The AMD Mantle Thread

[SiliconWars]

Money needs to come from somewhere. And its starting to be pretty obvious that AMD had to pay heavily for Mantle with DICE. Also, who do you think would be the first loser in that game? The one with the smallest or biggest marketshare? *Hint* AMD is the smallest player.

AMD will be close or have passed Nvidia's market share already due to console sales in Q3. By Q4 they will have.

Paying a few $million is peanuts for this exclusive access to some of the biggest games of the year.

 

[3DVagabond]

AMD will be close or have passed Nvidia's market share already due to console sales in Q3. By Q4 they will have.

Paying a few $million is peanuts for this exclusive access to some of the biggest games of the year.

Reading interviews after the 9/25 event, Dice and AMD have been working on this together for years. According to AMD this is something Devs have been crying about for at least 5 years. Of course there's a cost associated with the development. And you are right, taking into account the GCN arch used in the consoles, AMD's, and more specifically, GCN's market penetration is going to skyrocket.

 

[sontin]

AMD will be close or have passed Nvidia's market share already due to console sales in Q3. By Q4 they will have.

Paying a few $million is peanuts for this exclusive access to some of the biggest games of the year.

Sonic will be a must have this holiday. :awe:

Reading interviews after the 9/25 event, Dice and AMD have been working on this together for years. According to AMD this is something Devs have been crying about for at least 5 years. Of course there's a cost associated with the development. And you are right, taking into account the GCN arch used in the consoles, AMD's, and more specifically, GCN's market penetration is going to skyrocket.

Still believing that any sane company is developing a multi port game with a low level API in mind. A few facts for you guys:
All Multi-port games in the next 6 to 12 months will have no Mantle API except BF4: No Call of Duty, no Watch Dogs, no AC, no Batman, no Fifa, no PES...
GCN market penetration is playing no role because in a open standard world people will not limiting themself to <20% of the market for investing the same amount of money to design a DX path to archive 100%.

But we all need our dream world, i guess.

 

[Haider]

Does anyone have any real quotes from "respected" developers saying this? Because I've talked with quite a few developers and they don't believe DirectX is really that slow, especially compared to the ease of use and standarification it brings to PC development.

"2.5 times performance" is a rather bold claim, especially if you don't have any meaningful test to back it up.

Talking about something he wished he’d done differently during Rage’s development, Carmack said this:

“When we started on the game six years ago, I looked at the consoles and said ‘These are as good as the PC’, and our development strategy was to develop live on all the platforms. And now we’re looking at PCs that have ten times the horsepower of the consoles. I’m making a large change in my direction just saying ‘We should be building things efficiently on the PC and then deploying on the consoles.’ And we didn’t make that as crisp of a distinction as we could have.

“My development system now has twenty four threads and twenty four gigs of memory, and we can start putting on half a terrabyte of solid state drives, and these are the things that are gonna drive the development process on the PC. I’m actually as excited about how we’re developing tht titles in this coming generation as the graphic enhancements and things that I’m gonna make.

“…it is unhappily true that we have these consoles here running at sixty frames per second, and we could have these massively more powerful PC systems that struggle sometimes to hold the same framerate because of unnecessary overheads. If we were programming that hardware directly on the metal the same way we do consoles, it would be significantly more powerful.”


''The real reason to get excited about a PS4 is what Sony as a company does with the OS and system libraries as a platform, and what this enables 1st party studios to do, when they make PS4-only games. If PS4 has a real-time OS, with a libGCM style low level access to the GPU, then the PS4 1st party games will be years ahead of the PC simply because it opens up what is possible on the GPU. Note this won't happen right away on launch, but once developers tool up for the platform, this will be the case. As a PC guy who knows hardware to the metal, I spend most of my days in frustration knowing damn well what I could do with the hardware, but what I cannot do because Microsoft and IHVs wont provide low-level GPU access in PC APIs. One simple example, drawcalls on PC have easily 10x to 100x the overhead of a console with a libGCM style API.'' Timothy Lottes http://timothylottes.blogspot.fr


Thanks
Haider

 

[beginner99]

The only way that I can see Mantle working is if it is a straight port of the low-level console API. If it lets developers reuse the work they put into the optimized console version to get higher desktop performance, at a relatively low cost, I could see them going for it.

Yes. Thats how I understand it. You can port it like Java as long a a Mantle driver for your hardware exists.

Even if they did, developers would be faced optimizing the same piece of code for hundred pieces of hardware, as others have said repeatedly, DirectX being slow isn't because it's crap, it's because it can take the same codebase and make it work on any piece of hardware capable of implementing the DirectX API, in a world of thousands of different video cards compatibility is key.

Developers given the opportunity aren't going to optimize for all video cards individually the scope of that is just too big to be practical.

Mantle is an API not a specific implementation. There must be a mantle driver for your specific hardware but that is what AMD supplies so the developer only has to code against mantle and it will run on any hardware that has a mantle driver. And if mantle =XB1 API then they will code for it anyway -> no additional overhead, only for AMD in creating the drivers.

And it's only the engines that need to be optimized not all individual games. If you cover the most important engines, then a large fraction of all games are covered.

 

[sontin]

What? Mantle is a graphics API. This means you need to write your game for it. Mantle is not a middleware like PhysX which provides tools to create certain effects and compiles it to a target plattform automatically.

 

[3DVagabond]

Sonic will be a must have this holiday. :awe:



Still believing that any sane company is developing a multi port game with a low level API in mind. A few facts for you guys:
All Multi-port games in the next 6 to 12 months will have no Mantle API except BF4: No Call of Duty, no Watch Dogs, no AC, no Batman, no Fifa, no PES...
GCN market penetration is playing no role because in a open standard world people will not limiting themself to <20% of the market for investing the same amount of money to design a DX path to archive 100%.

But we all need our dream world, i guess.

Yes, BF4 will be the first. After that, we'll see. I can't see how it's going to work out like you think it will, though. This is just one big master plan that included the console wins. That would have been the make or break part of the plan, and AMD was successful there. Which game engine developer is going to want their competitors to be able to leverage this and not offer it themselves?

 

[beginner99]

What? Mantle is a graphics API. This means you need to write your game for it.

No, only the game engine. And obviously engines are re-used like frostbite 3 engine. BF4 developers code against frostbite 3 which has a render path for mantle. If you use frostbite 3 (or any other engine with a mantle render path) you notice nothing of it's existence besides potentially better performance on AMD GCN GPUs.

 

[3DVagabond]

No, only the game engine. And obviously engines are re-used like frostbite 3 engine. BF4 developers code against frostbite 3 which has a render path for mantle. If you use frostbite 3 (or any other engine with a mantle render path) you notice nothing of it's existence besides potentially better performance on AMD GCN GPUs.

I don't think it's quite that easy. The code for Mantle is going to have to be there. I don't think, for instance you could simply take any game that uses FB3 and simply a set of drivers is going to make it capable of running using the Mantle render path instead of Dx.

 

[sontin]

Yes, BF4 will be the first. After that, we'll see. I can't see how it's going to work out like you think it will, though. This is just one big master plan that included the console wins. That would have been the make or break part of the plan, and AMD was successful there. Which game engine developer is going to want their competitors to be able to leverage this and not offer it themselves?

In the next 12 months multi port games will come out on the current console generation, too. And right now there are no new consoles on the market. No publisher is targeting a <1% market with an 100% renderer investment.

No, only the game engine. And obviously engines are re-used like frostbite 3 engine. BF4 developers code against frostbite 3 which has a render path for mantle. If you use frostbite 3 (or any other engine with a mantle render path) you notice nothing of it's existence besides potentially better performance on AMD GCN GPUs.

Lol, just no. For that you need a wrapper and that would eliminate the positive effect of Mantle. :|

 

[beginner99]

Lol, just no. For that you need a wrapper and that would eliminate the positive effect of Mantle. :|

You obviously have no clue, NV shill. A game engine is to a certain extent a wrapper. Do you think all game developers code directly to D3D? they don't. They code against the game engine they use.

So the game engine wraps (the actual word is abstracts) whatever graphics APIs it is compatible with. And hence if the game engine uses mantle all games using that engine will profit from it. It's pretty easy to understand, at least when you are not trolling or pursuing some other ulterior motive.

The actual trick is the mantle driver and such a driver must exist for your hardware or it obviously won't work. And the driver contains the actual "to-the metal" coding.

And were exactly does this eliminate the positive effect of mantle? Care to explain?

EDIT:

Just look at this picture:

Graphics Application = Game Engine

Warning issued for personal attack.
-- stahlhart

 

[Haider]

Sonic will be a must have this holiday. :awe:



Still believing that any sane company is developing a multi port game with a low level API in mind. A few facts for you guys:
All Multi-port games in the next 6 to 12 months will have no Mantle API except BF4: No Call of Duty, no Watch Dogs, no AC, no Batman, no Fifa, no PES...
GCN market penetration is playing no role because in a open standard world people will not limiting themself to <20% of the market for investing the same amount of money to design a DX path to archive 100%.

But we all need our dream world, i guess.

It's not less than 20%, Mantle isn't completely low-level like assembler. It leverages an API/driver. I think that's why it's able to be cross platform and afford portability.

 

[sontin]

You obviously have no clue, NV shill. A game engine is to a certain extent a wrapper. Do you think all game developers code directly to D3D? they don't. They code against the game engine they use.

They code against the API. The Engine has no clue about the driver stack and the architecture. So they need to optimize the DX/OpenGL path per hand for every vendor.

So the game engine wraps (the actual word is abstracts) whatever graphics APIs it is compatible with. And hence if the game engine uses mantle all games using that engine will profit from it. It's pretty easy to understand, at least when you are not trolling or pursuing some other ulterior motive.

Yes, sure. I guess that's the reason why nearly every game comes at least with two renderer to the market... :\

And were exactly does this eliminate the positive effect of mantle? Care to explain?

If i dont care for low level optimizing and exclusives hardware features there is no reason to use Mantle. Which brings us back to the "no effect" situation.

 

[NTMBK]

Lol, just no. For that you need a wrapper and that would eliminate the positive effect of Mantle. :|

The engine is written against mantle, and the game is written against the engine. There's no extra wrapper involved.

Lots of other EA games are going to reuse the Frostbite 3 engine, and benefit from the Mantle optimizations.

 

[0___________0]

When they say open source they mean that you can use it on whatever platform you want. Meaning windows, linux, consoles, etc.

Who do you mean by "they"? AMD? AMD has stated it is not open source. You can't port the source code because it won't be available. I don't see anyone making the effort to create a Linux driver, and probably a necessary abstraction layer, from scratch. You're not going to create a polished product with the same performance. That's why I think AMD is only the one who can create a top notch solution and give momentum to Mantle on Linux.

 

[sontin]

The engine is written against mantle, and the game is written against the engine. There's no extra wrapper involved.

...
In a world with programmable shaders you dont write an engine against an API. Your engine has support for an API but the developer is responsible for the rest.

 

[smackababy]

Talking about something he wished he’d done differently during Rage’s development, Carmack said this:

“When we started on the game six years ago, I looked at the consoles and said ‘These are as good as the PC’, and our development strategy was to develop live on all the platforms. And now we’re looking at PCs that have ten times the horsepower of the consoles. I’m making a large change in my direction just saying ‘We should be building things efficiently on the PC and then deploying on the consoles.’ And we didn’t make that as crisp of a distinction as we could have.

“My development system now has twenty four threads and twenty four gigs of memory, and we can start putting on half a terrabyte of solid state drives, and these are the things that are gonna drive the development process on the PC. I’m actually as excited about how we’re developing tht titles in this coming generation as the graphic enhancements and things that I’m gonna make.

“…it is unhappily true that we have these consoles here running at sixty frames per second, and we could have these massively more powerful PC systems that struggle sometimes to hold the same framerate because of unnecessary overheads. If we were programming that hardware directly on the metal the same way we do consoles, it would be significantly more powerful.”


''The real reason to get excited about a PS4 is what Sony as a company does with the OS and system libraries as a platform, and what this enables 1st party studios to do, when they make PS4-only games. If PS4 has a real-time OS, with a libGCM style low level access to the GPU, then the PS4 1st party games will be years ahead of the PC simply because it opens up what is possible on the GPU. Note this won't happen right away on launch, but once developers tool up for the platform, this will be the case. As a PC guy who knows hardware to the metal, I spend most of my days in frustration knowing damn well what I could do with the hardware, but what I cannot do because Microsoft and IHVs wont provide low-level GPU access in PC APIs. One simple example, drawcalls on PC have easily 10x to 100x the overhead of a console with a libGCM style API.'' Timothy Lottes http://timothylottes.blogspot.fr


Thanks
Haider

I think that isn't so much of a "to the metal problem" as it is a DX stack problem. You can skip only one step (tessellation) in the DX stack IIRC. You have to go through a lot of others. If developers were given the chance to bypass (and write their own if they choose) any and all of these steps, a lot of custom implementations will spring up and be better (and worse). And if they allow a way to remove the abstraction for those that choose, it will give even more room for developers who want to go that route (such as Carmack and Lottes). However, you will notice, most developers won't bother, because DX is good enough and obsessing over graphics doesn't make your games any better.

 

[VulgarDisplay]

...
In a world with programmable shaders you dont write an engine against an API. Your engine has support for an API but the developer is responsible for the rest.

Which most developers are going to do regardless because of the consoles.

 

[blackened23]

I think that isn't so much of a "to the metal problem" as it is a DX stack problem. You can skip only one step (tessellation) in the DX stack IIRC. You have to go through a lot of others. If developers were given the chance to bypass (and write their own if they choose) any and all of these steps, a lot of custom implementations will spring up and be better (and worse). And if they allow a way to remove the abstraction for those that choose, it will give even more room for developers who want to go that route (such as Carmack and Lottes). However, you will notice, most developers won't bother, because DX is good enough and obsessing over graphics doesn't make your games any better.

Well, considering that PCs have GPUs 20-30 times faster than even the next gen consoles while only a small percentage of that 20-30 times faster can be leveraged, I think it is fair that many developers would like more control over the API situation. PC gamers expect more than console gamers do, on average, in terms of graphics. And that is what we get. But consider the cost: it is sad, that crysis 3 can chug along on a single GTX 780 maxed out. It is sad That Metro : LL can chug with pathetic frame rates when maxed out. Sure, you can lower 2-3 settings to rectify the framerate issue, but this should not be. The hardware is 20-30 times faster than consoles and is capable of delivering that same quality without the performance penalty, but DX has a ton of overhead. If a new API situation can be created to help facilitate that, i'm all for it.

I love consoles for gaming as well, but all of this "well it's good enough for me" graphics are better suited to consoles. PC gamers expect more and the fact is, we can only use a small percentage of the available power that we have - consider the 10 year old xbox 360. The current GPUs of that time were the ATI x1800 - can an x1800 even run Black Ops 2 at 720p at an acceptable frame rate? Hell no. Why? Because of DirectX. The xbox 360 can do more with less because it has an API similar to DX9 but it is tailored to the ATI GPU used within and allows direct hardware access.

The only sensible counter-argument to bare metal APIs is the fragmentation issue, but we will still have DirectX for common usage. "Good enough graphics" is not a proper argument. This is the PC gamer we're talking about, not the console gamer. And I say that as someone who appreciates and plays console games. We NEED a better situation for PC gaming than DirectX - hell, the transition from DX9 to DX11 took nearly 10 years. Ten years for us to get tessellation. That is pathetic. Maybe all of this talk of to the metal programming and developers that want it will force MS to look into the issue.

 

[sontin]

Which most developers are going to do regardless because of the consoles.

Which will be a high level API to port the game to different plattforms. :awe:

 

[smackababy]

Well, considering that PCs have GPUs 20-30 times faster than even the next gen consoles while only a small percentage of that 20-30 times faster can be leveraged, I think it is fair that many developers would like more control over the API situation. PC gamers expect more than console gamers do, on average, in terms of graphics. And that is what we get. But consider the cost: it is sad, that crysis 3 can chug along on a single GTX 780 maxed out. It is sad That Metro : LL can chug with pathetic frame rates when maxed out. Sure, you can lower 2-3 settings to rectify the framerate issue, but this should not be. The hardware is 20-30 times faster than consoles and is capable of delivering that same quality without the performance penalty, but DX has a ton of overhead. If a new API situation can be created to help facilitate that, i'm all for it.

You know damn well Metro: LL and Crysis 3 maxed out on PC isn't the same image quality as it is on Xbox 360.

I love consoles for gaming as well, but all of this "well it's good enough for me" graphics are better suited to consoles. PC gamers expect more and the fact is, we can only use a small percentage of the available power that we have - consider the 10 year old xbox 360. The current GPUs of that time were the ATI x1800 - can an x1800 even run Black Ops 2 at 720p at an acceptable frame rate? Hell no. Why? Because of DirectX. The xbox 360 can do more with less because it has an API similar to DX9 but it is tailored to the ATI GPU used within and allows direct hardware access.

Again, and I've asked this same question and been told a PC of 2005 running at the same settings is capable (just more expensive). Black Ops 2 is running at like 880 x 720 or something. It isn't running at 1080p. A think a x1800 could do that.

The only sensible counter-argument to bare metal APIs is the fragmentation issue, but we will still have DirectX for common usage. "Good enough graphics" is not a proper argument. This is the PC gamer we're talking about, not the console gamer. And I say that as someone who appreciates and plays console games. We NEED a better situation for PC gaming than DirectX - hell, the transition from DX9 to DX11 took nearly 10 years. Ten years for us to get tessellation. That is pathetic. Maybe all of this talk of to the metal programming and developers that want it will force MS to look into the issue.

Fragmentation is an issue and I never said "good enough graphics". I said obsessing over them. Crysis was a bad game, plain and simple. Crytek has even stated their games are 60% graphics, which is the reason they are bad. It doesn't matter how good your game looks if it is bad, which is the reason PC gaming is suffering. Getting an extra even 100% performance boost won't change the bad stories, dialog, design and gameplay choices that are being made already.

 

[Creig]

Guys it cant be done, thats what the abstraction layers are for. All hardware whould be have to be exactly the same to share any kind of low-level code.

Now Mantle works on PC and PC only, until information indicate otherwise, its PC only, its impossible to share low level code with xbox because of the ESRAM, its a critical GPU resource not present on pc.
Also no one will code for Mantle for consoles if its not inside the xbox/ps4 sdks and documentation.

http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn

Let’s be very clear here: AMD will not discuss the matter let alone confirm it, so this is speculation on our part. But it’s speculation that we believe is well grounded. Based on what we know thus far, we believe Mantle is the Xbox One’s low level API brought to the PC.

Low level api may seem to be a good thing, but think about Nvidia/Intel response to this, we really want a pc gaming world when Intel, Nvidia and AMD have their own low level apis?

Why would the introduction of Mantle cause Intel/Nvidia to bring out their own API's? The reason AMD can introduce their own new low-level API is because their GCN hardware is in all three next-gen consoles. Plus it's present in the PC market as well. Neither Intel nor Nvidia share that luxury.

As it stands now, devs are coding for PSGL (Playstation3), DX9 (Xbox 360) and DX11 (PC). But in the future, they will only have to code for Mantle (Playstation 4, XBox One & GCN PC) and DX11 (PC). Mantle actually makes it easier for devs since it reduces the number of APIs necessary to bring a game to all major platforms/GPUs. In addition, utilizing the Mantle API will result in higher performance over DX.

Mantle doesn't negatively affect Nvidia owners. It simply provides a performance boost for next gen console/AMD GCN owners.

 

[Creig]

All Multi-port games in the next 6 to 12 months will have no Mantle API except BF4: No Call of Duty, no Watch Dogs, no AC, no Batman, no Fifa, no PES...
GCN market penetration is playing no role because in a open standard world people will not limiting themself to <20% of the market for investing the same amount of money to design a DX path to archive 100%.

But we all need our dream world, i guess.

How about:

• Battlefield 4
• Command and Conquer
• Mirrors Edge
• Plants Vs Zombies: Garden Warfare
• Need For Speed: Rivals
• Dragon Age: Inquisition
• Star Wars: Battlefront
• Mass Effect (New Title in the Franchise)

All will using the Frostbite 3 engine which now includes Mantle.

 

[smackababy]

How about:

• Battlefield 4
• Command and Conquer
• Mirrors Edge
• Plants Vs Zombies: Garden Warfare
• Need For Speed: Rivals
• Dragon Age: Inquisition
• Star Wars: Battlefront
• Mass Effect (New Title in the Franchise)

All will using the Frostbite 3 engine which now includes Mantle.

BF4 requires Mantle to be patched in (so, most likely it won't use it very effectively).
The rest of the games are coming out exactly when? Battlefront was just announced after the acquisition of Lucas Arts by Disney. You think that is coming out before 2015?

Plants vs Zombies? Oh no! A game that looks like I could have played it in 2002 on Newgrounds as a flash game is going to use Mantle! I bet instead of 1000FPS, I'll get 1100!

Dragon Age? Better focus a lot more on substance than looks, otherwise it will be terrible.

Mass Effect? Better focus a lot more on the ending than looks, otherwise internet outrage.

 

[Creig]

Sontin claimed only 1 in the next 6-12 months. I showed there to be more. If you want to argue about whether or not a particular title will be popular or not, I suggest you start a thread in the PC Gaming forum.