.

ALSO...

There's a reason why it's being built just for GCN, and I think it's got a lot to do with some expanded instruction set that AMD is probably using. That's the only real reason I can see for them building a new driver along with a new API. If so, then no matter what is done OGL will not be the fastest low level API on consoles until Mantle releases it's source so that OGL can actually see these new instructions and understand how they are called, or at the very least releases a good portion of documentation on that. Remember GCN is really the first GPU built with running low level code for a long time. All GPUs before have been effectively OGL and DX ASICs with compute tacked on, which really just started from Nvidia tacking on PhysX hardware into it's GPU.

 

[youcanDUit]

does this mean that current hardware won't support DX12? i don't think i've seen a game support DX11.1 but if this is in lieu of AMD's efforts does that mean that upgrades will only benefit from it, even if it is 10-20 years (joke) down the line?