AMD Introduces Mantle At GPU 14

Both AMD and NV hardware can run any API, so the more the merrier, I'll be more then happy to see the long rain of DirectX everything come to an end.
 
From this article on Anandtech:

Consequently while Mantle is good for AMD users, is Mantle good for NVIDIA and Intel users? Do developers start splitting their limited resources between Mantle and Direct3D, spending less time and resources on their Direct3D rendering paths as a result?

At the risk of walking a very fine line here, like so many aspects of Mantle these are not questions we have the answer to today.

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

So, MS might have no choice, but to be in and looks like Nvidia and Intel wont even have a choice.

Again, nothing concrete.
 
From this article on Anandtech:



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

So, MS might have no choice, but to be in and looks like Nvidia and Intel wont even have a choice.

Again, nothing concrete.

We will have to see if this takes precedence over the traditional API's we have right now. I don't think it will at all but it may make it a bit more convoluted in the long run. We will have to see how developers like this approach and see if they decide to take this and run with it and how far they run with it.

Obviously, they (AMD) wanted to simplify the process to code lower level, which I love. They chose to stick with their current architecture and not their past hardware configs as those are not/will not be supported soon. Seeing traditional performance from other hardware won't be bad though.
 
We will have to see if this takes precedence over the traditional API's we have right now. I don't think it will at all but it may make it a bit more convoluted in the long run. We will have to see how developers like this approach and see if they decide to take this and run with it and how far they run with it.

Obviously, they (AMD) wanted to simplify the process to code lower level, which I love. They chose to stick with their current architecture and not their past hardware configs as those are not/will not be supported soon. Seeing traditional performance from other hardware won't be bad though.

Kudos to them for trying to prime the pump by sending a sizable wire to EA/DICE for them to even give Mandible the time of day, but will it be enough to interest other developers, I mean they're not dumb they know EA doesn't just get spontaneously interested in anything, or will other "investments" have to be made by AMD?

Still I can appreciate what they're trying to do given their current market position, and anything where performance improvement is concerned I care less about brand. I guess in a perfect world they'd be moving toward an OpenGL future like NVIDIA but for now at least they're giving the finger to DirectX.
 
We will have to see how developers like this approach and see if they decide to take this and run with it and how far they run with it.

Per the included link, looks like developers were already asking for this.

Would love if this can be integrated on SteamOS. As described, it will be lightweight and if you add the benefits of Mantle, there is the possibility that Windows just wont be able to get the same performance on the same hardware.
 
where is everybody getting this info that valve and nvidia partnered up?
 
Looks like its going to be fast:

"When we asked what this means in real-world terms, he stated, “with Battlefield 4 running with Mantel (AMD’s new graphics API), the card will be able to ‘ridicule’ the Titan in terms of performance.” "
http://www.maximumpc.com/amd_r9_290x_will_be_much_faster_titan_battlefield_4

Nvidia might be able to create a wrapper for this or something. They do expose low-level access to their GPU's via OpenCL already. I hope they do put the effort into supporting Mantle though, since as you say, we waste so much system resources on our PC's.

Oh Wow, this explains Dice having the new cards so early! Lets see... if the 290x by itself beats Titan, add Mantle at a conservative 3x overall performance increase (Take 9x draw calls / 3)... This will be Impressive! :eek::eek::eek:

ya-know the more im reading about mantle and looking back at AMD's bid for the console wars.
it looks like they are bending nvida over the table and saying "you like that huh?"

Yes, this is AMD completely outflanking nvidia's attempts at locking them out with proprietary products. All while keeping mantle open source at the same time, very impressively done! The more developers gain experience coding the new consoles, the more I see AMD's performance increase to grow over the coming years.
 
A few things-
1. Some early talk suggested it would work on nvidia cards somehow. Obviously it can't. It doesn't even work on 6000 series radeon cards.
2. Linux support theoretically can happen, but games with mantle end up requiring windows anyway, so amd probably won't bother making a linux mantle driver. They don't make great linux drivers atm anyway.

anyway, as someone who probably is gonna get BF4, I'm happy I won't have to buy windows 8 for directx 11.1. I can stick to windows 7 and use mantle.
 
If AMD plays their cards right, they will be the biggest thing for the next 10 years of gaming. Unbelievable move they just pulled here.
 
So has MS announced that they are totally on board with mantle? I feel like MS will see this ploy and try to do something to thwart it.

They will probably open the wallet and throw AMD a wad of cash to keep this Mantle on the Windows platform only (other than XB1 / PS4, since this appears to be just that same API moved over to the PC).

Remember that the primary point of Direct3D is platform lock-in. The fact that they can use it to mosey a handful of customers up to a new version of Windows is certainly a benefit, but not the primary one. Microsoft lives on lock-in. So if they can guarantee Mantle is Windows-only (which should be easy), I see no reason why they would feel uncomfortable.


Also... people in this thread keep saying Mantle is "open source". Why is this being said? Somebody show me an authoritative statement from AMD or DICE that says Mantle is "open source" because I can't find one (and I've looked). Or, better yet, link me to the git repo that has the source code.

There was one off-the-cuff statement about the Mantle API being "open"... and people seem to be highly misinterpreting that. You know the CUDA API is "open" too. So is the DirectX API. All that means is that programmers writing applications (in this case, game engines) have a public Application Programming Interface. That's it. It means that you can make program calls into the Mantle driver w/o having to engage some kind of private NDA and licensing contract to do it. Some guy on some website actually interpreted that to mean that NVIDIA can just go off and make their own version of Mantle, without understanding what the hell that even means. This Mantle would have to be pretty high-level to allow for that.

Now maybe Mantle will be more open than what I'm saying here. Maybe it will be "open source" (which would require AMD open sourcing their drivers -- lol good luck with that). And maybe someday it actually will run on Linux (lol, AMD and Linux). But so far we have no information from AMD that even hints at that. So it's just unfounded speculation.
 
anyway, as someone who probably is gonna get BF4, I'm happy I won't have to buy windows 8 for directx 11.1. I can stick to windows 7 and use mantle.

This is what I really like most about AMD's initiative even though I've got an NV card. The gotcha mentality and artificial barriers MS has created to try to force paid upgrades to the newest Windows has always rubbed me the wrong way. All it's ended up doing is discouraged developers from focusing their game on the latest rendering tech because such a small pool can even take advantage.
 
AMD's GPU division possesses vision and foresight that their CPU division desperately needs.
 
AMD's GPU division possesses vision and foresight that their CPU division desperately needs.

One and the same. AMD has made a point of how Mantle will be greatly accelerate 8 cores. The API allowing the handcuffs to be taken off the CPU.
Recall, the new consoles are lower powered 8 core CPU's. How do you think this will affect AMD's much higher powered but low cost 8 core PC cpu's when running XBone/PS4 derived software?
 
A few things-
1. Some early talk suggested it would work on nvidia cards somehow. Obviously it can't. It doesn't even work on 6000 series radeon cards.
2. Linux support theoretically can happen, but games with mantle end up requiring windows anyway, so amd probably won't bother making a linux mantle driver. They don't make great linux drivers atm anyway.

anyway, as someone who probably is gonna get BF4, I'm happy I won't have to buy windows 8 for directx 11.1. I can stick to windows 7 and use mantle.

Its open as it should be open to any developer wanting to implement it. I'm not sure how this would work on any other architecture as you would have to have access to every bit of logic on older versions/competitor cores (which is impossible).

It COULD world for somethings that different cores do exactly the same way but I'm not sure if anyone would really try that.
 
I am wondering if it's possible that AMD will re-enter mobile segment as well. Can you imagine how powerful AMD's position can be if low-level API such as Mantle can be used to make games that would scale from mobile devices all the way up to desktop? That would cement AMD's place in the 3D graphics scene for years to come.
 
Its open as it should be open to any developer wanting to implement it. I'm not sure how this would work on any other architecture as you would have to have access to every bit of logic on older versions/competitor cores (which is impossible).

It COULD world for somethings that different cores do exactly the same way but I'm not sure if anyone would really try that.
Could and should

The amount of overhead they are trying to eliminate with this is not that great, it exists but a ton of overhead was minimized moving from dx9 to dx10/11. Also makes me think of this
http://www.nvidia.com/docs/IO/8228/BatchBatchBatch.pdf

And while 8 cores sounds like a good idea it's up to game programmers to properly thread a game to begin with, most times developers just do simple version in which say sound gets it's own thread, game ai gets it's own et al. works fine except not all threads are created equal so a bottleneck occurs, my only problem is i noticed the words "upto" which makes me wonder how is paralleling done.

Exacerbating this is the fact DX11 is still just starting to be used but any gains from removal of overhead is often negated by the fact they bump up the visual features with things like tessellation(which to most people is so minimal they wont notice without side by side caps).
 
Are people really going to code for this API. OpenGL is supported by every major platform this generation. Why take the development time to create a Mantle version if it only works on windows AMD machines?
 
Are people really going to code for this API. OpenGL is supported by every major platform this generation. Why take the development time to create a Mantle version if it only works on windows AMD machines?

because mantile will basically be running on both xbone and ps4 and that allow developers to bring over existing code to the PC. According to AMD, this is what developers have been asking for, and they are delivering on it.
 
Well that isn't comfirmed but assuming it's true. How much mantle code will they really want to put in considering it wont translate to the portion of the PC market occupied by nvidia, nor to the linux based steamOS (amd linux support is much worse than nvidia, who knows if mantle will be available in the linux drivers either.) With OpenGL they hit all the platforms.
 
Well that isn't comfirmed but assuming it's true. How much mantle code will they really want to put in considering it wont translate to the portion of the PC market occupied by nvidia, nor to the linux based steamOS (amd linux support is much worse than nvidia, who knows if mantle will be available in the linux drivers either.) With OpenGL they hit all the platforms.

We will see how "open" mantle is. Here is a good article on the possible pros and cons.
http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn
 
Are people really going to code for this API. OpenGL is supported by every major platform this generation. Why take the development time to create a Mantle version if it only works on windows AMD machines?

So far only one game is supporting Mantle (not even at launch) and you can bet AMD payed them a ton of money to do it. It's just marketing really.
 
So far only one game is supporting Mantle (not even at launch) and you can bet AMD payed them a ton of money to do it. It's just marketing really.

i really hope that is true
last thing PC gaming needs is for NV to return fire with there own API
 
So far only one game is supporting Mantle (not even at launch) and you can bet AMD payed them a ton of money to do it. It's just marketing really.

You mean the same thing Nvidia does with TWIMTBP and their previous tactics. Pot, kettle black. Good for AMD pushing things forward and out of the box. Loved Glide.
 
AMD making custom silicon for both consoles at near zero profit mainly for the PR propaganda value doesn't put them in any commanding market position, sorry.

I think you might be slightly off with your "near zero profit" estimation:

"We reiterate our Buy rating and view our estimates for AMD as conservative in light of design wins in both the PS4 and xBox One," analyst Mark Lipacis said. This follows a win in the PS4.

"We believe AMD will receive revenues of $60-$80 per PS4 and xBox One shipped for its processor," he said.

He continued, "We believe that both the PS3 and xBox 360 shipped between 70-to-75 million units each over the past few years. Assuming similar success, for the PS4 and xBox One, and assuming $60 ASP per processor and 20% operating margin would translate to $9 billion in revenues and $2.40 in EPS over 6 years, or an average annual EPS of $0.40."

http://www.streetinsider.com/Analyst+Comments/XBox+One+and+PS4+Could+Add+$9B+in+Revenue+for+AMD+%28AMD%29+-+Jefferies/8360630.html
 
A few things-
1. Some early talk suggested it would work on nvidia cards somehow. Obviously it can't. It doesn't even work on 6000 series radeon cards.
2. Linux support theoretically can happen, but games with mantle end up requiring windows anyway, so amd probably won't bother making a linux mantle driver. They don't make great linux drivers atm anyway.

A low level API won't work on hardware that isn't similar. So if Nvidia did support Mantle, then it's for compatibility. Otherwise Mantle would perform nearly the same as OpenGL or Direct3D does. So there's really no point. Also I'm willing to bet that Mantle would work on 6000 series cards, but AMD won't because that's not how they make money. Remember they make money on hardware sales, and supporting older hardware just delays customers from buying updated hardware. It's the the first time AMD or er ATI has done this sort of thing. Remember the Radeon 8500 series cards, but then ATI created the 9000 which could use pixel shaders to render video, but the feature never made it to the 8500. Then they created the 9100 cards, which a lot of people noticed they're nearly the same as 8500, but can do video pixel shader acceleration. Turns out people took the bios and flashed it over and suddenly the support for this feature worked.

I'm sure Mantle will make it's way to Linux. Gabe Newell will have something to say about it.
i really hope that is true
last thing PC gaming needs is for NV to return fire with there own API
That's exactly what PC gaming needs right now. Like it or not, low level api's need to come back. Without them we're be wasting even more system resources. Back in the day OpenGL and Direct3D were good ideas. Developers had to support a lot of graphic cards, and everyone had their own API's that needed support. It really didn't matter if it was low level or high level API, because the difference from going software rendering to hardware was a huge leap. But today graphic cards demand a lot of resources, and those resources are wasted. This isn't a new issues either, as developers have requested this for some time. There was an article on HardOCP about code to metal a couple of years ago.

You can bet Microsoft and Sony aren't happy. Right now they don't want PC games to get any upper hand over consoles. Why you think Mantle for BF4 for PC is December? Even AMD knows they don't want to make Sony and Microsoft too pissed off, but yea they're mad as hell.
 
You can bet Microsoft and Sony aren't happy. Right now they don't want PC games to get any upper hand over consoles. Why you think Mantle for BF4 for PC is December? Even AMD knows they don't want to make Sony and Microsoft too pissed off, but yea they're mad as hell.

So Mantle is not available with the BF4 launch, but a BF4 update in "December"? I read it as it will be there at launch, with another update in December.
 
The 6xxx & 5xxx series cards do not use GCN architecture. And although they worked efficiently, programming them was such a giant pain most people who tried just gave up, nothing was stable.

AMD did their homework and hired the right people with GCN.

As for mantle ... if AMD didn't release it, how long do you think it would take for someone with a Dev kit to "leak" a hacked version? I'm sure Sony, MS, and AMD discussed this possibility early on and all agreed on a course of action.
 
The 6xxx & 5xxx series cards do not use GCN architecture. And although they worked efficiently, programming them was such a giant pain most people who tried just gave up, nothing was stable.
Programming them how?

As for mantle ... if AMD didn't release it, how long do you think it would take for someone with a Dev kit to "leak" a hacked version? I'm sure Sony, MS, and AMD discussed this possibility early on and all agreed on a course of action.[/QUOTE]

Hard to believe. Here's a article that shows Radeon HD 7850 vs 6850 vs 5850 at same clocks. They're not exactly the same, but they perform similar enough to support the idea that they're all based on the same architecture. Turns out the 6770 and 6750's are really 5700 cards. So yea, considering how similar the GPUS are in features and performance, do you really think the 7000 series cards got special treatment?

Of course we'll find out for sure once Mantle is released. Either someone is going to softmod the drivers to get mantle support for 6000 and 5000 series working, or flash a bios onto your older Radeon cards to ID as a 7000. It'll be interesting to see if this happens.
 
They're not exactly the same, but they perform similar enough to support the idea that they're all based on the same architecture.

GCN allowed for some pretty specific changes to the way memory addressing is allowed to happen - specifically with the way it handles x86 addressing...

Which I feel is a big feature in the new consoles and their integrated APU type chips.
 
Programming them how?

As for mantle ... if AMD didn't release it, how long do you think it would take for someone with a Dev kit to "leak" a hacked version? I'm sure Sony, MS, and AMD discussed this possibility early on and all agreed on a course of action.

Hard to believe. Here's a article that shows Radeon HD 7850 vs 6850 vs 5850 at same clocks. They're not exactly the same, but they perform similar enough to support the idea that they're all based on the same architecture. Turns out the 6770 and 6750's are really 5700 cards. So yea, considering how similar the GPUS are in features and performance, do you really think the 7000 series cards got special treatment?

Of course we'll find out for sure once Mantle is released. Either someone is going to softmod the drivers to get mantle support for 6000 and 5000 series working, or flash a bios onto your older Radeon cards to ID as a 7000. It'll be interesting to see if this happens.[/QUOTE]

AMD's actual shader processor (SPU) hasn't changed much since the 4XXX series. The biggest thing is the memory addressing/fetching and shader scheduler. Mantle has to keep all of this into consideration. If AMD added in support for VLIW4 and prior set ups (which uses very similar SPU's granted) they would further convolute the code and make that low level coding a nightmare. They are smart to support their best and current layout with this, its also their most used. I have a feeling the next GCN core type will be somewhat backwards compatible in regards to coding.
 
So Mantle is not available with the BF4 launch, but a BF4 update in "December"? I read it as it will be there at launch, with another update in December.

Dice will be releasing the patch to enable Mantle in December. It was in the notes over at Anand I believe if you want to read his release notes.
 
AMD is likely anticipating another 5-10 year console life cycle as well... And as much as we all hate to admit it, console gaming comprises the vast majority of the gaming ecosystem.

If they can stay with the GCN architecture for the same length of time, who really cares if the API is superceeded by something different in half a decade or longer. It allows some substantial benefits in the interim.

If this low level API approach proves to be a boon to developers, there's a good chance they'll land another contract moving forward... and I find it hard to believe that an API - about to be adopted by 90% of the gaming audience - won't have future parts with full backwards compatibility... at least until the whole discussion is relevant.
 
agreed. this is nothing but gravy for the PC gaming realm. BF4 is going to be huge and will need all the optimizations it can get, given how BF3 performs.
 
Back
Top