The AMD Mantle Thread

[SiliconWars]

http://www.gamesindustry.biz/articl...hooter-mold-how-oxide-plans-to-revive-the-rts

Baker said that supporting Mantle was a "straightforward" adaptation that only took a couple of months on an alpha API

So Mantle - or something close - has been available to other devs for a while now.

You'll see people misrepresenting that AMD's pushing this technology, but the reality is that a lot of people - including myself and other graphics architects - have been asking for this type of thing for a long time," said Baker.

He wanted it.

From our perspective, when AMD came around and said, 'we're actually going to do this,' we were very interested to try it.

AMD offered it.

Baker believes Mantle-enabled games will probably have a seperate executables from their Direct X versions, for efficiency. He said that players could probably expect an option to load the Mantle-enabled version, but otherwise they should not be affected by Mantle's implementation.

This seems like what they are all saying. Nothing to fear from it even if you don't currently have a GCN card. I mean it's not suddenly going to be made obsolete overnight, and if Mantle is a success then there's no good reason not to go with a GCN card for your next one.

Pretty excited about this now, it's gaining traction as I thought it would. I hope these weren't the only devs AMD was supposed to announce at APU13 next week and there are more, but even if these were them it's pretty good anyway.

 

[krumme]

I don't doubt that Crytek will support Mantle if it brings significant improvements, but it's possible that right now CIG just has a source license and implemented Mantle themselves with AMD's help. SC is looking to be a huge game, good to see Roberts try and take advantage of everything he can.

Their own reason for using mantle in sc is to use more ressources for the artistic side. I guess the major reason in general for using mantle is to prioritize money to the more content side. Another way of saying it bluntly: just to cut cost. Thats it. People in business normally understand that argument before anything else.
For the sc it looks like there is a certain budget and with mantle the can reach more.
The weird thing is as i understand it sc is only pc right???

If thats the case mantle is really really strong.

 

[Face2Face]

Mantle performance will dictate the next card I buy. SC is going to be an awesome game that will destroy most high-end PC's.

 

[frozentundra123456]

Their own reason for using mantle in sc is to use more ressources for the artistic side. I guess the major reason in general for using mantle is to prioritize money to the more content side. Another way of saying it bluntly: just to cut cost. Thats it. People in business normally understand that argument before anything else.
For the sc it looks like there is a certain budget and with mantle the can reach more.
The weird thing is as i understand it sc is only pc right???

If thats the case mantle is really really strong.

How does it cut costs to program for 2 APIs?

You still have to program for DX, and I dont see how mantle could make than any easier(cheaper). So no matter what mantle costs to implement, that cost will be added on to the cost for DX, unless the developer cuts corners and does a crappy job on DX.

 

[ShintaiDK]

Mantle performance will dictate the next card I buy. SC is going to be an awesome game that will destroy most high-end PC's.

Current GPUs are obsoleted by 20nm GPUs before Star Citizen even comes out. If it actually gets released.

 

[chimaxi83]

Current GPUs are obsoleted by 20nm GPUs before Star Citizen even comes out. If it actually gets released.

That's got nothing to do with Mantle. I'm fairly certain 20nm AMD cards will support it.

 

[smackababy]

How does it cut costs to program for 2 APIs?

You still have to program for DX, and I dont see how mantle could make than any easier(cheaper). So no matter what mantle costs to implement, that cost will be added on to the cost for DX, unless the developer cuts corners and does a crappy job on DX.

I think this is the biggest fear for the "non-believers". It is going to cost something to include an extra executable for Mantle, and how much will that bloat the already incredibly bloated development costs of games, or how much will it remove from the DX build.

It won't be free by any means. It will cost time to train developers and time for additional development.

 

[SiliconWars]

I think this is the biggest fear for the "non-believers". It is going to cost something to include an extra executable for Mantle, and how much will that bloat the already incredibly bloated development costs of games, or how much will it remove from the DX build.

It won't be free by any means. It will cost time to train developers and time for additional development.

The newly-announced developers have come out and stated explicitly that the extra Mantle development was not an issue. In fact :- http://www.oxidegames.com/mantle

Q. Why are you supporting Mantle?A. Mantle has two big things going for it. One, adding Mantle support doesn’t require us to write a whole new engine – we’re able to build Nitrous the way we want to, and it simply uses Mantle with compatible hardware. Because of that, Mantle support doesn’t cost us an absurd amount of time or money. Integrating Mantle into Nitrous isn’t trivial, but we think the benefits are easily worth the development effort on our part

That can't be any clearer than what it is.

Of course there are costs involved, but so far from what we've seen they are not big costs, and the expertise does not appear to be limited to gaming gods like Repi at DICE either.

 

[ruhtraeel]

IMO Mantle isn't necessary. Games run fine nowadays on DirectX and OpenGL (for anything that DirectX doesn't run on).

As a 7970 owner, I'm not excited at all.

The thing that pisses me off most is when AMD criticizes Nvidia's PhysX and CUDA for being proprietary, and then pull this proprietary bullcrap.

I'm sort of glad that my dad's company uses only Nvidia cards (and they don't even look at the cost of the card when they decide which GPUs to buy)

 

[bystander36]

The newly-announced developers have come out and stated explicitly that the extra Mantle development was not an issue. In fact :- http://www.oxidegames.com/mantle

That can't be any clearer than what it is.

Of course there are costs involved, but so far from what we've seen they are not big costs, and the expertise does not appear to be limited to gaming gods like Repi at DICE either.

The concerns are still concerns. They just said it won't be absurdly more expensive. That doesn't mean there won't be all the complications that are feared, such as poorer patch fixes.

I just hope we don't go back to the quality of things when OpenGL and DirectX were both supported on many games.

 

[frozentundra123456]

The newly-announced developers have come out and stated explicitly that the extra Mantle development was not an issue. In fact :- http://www.oxidegames.com/mantle

That can't be any clearer than what it is.

Of course there are costs involved, but so far from what we've seen they are not big costs, and the expertise does not appear to be limited to gaming gods like Repi at DICE either.

So any cost less than "absurd" is not big?

 

[VulgarDisplay]

So any cost less than "absurd" is not big?

Your concern for game devs financial situation is touching.

Is it free to add physx support to games? Physx gets flak because its proprietary and because of that is relegated to being performance killing eye candy with no effect on gameplay. Nvidia still invests money into it to add value to their GPUs. Why is it such a problem now that amd has their own value adding idea for their own products?

I will gladly take extra performance and increased quality over blowing newspapers and slime any day.

 

[Saylick]

So any cost less than "absurd" is not big?

You bring up a good point, but also note the second half of the statement, i.e. "Integrating Mantle into Nitrous isn’t trivial, but we think the benefits are easily worth the development effort on our part."

In other words, from a cost-benefit perspective, they believe implementing Mantle is worth it.

If you could invest $1 million dollars for a guaranteed 1.5x return rate, would you do it? Finding $1 million dollars to spare isn't easy, but that doesn't mean making the investment isn't worth it.

 

[SiliconWars]

I have no idea how you quantify these things in your own minds, I simply linked what actual devs that are going to be using Mantle said.

A new startup company says the cost wasn't absurd, and an easy choice considering the benefits. It's going to be used in a Kickstarter project. If it's so easy for minor studios what do you think it'll be like for the big guys?

I'm trying to do you a favour here. Just forget about the ease of integration and cost aspect because it's clearly not an issue. There may be issues left with Mantle but this isn't one of them.

 

[AnandThenMan]

IMO Mantle isn't necessary.

Game developers don't agree, and is the major reason Mantle was created in the first place.

 

[SiliconWars]

Game developers don't agree, and is the major reason Mantle was created in the first place.

Never said better than here...

"You'll see people misrepresenting that AMD's pushing this technology, but the reality is that a lot of people - including myself and other graphics architects - have been asking for this type of thing for a long time," said Baker. "I call it a contract of trust. The problem is driver models and APIs are built to protect against any random scenario which might happen. That costs you a lot of performance. Our game engine is already carefully engineered not to do certain bad things. From our perspective, when AMD came around and said, 'we're actually going to do this,' we were very interested to try it."

I guess it's going to take another 100 postings of this and similar comments from other industry veterans until it's accepted as fact.

 

[frozentundra123456]

I have no idea how you quantify these things in your own minds, I simply linked what actual devs that are going to be using Mantle said.

A new startup company says the cost wasn't absurd, and an easy choice considering the benefits. It's going to be used in a Kickstarter project. If it's so easy for minor studios what do you think it'll be like for the big guys?

I'm trying to do you a favour here. Just forget about the ease of integration and cost aspect because it's clearly not an issue. There may be issues left with Mantle but this isn't one of them.

Well, I have a lot more to be concerned about in my life than a graphics api, so you dont have to do me any favors. Beyond that, I prefer to make my own decisions, waiting till there is hard data, rather than continually tout mantle based on statements made by the most biased sources possible, those who are actually using it, and who undoubtedly are subsidized in some way by AMD. We will know eventually what the performance increase is, but we may never know the real cost in development costs, time spent, and how much it hinders DX development for those who do not have an amd card.

And by the way, I have an AMD card, so dont go saying I hate AMD.

 

[SiliconWars]

Well, I have a lot more to be concerned about in my life than a graphics api, so you dont have to do me any favors. Beyond that, I prefer to make my own decisions, waiting till there is hard data, rather than continually tout mantle based on statements made by the most biased sources possible, those who are actually using it. It is like going to the chevrolet dealer and asking if a chevy is better than a ford.

I'm really happy for you to make your own decision on Mantle. As you clearly have no discussion to add you can keep it to yourself. This is an AMD thread discussing AMD technology and if you don't like it there are plenty of threads that you can discuss technology where your decision hasn't already been made.

 

[frozentundra123456]

I'm really happy for you to make your own decision on Mantle. As you clearly have no discussion to add you can keep it to yourself. This is an AMD thread discussing AMD technology and if you don't like it there are plenty of threads that you can discuss technology where your decision hasn't already been made.

I have not made a decision yet, because there is no concrete information to make a decision on. That is the whole point.

 

[SiliconWars]

So all you have to add to the thread is that you haven't decided? That doesn't take 50+ posts frozentundra.

Wait until you have decided then come back and let us know. The rest of us wouldn't mind discussing Mantle and if we get a bit carried away with it then you know...you can always just keep on waiting because nobody is forcing you to read it either. This is an AMD thread, respect it.

 

[psoomah]

When developers signing on for Mantle there will be a lot of the usual Marketing BS.

Taking that marketing bs talk here and using it as strawmen is thread crapping imho. Nothing good comes from using Marketing bs as a foundation of a discussion.

The same goes for the 10x faster than DirectX and nonsense like that, taken out of context. Nothing but Strawmen.

We have had several threads here with the new 290 cards that went into smoke.
If you dont like Mantle fine, present the arguments instead of strawmen.

For my perspective it would be nice if we could get some more info about the content- the games, but i dont know where to find more. Anyone?

Information contained in the slide presentations at GPU13 is very carefully worded and vetted. There are legal considerations with such official presentation. Unless proven otherwise the information in those slides can be considered valid and meaning exactly what they say.

 

[SiliconWars]

You're just going to have to wait a few more days lol.

As ever, anyone expecting to see raw numbers is going to be disappointed. I think AMD will use this to showcase the actual visual differences, ie best case loaded Mantle with zillions of objects that chokes Nvidia cards but runs smooth as butter on GCN. This is AMD's general tactic and I don't see them deviating from it.

We'll get some kind of BF4 demo as well but with the same rules. I expect extra eye-candy to be the main showpiece far more than fps totals. Maybe they'll tease us with comparative fps with some obscure game that never gets benchmarked, but there's zero chance of them giving the whole game away until the BF4 patch launches.

 

[psoomah]

IIn many ways I am surprised developers are signing on, I suspect AMD is paying them to add support, or the typical thing is to give them developers to add support into their games alongside their other devs.

DICE flat out said an AMD team was implementing Mantle into Frostbite with EA/DICE 'assistance'. While AMD said Frostbite was it's lead engine they also said they were working with several developers and it's reasonable there are other AMD teams working in collaboration with the Frostbite team implementing Mantle on other engines and when the Frostbite engine is completed, that team spread out to other engines with a now fully fleshed out out rendering path in hand and the experience gained in implementing it into Frostbite and so on until every developer that want's Mantle has it.

It's hard to imagine a better return on investment for AMD than getting Mantle into every high end gaming engine out there.

 

[psoomah]

I'd be surprised if there's much re-usable code, these are different APIs with different languages. But for sure a lot of the tricks used to optimize the performance, the actual concepts themselves will be transferable, so some of the work done to optimize consoles and figure out how to best use the hardware will transfer but it'll likely need the code refactored somewhat.

AMD of necessity wrote the initial abstraction console APIs, it is their architecture after all and no one else would have the intimate knowledge of the architecture to enable them to write that code. Since the consoles architectures are nearly identical, AMD"s low level 'metal' APIs would be nearly identical for both machines. A variation of those APIs would then form the foundation of Mantle for AMD"s GCN based APU and CPU architectures.

 

[krumme]

How does it cut costs to program for 2 APIs?

You still have to program for DX, and I dont see how mantle could make than any easier(cheaper). So no matter what mantle costs to implement, that cost will be added on to the cost for DX, unless the developer cuts corners and does a crappy job on DX.

Good question. A fact is, they explicitely said they could cut cost. They could say anything else but thats what was drawn into attention. So how?

Here is my take why:
Programming for mantle and get okey performance is extremely cheap. Is far far more simple than dx. Especially for the heavy classes/layers on the lower level on the gfx side.
If you dont demand the last level of performance on dx you save a ton of proramming time.
The total calculation of the above is saved cost in this case for the content side.

Add to the above that if nv want better dx perf they can just pay for it. And to the degree they want. Its their business.

But it simply shows the benefit of mantle to dx.
Its both cheaper ro program for.
You get better performance for less.
Porting is a walk in the park when the job is done.

Go imagine you were a developer. I would take that route everyday. Who wouldnt want mantle instead of dx. Dx is the old world focusing on hardware and lots of programming ressources for the low level to get good performance. Be it pc or consoles. For the gaming market that is bad use of ressources. Why should they care.

The story that dx was one api is off. Its a complicated, heavy and fragmented huge pile of rocks remisent of the ibm pc era.

Time goes on. The situation with similar pc and consoles changed it.
Its ready to die of age.