The AMD Mantle Thread

[Silverforce11]

Some of you guys are dreaming if you think Mantle is indifferent to the architecture, it's simply not true and many presentations already pointed it out, re: GCN specific optimization AND FEATURES. Mantle unlocks additional rendering features unique to GCN above and beyond DX.

"Andersson said that most Mantle functionality can work on most modern GPUs out today." Note the "MOST" and not "ALL".

It's AMD being silly, as if Intel or NV would have AMD determine their future GPU architectures in terms of features.

 

[3DVagabond]

Some of you guys are dreaming if you think Mantle is indifferent to the architecture, it's simply not true and many presentations already pointed it out, re: GCN specific optimization AND FEATURES. Mantle unlocks additional rendering features unique to GCN above and beyond DX.

"Andersson said that most Mantle functionality can work on most modern GPUs out today." Note the "MOST" and not "ALL".

It's AMD being silly, as if Intel or NV would have AMD determine their future GPU architectures in terms of features.

Of course they wouldn't. Mantle would have to become an open standard before any of the other vendors would adopt it. It could get interesting though if a majority of the AAA titles used it and it was appreciably faster.

 

[blackened23]

I can't possibly see why AMD would want Mantle as an open technology. How does Mantle being available to other vendors help sell AMD cards? That's the goal here, right? For AMD to sell more of their APUs and GPUs? It's a pretty ridiculous notion to suggest that anyone can use it if you ask me - let's say that nvidia theoretically uses Mantle, but Mantle titles somehow run better on NV cards a year from now. So how does that help AMD add value to their cards? From a business and financial perspective, a Mantle available to everyone would just lead to AMD selling fewer GPUs and APUs. AMD needs to stop trying to be the "nice" guy. The nice guys never win in the business world, because rest assured the competition doesn't share that mindset.

I mean, let's look at nvidia value adds. Aside from the user experience, they have shadowplay, physx, TXAA, among a variety of other NV specific features. Does NV rush out to intel and AMD and say "hey, we want you guys to use this!"? No of course not. These features help sell their cards. Mantle will help sell AMD cards.

IMO, unless AMD is stupid, Mantle will be AMD only. All they need to do is get it baked into several next generation engines and that will make it available to the vast majority of PC Games, since most AAA games use 4-5 game engines at the most. If Mantle somehow made it into UE4, well bam, that would make Mantle available to a TON of games. It doesn't need to be "open" and available to everyone.

The ultimate goal of Mantle is to serve as a value add for gamers to go out and buy AMD GCN cards and GCN APUs. Period. Making it available to everyone runs quite contrary to that.

 

[Silverforce11]

The ultimate goal of Mantle is to serve as a value add for gamers to go out and buy AMD GCN cards and GCN APUs. Period. Making it available to everyone runs quite contrary to that.

Spot on. That, we can all hopefully agree on.

Mantle is unique to GCN architecture, it doesn't even support older gen Radeons or APUs. If that isn't an indication of hardware specificity, not sure what other evidence some of you need. Again, Intel and NV is NOT going to support Mantle. Not just because the architectures differ, but because they would never devote resources to help their competitor.

 

[3DVagabond]

I can't possibly see why AMD would want Mantle as an open technology. How does Mantle being available to other vendors help sell AMD cards? That's the goal here, right? For AMD to sell more of their APUs and GPUs? It's a pretty ridiculous notion to suggest that anyone can use it if you ask me - let's say that nvidia theoretically uses Mantle, but Mantle titles somehow run better on NV cards a year from now. So how does that help AMD add value to their cards? From a business and financial perspective, a Mantle available to everyone would just lead to AMD selling fewer GPUs and APUs. AMD needs to stop trying to be the "nice" guy. The nice guys never win in the business world, because rest assured the competition doesn't share that mindset.

I mean, let's look at nvidia value adds. Aside from the user experience, they have shadowplay, physx, TXAA, among a variety of other NV specific features. Does NV rush out to intel and AMD and say "hey, we want you guys to use this!"? No of course not. These features help sell their cards. Mantle will help sell AMD cards.

IMO, unless AMD is stupid, Mantle will be AMD only. All they need to do is get it baked into several next generation engines and that will make it available to the vast majority of PC Games, since most AAA games use 4-5 game engines at the most. If Mantle somehow made it into UE4, well bam, that would make Mantle available to a TON of games. It doesn't need to be "open" and available to everyone.

The ultimate goal of Mantle is to serve as a value add for gamers to go out and buy AMD GCN cards and GCN APUs. Period. Making it available to everyone runs quite contrary to that.

What do we care? It's pretty obvious that DX sucks and the devs want to kick it to the curb. I hope they're successful and it breaks M$'s choke hold on gaming.

AMD can't win here. When Mantle was first announced people were crying that it would fragment the market. It'll end up like glide and we'll have all kinds of bugs. It's a monopoly that will cut out the other hardware manufacturers. Devs are going to have to develop for more API's and it'll raise costs. All the other hardware manufacturers will have to develop their own API's to compete and gaming will revert to the dark ages. On and on... Now they're saying that actually it's not hardware specific and it only requires a compatible feature set that could be implemented by other hardware vendors and people are still complaining.

Spot on. That, we can all hopefully agree on.

Mantle is unique to GCN architecture, it doesn't even support older gen Radeons or APUs. If that isn't an indication of hardware specificity, not sure what other evidence some of you need. Again, Intel and NV is NOT going to support Mantle. Not just because the architectures differ, but because they would never devote resources to help their competitor.

You are saying it's unique to GCN architecture. They are saying it's not. Why do you know better than them? It's early days yet. Wait until someone who knows says it won't work on anything except GCN. It actually makes sense that it isn't solely reliant on GCN. That would tie AMD's hands and lock them into GCN preventing them from advancing, or ruining backward compatibility with any new arch they developed. They might just be telling the truth. It makes sense that they'd be able to develop new architectures that would maintain Mantle compatibility.

 

[Obsoleet]

It's AMD being silly, as if Intel or NV would have AMD determine their future GPU architectures in terms of features.

Intel should definitely work with AMD and get in on Mantle.
In fact, now is the time for Intel and AMD to work together to defeat common enemies.

-Intel to face ARM with their low power hardware, which will benefit AMD if x86 gains more traction there as they're working towards the same goal.
-AMD who already defeated everyone in the next-gen consoles handily, and this will only further x86's traction there.
-And they both have common interests and common enemies.

What would be ideal: AMD should work with Intel and get them in on Mantle. Then cooperate together for an open standard Gsync clone- use Intel's might to get it in every LCD display sold on the market. Destroying NV's initiative. Maybe even making this 'open standard' somehow difficult or impossible for Nvidia to support through legal barriers. Don't know, but it would be a good idea to do so, as it won't be long and OpenSync will be all over and GSync will be another Physx.
Lower performance parts benefit more from a Gsync type solution than high end anyway. It's ideal for APUs from Intel and AMD.

I personally like both Intel and AMD, and am pretty negative on NV and most competitors to the esteemed Club x86, sorry Nvidia.
I see a lot of market power, and joint interests between AMD and Intel now more than ever. I'd like to see the two create more backroom deals and put some hurt on their respective foes. Both are going to benefit in the end as long as x86 prevails.

 

[Silverforce11]

You are saying it's unique to GCN architecture. They are saying it's not.

They are the ones saying its for GCN. They are also the ones saying it unlocks GCN exclusive features. They are saying its NOT compatible with older radeons and APUs.

"Andersson said that most Mantle functionality can work on most modern GPUs out today."

Most != All. If you don't have ALL, you don't have Mantle support. re: DX10 or DX11

 

[krumme]

So, I will take this as a no then? There is absolutely no real demos or code examples showing the similarities of the APIs nor are there any demos or code examples showing Mantle actually performing better even in synthetic benchmarks?

Obviously, I'm the one threadcapping because I went on a broken English rant about Nvidia being bad and no real substance other than "Mantle is great because someone else said so."


I am done with this thread until the SDK comes out. The Mantle fanboys are rather silly.

We have known since page 1 the SDK will be final h2 2014. We have known since page 1 you will never see the ms drivers or the sony drivers. Ever.
So where are you going with your question of similarity?

Go see the q&a. No matter the coding differences. By looking how different mantle is to dx for working complexity it surely is fast to work with. Its extremely easy to implement. And thats even without a console version.

 

[beginner99]

Hint to those thinking iGPUs are useless: what's to stop a programmer from using the CPU, iGPU, AND dGPU all at the same time?

Yes, at that point all these iGPUs finally become usable and are not dead-silicon anymore.

They would become reliant on AMD to play nice and keep Mantle working properly on their cards going forward.

Unless it becomes a truly open platform, out of AMD's control, I do not see Nvidia willingly getting involved.

If you replace AMD with Microsoft and Mantle with DX, you have the situation as it is now.

Some of you guys are dreaming if you think Mantle is indifferent to the architecture, it's simply not true and many presentations already pointed it out, re: GCN specific optimization AND FEATURES. Mantle unlocks additional rendering features unique to GCN above and beyond DX.

Your obviously clueless about programming. Else you would easily see, how this can be hardware agnostic. The only requirement is that the hardware supports the minimal required features. Mantle is an API. You can program against it and the important part you can create your own implementation (speak mantel driver) of it for your GPU uArch.

It's no different than DX. DX also requires a certain minimum feature set so that you can name your card DX-compatible.



Personal Attacks are against the forum rules.

-Rvenger

 

[krumme]

I don't have a problem with the dev's taking control, but don't act like MS hasn't tried to improve gaming. Even if they do it so they can sell their latest OS. If they weren't trying, OpenGL would rule the roost.

Amd have competences in house that ms dont have to make the synergy between hardware, driver and api faster and easier to work with.

 

[Silverforce11]

Your obviously clueless about programming. Else you would easily see, how this can be hardware agnostic. The only requirement is that the hardware supports the minimal required features. Mantle is an API. You can program against it and the important part you can create your own implementation (speak mantel driver) of it for your GPU uArch.

It's no different than DX. DX also requires a certain minimum feature set so that you can name your card DX-compatible.

You just contradicted your insult in one post.

Fact is, the ONLY current architecture that has full Mantle compatibility is GCN. The rest, while they could support "most Mantle features", it won't support ALL. Therein lies the small detail you choose to ignore.

For your DX comparison, its like getting a DX10.1 compatible card to run DX11 only features, not happening.

 

[bystander36]

Amd have competences in house that ms dont have to make the synergy between hardware, driver and api faster and easier to work with. The leverage is ofcource they dont have to design for a lot of different hardware.

In what way does this mean MS isn't trying to improve gaming?

 

[3DVagabond]

They are the ones saying its for GCN. They are also the ones saying it unlocks GCN exclusive features. They are saying its NOT compatible with older radeons and APUs.

"Andersson said that most Mantle functionality can work on most modern GPUs out today."

Most != All. If you don't have ALL, you don't have Mantle support. re: DX10 or DX11

So because it's not compatible with older Radeons and APU's that means it's not compatible with anything else? Not necessarily.

Don't get me wrong. I'm not saying it is or it isn't. What I'm saying is that I don't know, and you don't either. You talk though like you have more knowledge of it than Dice does. See my point?

 

[Skurge]

You just contradicted your insult in one post.

Fact is, the ONLY current architecture that has full Mantle compatibility is GCN. The rest, while they could support "most Mantle features", it won't support ALL. Therein lies the small detail you choose to ignore.

For your DX comparison, its like getting a DX10.1 compatible card to run DX11 only features, not happening.

It's not compatible with older cards not because they are not GCN, but because they do not support the minimum featureset.

It's been mentioned before. I don't know if you missed it. Minimum featureset. If you support it and you can write a driver it should work. Why wouldn't it? It still has a hardware abstraction layer. It's just much thinner.

Even ignoring all that. Why would Johan build an API that is not only AMD only, but GCN only. If I had to take a guess. The reason so many devs are throwing their weight behind it so soon is because it will be able to work on other vendors. If you watch APU13 interview. You can just see how the devs hate working with DX. They are trying to push mantle to either get rid of having to work with DX or get MS to completely revamp it.

 

[3DVagabond]

It's not compatible with older cards not because they are not GCN, but because they do not support the minimum featureset.

It's been mentioned before. I don't know if you missed it. Minimum featureset. If you support it and you can write a driver it should work. Why wouldn't it? It still has a hardware abstraction layer. It's just much thinner.

Even ignoring all that. Why would Johan build an API that is not only AMD only, but GCN only. If I had to take a guess. The reason so many devs are throwing their weight behind it so soon is because it will be able to work on other vendors. If you watch APU13 interview. You can just see how the devs hate working with DX. They are trying to push mantle to either get rid of having to work with DX or get MS to completely revamp it.

Johan? What makes you think Johan wrote it?

 

[Skurge]

Johan? What makes you think Johan wrote it?

He built it in collaboration with AMD.

 

[3DVagabond]

He built it in collaboration with AMD.

I think it might be the other way around.

 

[Silverforce11]

It's not compatible with older cards not because they are not GCN, but because they do not support the minimum featureset.

It's been mentioned before. I don't know if you missed it. Minimum featureset. If you support it and you can write a driver it should work. Why wouldn't it? It still has a hardware abstraction layer. It's just much thinner.

I don't know where the confusion is.. obviously to support an API you absolutely need a minimum featureset. Currently only AMD GCN (as indicated by all the presentations so far with their language usage) meets that minimum requirement. Neither AMD past radeons nor NV's architecture meets the requirement, else AMD and other devs would not make their statements regarding "most features". Note the language used. If you don't meet the featureset requirements (NV may or may not meet most of the features, still not ALL), you aint got compatibility with that API.

@Vagabond, I dont know (and nobody here either), I am merely interpreting the presentations and information that has been released. It clearly states Mantle has GCN exclusive features. How much clearer does it need to be?

 

[Minkoff]

Well this is another thing they said for the support...

AMD Mantle: Empowering 3D Graphics Innovation at APU13
http://www.youtube.com/watch?v=r7Q_nAnnrnI#t=189

2:00 min. - "Floating abstraction layer" and not stuck to GCN and possible compatibility with competitor hardware
As I understand it, Mantle is not a low level API nor a High level one (that's why I use "floating" as a word). So it's up to AMD to decide (and probably some devs) what to do with it.

But...there is another thing that I find very interesting at 37:15min. regarding the assymetric GPU configuation (like APU and high end GPU or different GPU crossfire). It reminds me of times long gone, when ATI said, that it's possible to use one GPU for physics or AI etc...

 

[csbin]

http://www.legitreviews.com/rebelli...g-amd-mantle-games_129404#DrUmS8kx02mCvxZd.99​

Rebellion Announces They Will Be Using AMD Mantle In their Games

Rebellion announced today that they will be using AMD’s Mantle graphics API to get the very best hardware performance with their in-house game engine, Asura. The Asura game engine powers all of the Rebellion game titles, from advanced PC titles to mobile and next-generation console games. The first Rebellion Entertainment game title to use AMD Mantle technology will be Sniper Elite V3!
Quotes from Rebellion’s CTO and Co-Founder, Chris Kingsley:
“It supports more platforms than almost any other engine we can think of and has important cutting edge features, including: tessellation, DirectX 11 Compute Shaders, AMD Eyefinity technology, multi-GPU support and more.”
“As one of Europe’s leading independent games studios we’re dedicated to constantly pushing our technology as far as we can, and we’re excited about the possibilities that Mantle brings to PC gaming and the industry as a whole.”
“We believe that supporting Mantle will enable us to stay on the bleeding edge of PC gaming and ensure that we don’t leave any performance on the table when it comes to offering gamers amazing experiences.”

 

[blastingcap]

Yeah someone already posted that Rebellion thing. Not sure if they are major (apparently Sniper Elite v2 is their claim to fame, and I hear that game was repetitive and sucky), though every bit counts.

 

[Noctifer616]

@Vagabond, I dont know (and nobody here either), I am merely interpreting the presentations and information that has been released. It clearly states Mantle has GCN exclusive features. How much clearer does it need to be?

The presentation could also be interpreted in the way that right now Mantle is GCN only and they want to finish it before they expand their support to other vendors or platforms.

It's in alpha right now with beta coming in a few months.

It's also possible that they can't add support for other vendors without the help from Nvidia and Intel.

It's possible that Kepler and GCN have all the feature sets required for Mantle.

 

[MeldarthX]

They are the ones saying its for GCN. They are also the ones saying it unlocks GCN exclusive features. They are saying its NOT compatible with older radeons and APUs.

"Andersson said that most Mantle functionality can work on most modern GPUs out today."

Most != All. If you don't have ALL, you don't have Mantle support. re: DX10 or DX11

APU13 they stated its not tied to GCN; while in its early stage yes it is tied to gcn its going to be opened up to all gpus; which also means they will put in for their older tech and tech that comes after gcn.

This argueing it has to be on gcn is silly when they've said mantle won't be. Yes in the early stages it is because its just starting out; they want to develop it into a fully fledge open standard.

Look at history with AMD; when have they ever locked out ? Name one time? I can name plenty for Intel and Nvidia........Name one for AMD? AMD standardized GDDR - did they lock out nvidia from using it? Nope; They are leading the development GDDR 6.

Mantle will be the driving force for systems for next 5 years; it will be open. Will Intel get on board; *most likely though they'll come kicking and screaming like they did with AMD64; but they will come.* Will Nvidia? Now that I don't know because Nvidia will do things like cut off their nose to spite their face.

G-sync is not special; that can be all done actually threw displayport. Nvidia does not do anything that won't allow them to try and lock it completely to them so they can charge more. They are not consumer friendly.

Out of the three - whether it was to survive or not in the end it honestly doesn't matter. AMD has been over the years the most consumer friendly company out of the 3.

TressFX; locked? No........Trueaudio locked? No; TressFX 2.0 locked? No; Mantle......locked NO.. No matter how many times you try to say it is; it won't be.

If AMD does 180 on this; then I'll eat my hat......but I can say without a doubt they won't. Specially after APU 13 completely destroys the arguements its locked and won't be open.

 

[SiliconWars]

Yeah someone already posted that Rebellion thing. Not sure if they are major (apparently Sniper Elite v2 is their claim to fame, and I hear that game was repetitive and sucky), though every bit counts.

They made AvP as well. Not huge by any means but another win.

I can't possibly see why AMD would want Mantle as an open technology. How does Mantle being available to other vendors help sell AMD cards? That's the goal here, right? For AMD to sell more of their APUs and GPUs? It's a pretty ridiculous notion to suggest that anyone can use it if you ask me - let's say that nvidia theoretically uses Mantle, but Mantle titles somehow run better on NV cards a year from now. So how does that help AMD add value to their cards? From a business and financial perspective, a Mantle available to everyone would just lead to AMD selling fewer GPUs and APUs. AMD needs to stop trying to be the "nice" guy. The nice guys never win in the business world, because rest assured the competition doesn't share that mindset.

I mean, let's look at nvidia value adds. Aside from the user experience, they have shadowplay, physx, TXAA, among a variety of other NV specific features. Does NV rush out to intel and AMD and say "hey, we want you guys to use this!"? No of course not. These features help sell their cards. Mantle will help sell AMD cards.

IMO, unless AMD is stupid, Mantle will be AMD only. All they need to do is get it baked into several next generation engines and that will make it available to the vast majority of PC Games, since most AAA games use 4-5 game engines at the most. If Mantle somehow made it into UE4, well bam, that would make Mantle available to a TON of games. It doesn't need to be "open" and available to everyone.

The ultimate goal of Mantle is to serve as a value add for gamers to go out and buy AMD GCN cards and GCN APUs. Period. Making it available to everyone runs quite contrary to that.

If AMD can convince the devs to adopt Mantle without the support of anyone else, sure. If not, they might need to open it to gain mass adoption. Now we'd be in the situation where AMD has CPU parity with Intel but a stronger APU that can be combined with a discrete card.

Basically, AMD would overtake Intel in gaming overnight. Nvidia has no proper CPU that they can combine their own graphics with, so AMD will still have that advantage over them.

Either way AMD wins out of this, it's sheer genius. Personally I would rather they finished Nvidia first but with both Nvidia and Intel absolutely motivated to not allow that to happen I doubt it will.

The second, more likely option is that one of the big mobile players decides they need this advantage. This is the sort of thing that would save Intels bacon in mobile, but they are far too arrogant to even consider it. Others must be though.

 

[Erenhardt]

Mantle is not only providing low level programming performance benefits, but also helps speed up developing/optimizing/debug process, which is (probably) brand agnostic.

The presentation could also be interpreted in the way that right now Mantle is GCN only and they want to finish it before they expand their support to other vendors or platforms.

It's in alpha right now with beta coming in a few months.

It's also possible that they can't add support for other vendors without the help from Nvidia and Intel.

It's possible that Kepler and GCN have all the feature sets required for Mantle.

It is almost impossible to add other vendor support without detailed information about architecture and drivers they use.
It is almost like with open gpu drivers in linux. Without a help from hardware engineers, open drivers can't get anywhere close to official drivers.

Intel should definitely work with AMD and get in on Mantle.
In fact, now is the time for Intel and AMD to work together to defeat common enemies.

-Intel to face ARM with their low power hardware, which will benefit AMD if x86 gains more traction there as they're working towards the same goal.
-AMD who already defeated everyone in the next-gen consoles handily, and this will only further x86's traction there.
-And they both have common interests and common enemies.

What would be ideal: AMD should work with Intel and get them in on Mantle. Then cooperate together for an open standard Gsync clone- use Intel's might to get it in every LCD display sold on the market. Destroying NV's initiative. Maybe even making this 'open standard' somehow difficult or impossible for Nvidia to support through legal barriers. Don't know, but it would be a good idea to do so, as it won't be long and OpenSync will be all over and GSync will be another Physx.
Lower performance parts benefit more from a Gsync type solution than high end anyway. It's ideal for APUs from Intel and AMD.

I don't understand. What exactly benefit do AMD have from working on mantle with intel? Just so intel may help them develop g-sync like hardware? I don't think it is that complicated. Intel's power to help deep integration might help, but there are different paths.

I think that it could be done on graphics card, without special display AIB. AMD have introduces a few open industry standards already. I think they can make it. Consoles are done, Hawaii is done, Kaveri is done. I guess they will have some spare resources soon.