The AMD Mantle Thread

[SiliconWars]

I just can't believe people keep claiming mantle will cut costs. The only way it can cut costs is if fewer resources are directed toward direct x. Or at least for the developers, if AMD subsidizes the development. The additional resources needed may be worth it, but it will require additional time (money) to add another code path and test/debug it.

In the end, AMD doesn't really care if DX gets less resources directed towards it. In fact I'm sure that's their ultimate aim regardless of what they are claiming otherwise.

I also an not convinced it will lead to a huge increase in game sales. Console sales are probably 90% over the market for new big name titles. Even if mantle were to increase PC sales of a game 50% (highly unlikely in my opinion) that would only be a 5% increase in total sales, assuming none of those sales were taken away from consoles.

It seems clear that the easy path to a PC game is Mantle, far more than DirectX is, so if the PC market is so worthless then that just means the devs are even more likely to stop caring about DirectX and put more "effort" into Mantle instead. The problem is the rest of the PC industry on DirectX - the devs would much rather this problem went away.

I believe that less effort will be put into DirectX, yes. Over time the performance/quality gap will grow until Mantle is the dominant PC API. This is AMD's aim, whether or not they will fulfil it is another matter. They have the clear technology lead but I don't see Nvidia and Intel just sitting on their backsides letting it happen. In this regard I'm far more concerned about what damage Intel can do to Mantle than what Nvidia can do.

 

[krumme]

In the end, AMD doesn't really care if DX gets less resources directed towards it. In fact I'm sure that's their ultimate aim regardless of what they are claiming otherwise.

It seems clear that the easy path to a PC game is Mantle, far more than DirectX is, so if the PC market is so worthless then that just means the devs are even more likely to stop caring about DirectX and put more "effort" into Mantle instead. The problem is the rest of the PC industry on DirectX - the devs would much rather this problem went away.

I believe that less effort will be put into DirectX, yes. Over time the performance/quality gap will grow until Mantle is the dominant PC API. This is AMD's aim, whether or not they will fulfil it is another matter. They have the clear technology lead but I don't see Nvidia and Intel just sitting on their backsides letting it happen. In this regard I'm far more concerned about what damage Intel can do to Mantle than what Nvidia can do.

My take on the Intel side is that they are very happy the console win got the amd way because it meant x86 support. Very happy...

What Intel dont want is some linux on arm ecosystem to start to grow.
Mantle is dangerous but also safeguarding the x86 system. Its got two sides. And i am pretty sure the non arm approach of Mantle makes sure Intel will not hinder it.

An unsuccessfull mantle will lead to quicker and stronger arm adoption on future consoles and multimedia platforms.

Ms dont have other choice than to lean the same way because win8 have proved ms is tied to x86 success. Dx is just losing it. With the new consoles its just a bad solution and mantle is just flat out better by a huge margin. Either mantle will succeed or something else like mantle in combination with new consoles.

 

[frozentundra123456]

In the end, AMD doesn't really care if DX gets less resources directed towards it. In fact I'm sure that's their ultimate aim regardless of what they are claiming otherwise.

It seems clear that the easy path to a PC game is Mantle, far more than DirectX is, so if the PC market is so worthless then that just means the devs are even more likely to stop caring about DirectX and put more "effort" into Mantle instead. The problem is the rest of the PC industry on DirectX - the devs would much rather this problem went away.

I believe that less effort will be put into DirectX, yes. Over time the performance/quality gap will grow until Mantle is the dominant PC API. This is AMD's aim, whether or not they will fulfil it is another matter. They have the clear technology lead but I don't see Nvidia and Intel just sitting on their backsides letting it happen. In this regard I'm far more concerned about what damage Intel can do to Mantle than what Nvidia can do.

I never said PC gaming is worthless. In fact I dont even own a console, and all the games I play are on the PC. The simple fact is though that console sales far outweigh PC sales.

As far as Intel and nVidia trying to counter mantle, what do you expect? Of course they will. AMD is basically trying to take over the gaming market, which even if mantle is more efficient in the short term, I dont see as a good thing. Not to mention that direct x will still be required for older games and newer games that do not use mantle.

 

[desprado]

I never said PC gaming is worthless. In fact I dont even own a console, and all the games I play are on the PC. The simple fact is though that console sales far outweigh PC sales.

As far as Intel and nVidia trying to counter mantle, what do you expect? Of course they will. AMD is basically trying to take over the gaming market, which even if mantle is more efficient in the short term, I dont see as a good thing. Not to mention that direct x will still be required for older games and newer games that do not use mantle.

DX is going no where though even Mantle is using Dx features and it is implemented on Dx 11

 

[frozentundra123456]

My take on the Intel side is that they are very happy the console win got the amd way because it meant x86 support. Very happy...

What Intel dont want is some linux on arm ecosystem to start to grow.
Mantle is dangerous but also safeguarding the x86 system. Its got two sides. And i am pretty sure the non arm approach of Mantle makes sure Intel will not hinder it.

An unsuccessfull mantle will lead to quicker and stronger arm adoption on future consoles and multimedia platforms.

Ms dont have other choice than to lean the same way because win8 have proved ms is tied to x86 success. Dx is just losing it. With the new consoles its just a bad solution and mantle is just flat out better by a huge margin. Either mantle will succeed or something else like mantle in combination with new consoles.

You have a point, but I dont see mantle as a plus for intel, if it becomes widely adopted and brings tremendous gains, neither of which I take as a given. With high end systems it is probably not a problem for intel, as a discrete gpu will still be used, and Intel doesnt really care whether it is nVidia or AMD. However, in mobile, Intel has been steadily gaining ground on AMD APU graphics performance, and IF mantle delivers as some expect, it would be a big setback for Intel in that area. Again though, we come to the fragmentation problem. Considering the big lead in market share intel has in mobile, and the apparent gains in tablets with the new atom, it is highly unlikely DX will disappear.

 

[SiliconWars]

Krumme makes a good point in that x86 stays dominant under Mantle. It's just really bad news for Intel on the APU front, but do they really care about that anyway?

 

[redzo]

Yes,except GCN hardware is the basis for the two most powerful/popular consoles on the market that will sell in the 10's of millions and also AMD graphics cards which have about 40% of the discreet graphics card markets.
That's a lot different to Physx which is limited to NVDA graphics cards.

As a game studio, you're a money making machine. You cannot/will not ignore nvidia's x86 market share, nor you cannot direct dev resources from dx to mantle(this will translate into worse dx perfomance, more bugs, etc..). This requires additional dev resources, which means additional dev cost. No company follows this logic. It's the other way around: decrease dev cost by any means possible, increase profits. DX's and opengl's strength is that they are portable between the two of the most dominant gpu manufacturers. This is how and why they were born in the first place. I can't see how people keep saying that mantle reduces dev cost.

 

[dacostafilipe]

I just can't believe people keep claiming mantle will cut costs. The only way it can cut costs is if fewer resources are directed toward direct x. Or at least for the developers, if AMD subsidizes the development. The additional resources needed may be worth it, but it will require additional time (money) to add another code path and test/debug it.

Working with a SDK/Framework/Library limits what you can do with your code. And that's okay in most cases, because it speeds development time.

But, sometimes you do have the time/money to spend on optimizing your application. And when you do, you find yourself spending more time debuging the SDK/Framework/Library then writing real code.

You then try to use an alternate SDK/Framework/Library that let's you go deeper (ex: moving Cocoa code directly to Quartz, Cocoa's underlaying interface).

Mantle is just that deeper SDK/Framework/Library.

Yes, it will take some time to learn a new Library and yes not every dev will be able to use it (know-how missing), but in the end (mid- to long-term) you will win a lot of time because you reach your goal faster.

So yes, it can (!) cut costs ...

 

[psoomah]

As a game studio, you're a money making machine. You cannot/will not ignore nvidia's x86 market share, nor you cannot direct dev resources from dx to mantle(this will translate into worse dx perfomance, more bugs, etc..). This requires additional dev resources, which means additional dev cost. No company follows this logic. It's the other way around: decrease dev cost by any means possible, increase profits. DX's and opengl's strength is that they are portable between the two of the most dominant gpu manufacturers. This is how and why they were born in the first place. I can't see how people keep saying that mantle reduces dev cost.

Might be a compelling argument if actual game developers weren't showing so much enthusiasm for Mantle and both EA/DICE and Oxide indicating once integrated into an engine, providing a Mantle path in a game is sufficiently resource light the expect it will be a clear cost/profit winner.

As for Nvidia, with Mantle in play the devs would applaud AMD cornering 80% of new AIB sales.

 

[DarkKnightDude]

Can anyone explain what's the benefit of 10 times more drawcalls?

 

[SiliconWars]

Can anyone explain what's the benefit of 10 times more drawcalls?

In a nutshell, improved graphics, and more easily. Instead of devs being forced to use "instancing" (which is basically using the same graphic multiple times) because of the DirectX draw call issue, they will be able to use different graphics far more often.

By removing this overhead they actually make it easier for the really good artists to make much better graphics, and the really good coders no longer have to employ so many tricks and hacks to get it to work.

http://www.bit-tech.net/hardware/graphics/2011/03/16/farewell-to-directx/2

 

[dacostafilipe]

As a game studio...

Yes, for a game studio that has to balance devs between the game engine and the game creation (NOT the same thing!) this will be difficult to support.

But that's not the case for the game studios that use 3rd-party engines (Unreal, Crytek, ...) or for those who have a special division for the game engines (EA with Dice). The people creating the engines already spend a lot of time maxing DX out for specific brands/models and Mantle will simplify their work.

 

[krumme]

I never said PC gaming is worthless. In fact I dont even own a console, and all the games I play are on the PC. The simple fact is though that console sales far outweigh PC sales.

As far as Intel and nVidia trying to counter mantle, what do you expect? Of course they will. AMD is basically trying to take over the gaming market, which even if mantle is more efficient in the short term, I dont see as a good thing. Not to mention that direct x will still be required for older games and newer games that do not use mantle.

I am pretty sure its around 50/50% now from the recent estimates i have seen.

 

[krumme]

You have a point, but I dont see mantle as a plus for intel, if it becomes widely adopted and brings tremendous gains, neither of which I take as a given. With high end systems it is probably not a problem for intel, as a discrete gpu will still be used, and Intel doesnt really care whether it is nVidia or AMD. However, in mobile, Intel has been steadily gaining ground on AMD APU graphics performance, and IF mantle delivers as some expect, it would be a big setback for Intel in that area. Again though, we come to the fragmentation problem. Considering the big lead in market share intel has in mobile, and the apparent gains in tablets with the new atom, it is highly unlikely DX will disappear.

I agree its not a plus. But its unavoidable. Mantle is dx a long way as i understand it. And dx will not disapear as long as windows is ruling. Mantle will probably just be dominant on the performance side for both gfx and sound for the next 5 years.

But what can intel do to counter it without speeding arm adoption?

As it seems we will eg get ps4.1 4.2... with improved gfx. Right now bf4 runs 900p an probably at something like high medium ultra. I think we will see future small iteration of ps that can handle 1080p and ultra quality. Probably already on 20nm we will see jaguar++ cores and perhaps double the shaders.

Intel must be very interested in supporting that development because it keeps arm at distance from their cashcows while they grow new business or go more the foundry way. It buys them valuable time.

 

[krumme]

Krumme makes a good point in that x86 stays dominant under Mantle. It's just really bad news for Intel on the APU front, but do they really care about that anyway?

Yes they care but its hardly very important for profit. But they care far more for their cashcows. Muhhh

We will see if mantle helps kabini. If it does make a difference ofcource it makes the Intel solution less competitive but i hardly think it matters compared to securing their core business.

The importance on the low end becomes far more important on that perspective. I would like to see if a kabini 5200 2ghz could actually be playable in bf4 at lowest res 720p at lowest quality with mantle at around 25 fps. That would be funny but hardly doable

 

[krumme]

Working with a SDK/Framework/Library limits what you can do with your code. And that's okay in most cases, because it speeds development time.

But, sometimes you do have the time/money to spend on optimizing your application. And when you do, you find yourself spending more time debuging the SDK/Framework/Library then writing real code.

You then try to use an alternate SDK/Framework/Library that let's you go deeper (ex: moving Cocoa code directly to Quartz, Cocoa's underlaying interface).

Mantle is just that deeper SDK/Framework/Library.

Yes, it will take some time to learn a new Library and yes not every dev will be able to use it (know-how missing), but in the end (mid- to long-term) you will win a lot of time because you reach your goal faster.

So yes, it can (!) cut costs ...

And it will do the above because the learning cost is covered by the consoles the rest is covered by the major engines.

 

[krumme]

Amd say they will make mantle api open. Its easy to say when they are in controll of the mantle driver. Its bs. Its excactly the same strategy as Intel is using. We have been over it with Intels dispatcher in their compilers.
The core of Mantle is the driver. And the connection here with future hardware upgrades will keep amd constantly at an advantage. Ofcourse. Its the purpose.

 

[Paul98]

Has there been any info from this "Mantle: Empowering 3D Graphics Innovation"? I was looking online but didn't see anything.

 

[Kenmitch]

Has there been any info from this "Mantle: Empowering 3D Graphics Innovation"? I was looking online but didn't see anything.

Later today? Think that was stated in thread yesterday?

 

[moonbogg]

Why doesn't Microsoft revolutionize direct X and give it low level capabilities? Wouldn't that bury Mantle and give everyone a single, low level API to use as well as secure Microsoft's dominance at the same time?

 

[smackababy]

Why doesn't Microsoft revolutionize direct X and give it low level capabilities? Wouldn't that bury Mantle and give everyone a single, low level API to use as well as secure Microsoft's dominance at the same time?

Because they haven't had to. They went away from the direct, low level access when they abandoned DOS. A bunch of devs at the time thought this abstraction from the hardware would stall game development and make it terrible on Windows. MS introduced DirectX (or whatever it was called at the 1.0 stage) and all was well. It performed well enough, considering the hardware limitations. Now though, developers want to spend all their resources on making the game look as good as possible, so the "to the metal" argument as resurfaced. Too bad a bunch of those developers (including ones that claim their games are 60% graphics) haven't figured out to how to spend some of that money on a good writer or two or actually think about gameplay decisions before they go "well this will look really good!".

Gaming has suffered since we were no longer really technologically held back. Rather than coming up with new, inventive game play ideas or good stories to carry the low res sprites and text dialog, we get "need more graphics! omg!" But, I guess that is what sells. People still love Crysis and always cite how good it looks. The game sucked. If it had used the source engine, we'd never had heard of it.

 

[Saylick]

Why doesn't Microsoft revolutionize direct X and give it low level capabilities? Wouldn't that bury Mantle and give everyone a single, low level API to use as well as secure Microsoft's dominance at the same time?

Due to how DirectX inherently works, MS can't just change a few lines of code and voila, Mantle-killer.exe.

Simply put, DirectX was designed from the get-go to hide the physical graphics hardware behind an abstraction layer. This enables differing graphics architectures to all run the same DirectX code so long as the architectures in question support a myriad of specifications, which we know as DirectX 1.0 - 11.0, etc. Mantle, on the other hand, is specific to the GCN architecture. Code written in Mantle will call on specific execution units within the GCN-based graphics card, so nVidia GPUs can't run Mantle code unless you compile the Mantle code into something that an nVidia GPU can understand.

If you are familiar with high-level vs. low-level computer languages, this all should make sense to you. Think of DirectX as something along the lines of Python while Mantle is something closer to C. There is a speed penalty associated with translating higher-level languages into a lower-level language, but the benefit of a higher-level language is that the syntax is generally more "natural" to the user and is thus more user friendly. Note that in my analogy, Mantle isn't a super low-level language like assembly language, for example, because there is still a degree of user friendliness that is retained. In this case, DirectX has a higher abstraction penalty than Mantle, which is why we will see performance gains (via higher efficiency and better utilization of the underlying hardware) when Mantle is used.

 

[mak360]

Does anyone know what time APU13 starts? Is there a live stream going on?

Yes I'm too lazy to use google.

Register on this Link and confirm email and you're done. Next one could be interesting :whiste:

Search for
APU365 Registration - InXpo

Need to confirm Email and you're done.

 

[smackababy]

I do think MS could make a direct hardware access path (pun? what?), even if it wasn't native DirectX. They could add in a low level API to the next version of DX and give developers the chance to do it. And then, given specific calls, "force" the hardware vendors make their drivers accept such calls in accordance with the new API.

If they put the ball in the hardware maker's court and say "make your drivers recognize this call and act the same on all supported hardware to be DX 12 certified", it would gain a lot of steam I believe. MS could still obtain an abstraction layer above hardware (or architecture) specific, and allow developers who want to skip the DX path (which is lengthy) in favor of their own, possibly more efficient or powerful, path.

 

[SiliconWars]

I do think MS could make a direct hardware access path (pun? what?), even if it wasn't native DirectX. They could add in a low level API to the next version of DX and give developers the chance to do it. And then, given specific calls, "force" the hardware vendors make their drivers accept such calls in accordance with the new API.

If they put the ball in the hardware maker's court and say "make your drivers recognize this call and act the same on all supported hardware to be DX 12 certified", it would gain a lot of steam I believe. MS could still obtain an abstraction layer above hardware (or architecture) specific, and allow developers who want to skip the DX path (which is lengthy) in favor of their own, possibly more efficient or powerful, path.

Meanwhile AMD is busy attempting world domination with Mantle and Nvidia is basically all-in with OpenGL.

DirectX is finished, it's just that AMD and Nvidia went about it in different ways.