The AMD Mantle Thread

[Creig]

I love being on this board during GPU releases. If you predict that an unreleased product is going to bring down Microsoft and Nvidia, nobody calls you on it.

You mean like somebody predicting that an unreleased product will somehow spell the eventual doom of the two most popular GPU manufacturers on the planet?

Unfortunately AMD would go out of business along with NVidia if this eutopian dream panned out that way.

Everything developed to run on old hardware = no need for new discreet GPUs = bye bye.

However, it isn't the actual API that could be the long-term problem. It is the continued dumbing down of PC games, and anything that makes it easier for a console-first developer to port a game to a PC, contributes to that process.

Just because it makes things easier to program across multiple platforms doesn't automatically mean that graphics levels will suffer. Some games will be written for consoles first and the PC second. And some games will be written for the PC first, then ported to consoles. The same as always.

What this really allows is more games to be more easily ported between consoles and PCs. This means that some games that ordinarily would ordinarily be tied to a single platform can instead be offered to all three. Don't you think that more games for more platforms is a worthwhile benfit?

You can scream from the rooftops that it is about AMD getting a couple FPS boost in a random benchmark that will be forgotten a week later, but it doesn't make it so.

I see. So from your viewpoint, Mantle is nothing more than a minor FPS bump in a random benchmark. Okay then.

Given the amount of effort the Nvidia fans are putting into downplaying Mantle, I'd say AMD must be on the right track with this.

 

[StrangerGuy]

Heh...I'd rather see 4k come along and force these companies to give me 60fps on it with more power. That's just me.

It's not limited to more GPU to up the resolution further. If you have the gall to require more GPU power than running Crysis I fully expect your game to look better, not worse. Otherwise go back and optimize that lazy POS of a game.

 

[cmdrdredd]

What this really allows is more games to be more easily ported between consoles and PCs. This means that some games that ordinarily would ordinarily be tied to a single platform can instead be offered to all three. Don't you think that more games for more platforms is a worthwhile benfit?



Given the amount of effort the Nvidia fans are putting into downplaying Mantle, I'd say AMD must be on the right track with this.

1) Sure if most of the best console games weren't developed by first party studios or bought and paid for exclusives. You'll never see Naughty Dog releasing PC versions of their games, nor many japanese developers doing it.

2) How do you know they are nvidia fans? Some people here that dislike this approach own GCN cards themselves.

It's not limited to more GPU to up the resolution further. If you have the gall to require more GPU power than running Crysis I fully expect your game to look better, not worse. Otherwise go back and optimize that lazy POS of a game.

So you're one of those. Crysis was good back then but things have progressed since then. I think there are a ton of games that look better than Crysis. Crysis was a very bland game once you got off the island area. In both graphics and gameplay.

Also yes it is up to your GPU when you go up in resolution. Why do you think a 6970 can't run playable framerates on 2560x1440 when it can at the same settings at 1920x1080 but a 7950 can? Simple...GPU power. At some point only faster GPUs can do more. You'll get to a point, even with mantle that you max out the potential. Consoles have done it, it takes a while though. I think this is what AMD wants, to never have to try chasing Nvidia's Maxwell if it delivers what I think it potentially could.

 

[Enigmoid]

I get that mantle is good and can see the benfits but I'm wondering by what factor does actual performance increase by reducing draw calls 9x? BF3 singleplayer is virtually CPU agnostic despite rendering a fairly complex, well detailed and visually impressive world. Simply put, BF3 singleplayer does not really need mantle and using mantle would not improve performance at all as you are pretty much always GPU limited. MP seems to be CPU limited due to the number of players, explosions, and physics that are going on, rather than draw calls. Even crysis 3 has its fps tank when you enter an area with a lot of moving grass (physics calculations). Can someone please explain?

 

[cmdrdredd]

I get that mantle is good and can see the benfits but I'm wondering by what factor does actual performance increase by reducing draw calls 9x? BF3 singleplayer is virtually CPU agnostic despite rendering a fairly complex, well detailed and visually impressive world. Simply put, BF3 singleplayer does not really need mantle and using mantle would not improve performance at all as you are pretty much always GPU limited. MP seems to be CPU limited due to the number of players, explosions, and physics that are going on, rather than draw calls. Even crysis 3 has its fps tank when you enter an area with a lot of moving grass (physics calculations). Can someone please explain?

I think we will have to wait for the benchmarks and someone to dive into the explanation about what's going on at that time. I can see what you're getting at I believe.

 

[Carfax83]

http://tech.slashdot.org/story/13/04/12/1847250/amd-says-there-will-be-no-directx-12-ever

It all makes sense now. AMD stonewalling the development of DirectX 12 by saying they won't make any cards that support any proposed DirectX 12 standard because they want to milk GCN for the 5-10 years the PS4 and Xbone are projected to last.

Well that's informative

I wrote a post earlier speculating how many draw calls a game like Crysis 3 has. Crysis 3 is extremely detailed, and has the best DX11 implementation of any game to date, plus it seems to have no problem utilizing even the most advanced multicore processor, and making sure the GPUs are well fed.

There is no object pop in in Crysis 3 either, at least none that I could obviously see.

Anyway, the point being that DX11 did a LOT to fix many of the problems developers are complaining about. It's not perfect, but it's a heckuva lot better than DX9.

DX12 will do even more I'm certain. So I find it interesting that AMD is all of a sudden trying to get rid of DirectX, especially after all these years.

Interesting..... :sneaky:

 

[24601]

Well that's informative

I wrote a post earlier speculating how many draw calls a game like Crysis 3 has. Crysis 3 is extremely detailed, and has the best DX11 implementation of any game to date, plus it seems to have no problem utilizing even the most advanced multicore processor, and making sure the GPUs are well fed.

There is no object pop in in Crysis 3 either, at least none that I could obviously see.

Anyway, the point being that DX11 did a LOT to fix many of the problems developers are complaining about. It's not perfect, but it's a heckuva lot better than DX9.

DX12 will do even more I'm certain. So I find it interesting that AMD is all of a sudden trying to get rid of DirectX, especially after all these years.

Interesting..... :sneaky:

Considering AMD/ATi has always been enthusiastic about supporting new DX versions before, I'm guessing this is how it went down.

August 2011: Rory Reed came into power
2012?: AMD/ATi becomes sole CPU and GPU supplier for PS4 and Xbone
Rory Reed, in his infinite IBM wisdom, decides to make the obvious choice in Capitalism, Decommoditizing his Commodity.

Well played Rory Reed, Well played.

EDIT:

This would also fit into the layoffs on the engineering team.

You don't need as large an engineering team if your plan is to rehash the same architecture for 5-10 years.

 

[VulgarDisplay]

I agree with almost everything you said here. :thumbsup:
There will be some major difficulties getting other developers on board here. Delays, money and frustration is what comes to mind.
Major AA titles take a few years to create. Doing two APIs at the same time, means hiring bigger staff, which AMD would have to pay since they are the only ones that use this system. Unless they make Nvidia pay for using it, but that will be the day when hell freeze over. I think Nvidia would rather make their own version instead.

Except, I still have difficulty believing that a brand new API will trump a well established and well polished system like DX right now. Perhaps several years down the line.


Yes I dont know a lot of the software side, thats why Im asking questions

Why would they have trouble getting developers on board? Its baked into the consoles. The vast majority of all developers will be coding for mantle capable hardware.

PC exclusives are the only ones that will be hard to get onboard but there are so few of those that it doesn't matter. Most devs beg for this low level access so getting them onboard will not be a huge issue I assume.

This is as much to combat the rising costs of game development as it is to fight nvidia. Amd is appealing directly to developers wallets with this.

 

[Creig]

To me this feels like AMD couldn't bring a card to beat the GTX 780 so they need to try convincing developers to use this new API so their cards can have better performance.

I think it's more at looking ahead and trying to solve your minor problems before they become major problems. In this case, it's die shrinks. As the manufacturing process gets smaller and smaller, we seem to have to wait longer and longer for the next process node to be mature enough for volume GPU production. In addition, GPUs just keep getting bigger and more complex with every generation. Troublesome dies and complex designs do not make for a pretty picture.

I believe one of the reasons behind the creation of Mantle is to try and squeeze as much performance out of current silicon as possible. Simply relying on a die shrink to increase performance isn't enough anymore.

 

[cmdrdredd]

Why would they have trouble getting developers on board? Its baked into the consoles. The vast majority of all developers will be coding for mantle capable hardware.

PC exclusives are the only ones that will be hard to get onboard but there are so few of those that it doesn't matter. Most devs beg for this low level access so getting them onboard will not be a huge issue I assume.

This is as much to combat the rising costs of game development as it is to fight nvidia. Amd is appealing directly to developers wallets with this.

Because Sony has their own API that developers have already been working with, Microsoft has redesigned DX11 for the Xbox One to be efficient as possible on that platform. Developers have already been using it. The dev boxes at the MS booth at e3 were Nvidia PCs in some cases too.

So devs already working on games aren't using mantle right now. It's just DICE for their engine and nobody else.

I think it's more at looking ahead and trying to solve your minor problems before they become major problems. In this case, it's die shrinks. As the manufacturing process gets smaller and smaller, we seem to have to wait longer and longer for the next process node to be mature enough for volume GPU production. In addition, GPUs just keep getting bigger and more complex with every generation. Troublesome dies and complex designs do not make for a pretty picture.

I believe one of the reasons behind the creation of Mantle is to try and squeeze as much performance out of current silicon as possible. Simply relying on a die shrink to increase performance isn't enough anymore.

Well I hope for their sake they don't expect us to sit on the same GPU for 5 years.

 

[VulgarDisplay]

Because Sony has their own API that developers have already been working with, Microsoft has redesigned DX11 for the Xbox One to be efficient as possible on that platform. Developers have already been using it. The dev boxes at the MS booth at e3 were Nvidia PCs in some cases too.

So devs already working on games aren't using mantle right now. It's just DICE for their engine and nobody else.



Well I hope for their sake they don't expect us to sit on the same GPU for 5 years.

Did you not read anything about mantle? Anand suspects that it may actually be the low level api in use on the Xbox. The devs showing their games on nvidia hardware at e3 were working on their games years prior to that trade show and the true specs of the next gen console were barely drying on the paper they were printed on. Mantle is obviously a very new development because it has surprised everyone. Given time more devs will pick this up.

 

[Shivansps]

What i do understand is that cant be a low-level optimisation if its has to be plug&play on any hardware. Its not low-level anymore, it has one or several layers on top.

 

[tviceman]

1. AMD's overall low PC adoption rate is going to hurt Mantle's ability to sway developers.
2. Mantle being proprietary to AMD PC hardware will hurt Mantle's ability to sway developers. See #1 which makes #2 an even harder sell.
3. Sony and MS have had development kits for the consoles for some time now. It will be a tough sell to convince developers to drop these development kits in favor of AMD's solution, even if better performance can be tapped.
4. Back in the day, native glide games didn't provide leaps and bounds better performance than directx did. It helped, but it wasn't a magical leap like AMD is claiming will be done with Mantle. Just like any other company hyping their new super tech, AMD is likely only claiming best case scenarios that will never be fully achieved.

Because of all this, I again say that Mantle won't have penetration beyond what physx has.

 

[VulgarDisplay]

1. AMD's overall low PC adoption rate is going to hurt Mantle's ability to sway developers.
2. Mantle being proprietary to AMD PC hardware will hurt Mantle's ability to sway developers. See #1 which makes #2 an even harder sell.
3. Sony and MS have had development kits for the consoles for some time now. It will be a tough sell to convince developers to drop these development kits in favor of AMD's solution, even if better performance can be tapped.
4. Back in the day, native glide games didn't provide leaps and bounds better performance than directx did. It helped, but it wasn't a magical leap like AMD is claiming will be done with Mantle. Just like any other company hyping their new super tech, AMD is likely only claiming best case scenarios that will never be fully achieved.

Because of all this, I again say that Mantle won't have penetration beyond what physx has.

As for number 1 and 2. Its a good thing amd PC hardware is in all next gen consoles and mantle works there as well. Should be good for an additional 150 million plus gcn GPUs to sway those pesky developers over the next 5-10 years.

 

[Enigmoid]

As for number 1 and 2. Its a good thing amd PC hardware is in all next gen consoles and mantle works there as well. Should be good for an additional 150 million plus gcn GPUs to sway those pesky developers over the next 5-10 years.

I don't think there are 150 million gcn gpus in existence at this point in time. You are probably off by at least an order of magnitude though I could be completely wrong.

 

[Rvenger]

Because of all this, I again say that Mantle won't have penetration beyond what physx has.

Completely 2 different animals and that is your opinion. We will see what mantle brings.

 

[flopper]

1. AMD's overall low PC adoption rate is going to hurt Mantle's ability to sway developers.


Because of all this, I again say that Mantle won't have penetration beyond what physx has.

with mantle the pc power will be even out.
you dont need as much to gain benefit for the game.
so this changes how the configuration of a normal gaming system due to bypassing DX so all the overhead and bottlenecks wont happen the way it does today.

its a new rulebook, physX was Nvidia trying to lock it in.
any such standard will be hard to do as none want to pay extra for stuff, the main difference from Glide is, the developers asked for mantle, and anyone I know off - if you ask for it and recive your going to use it.

when mantle is out, the fotnot might be darn...its good

 

[wlee15]

1. AMD's overall low PC adoption rate is going to hurt Mantle's ability to sway developers.
2. Mantle being proprietary to AMD PC hardware will hurt Mantle's ability to sway developers. See #1 which makes #2 an even harder sell.
3. Sony and MS have had development kits for the consoles for some time now. It will be a tough sell to convince developers to drop these development kits in favor of AMD's solution, even if better performance can be tapped.
4. Back in the day, native glide games didn't provide leaps and bounds better performance than directx did. It helped, but it wasn't a magical leap like AMD is claiming will be done with Mantle. Just like any other company hyping their new super tech, AMD is likely only claiming best case scenarios that will never be fully achieved.

Because of all this, I again say that Mantle won't have penetration beyond what physx has.

1. There's a big difference between the Glide, DX7/8 era and the current generation. If you're writing a shader during the Glide era you're basically writing ASM. Now you're writing with high level shading languages such as HLSL, GLSL, Cg, PSSL.

2. There's also far less games being made from the ground up now, so rather than convincing hundreds of game developers to use Mantle you need to get makers of graphics engine and middleware onboard. They've got EA onboard with Frostbite 3 which is very important now it's going to be important that they get your Unreal, and Crytech.

 

[Silverforce11]

1. There's a big difference between the Glide, DX7/8 era and the current generation. If you're writing a shader during the Glide era you're basically writing ASM. Now you're writing with high level shading languages such as HLSL, GLSL, Cg, PSSL.

2. There's also far less games being made from the ground up now, so rather than convincing hundreds of game developers to use Mantle you need to get makers of graphics engine and middleware onboard. They've got EA onboard with Frostbite 3 which is very important now it's going to be important that they get your Unreal, and Crytech.

This glide comparison is nonsense. AMD have the benefit of their architecture in the major gaming machines of the future, 3dfx did not command the same leverage.

Mantle only needs to be supported in a few major game engines mentioned above (including SquareEnix's and Unity too), and nearly all the major games will run faster on GCN. That is the difference these days, with many studios licensing a game engine and build their game from there. Very few actually go out and create a new engine for themselves, it simply is too much $$ and effort involved.

 

[VulgarDisplay]

I don't think there are 150 million gcn gpus in existence at this point in time. You are probably off by at least an order of magnitude though I could be completely wrong.

According to wikipedia ( best source right?) the ps3 and Xbox 360 sold over 150 million units. I fully expect more than that to sell this time around because gaming gets more mainstream every year. So there aren't that many yet, but there will be in short order. Nvidia should be very worried about this news. They will do fine in the hpc sector but they unwittingly dug themselves into a hole by not going after consoles.

Imagine if china allows these consoles into their country? They won't be able to make enough of the things.

 

[Zanovar]

Hmm i think quite a few poeple are getting quite ahead of themslves.tbh im not sure if this is is a good thing or not.(more reading to do).Ill just buy an amd card *shrugs*haha

 

[blackened23]

Glide is not a valid comparison to Mantle. Anyone that had a 3dfx card back in those days, around 1996 or so, remember that there were 6-8 different brands competing with each other and 3D rendering APIs did not exist whatsoever. There was no D3D. There was no DirectX - so everything *had* to be a native API. This led to a mess with the PowerVR using a specific rendering API, the rendition Verite using another one, and 3dfx using glide, of course. And there was the S3 Virge which also had it's own thing going on. The point here is that was no single standard, there 6 disparate chips all competing with each other and they all had different rending APIs. This led to a MESS with any games that had built in 3d rendering - one example coming to mind is quake. There was the 3dfx mini GL wrapper. Then there was vQuake for rendition verite cards. Then PowerVR? Sorry, no go, you're using software quake. And then the initial version of Tomb Raider supported the PowerVR chip. But it did not support Glide or Rendition. It was a mess. Basically some games used proprietary APIs, and others were forced to use software rendering (which looked like garbage).

A common unified APi made sense then because there were 6 competing standards and none of them were obviously compatible with each other. Now? Mantle will work on GCN architecture, and you will still get DirectX. Developers will still make DX versions of their games - let's not kid ourselves - some games will use mantle and GCN will benefit, but this will in no way prevent nvidia from playing those same games with DirectX. So I find the hostility and ruffled feathers to be absolutely hilarious. What do YOU guys care if Mantle isn't adopted? Taking this awfully personally eh? I use a 780 and I don't really care if it does or doesn't take off. If it does take off, my hats off to AMD. They apparently aren't being idiots for once in the past ten years and are using their hardware as leverage over the PC gaming ecosystem JUST as nvidia has done for years.

Now with regard to Mantle being adopted on a wide scale, that's a tough question. Probably not. But it is safe to say that AMD has SOME leverage on various vendors who can integrate the API into their engines (Frostbite 3 is due to be used in roughly 10 EA games) and then there is the rumor that Mantle is similar the XB1 API. Nobody knows the specifics of the next-generation console APIs, not a single person here, so no one can confirm or deny this. But it could be possible. If it is intertwined with any next-generation console API, that would be a big win for AMD for sure.

But in the end, I don't understand the ruffled feathers. It's pretty hilarious. Even more hilarious are the folks trying to downplay BF4 being a win for AMD. Come on, who are you trying to kid here? BF3 far outsold any PC game by a huge amount in the past 3 years, and it is still selling. The same will happen with the PC version of BF4. Anyway, Nvidia users will still get their directx. If mantle doesn't take off, that's fine, nothing will change. But Mantle will be available in a wide variety of EA games since many of their games will be using Frostbite 3.

 

[Shivansps]

If what Anadtech suspects its true i can see how this can made easier to port PC games to xbox consoles, but i dont see much gain on the PC side, i can see fragmentation, split already limited work time on 2 APIs and low-level programing incompatibilities on Nvidia and Intel(if they ever implement it)

 

[blackened23]

There is no fragmentation because everything still has D3D. This isn't 1996 where you had software rendering or proprietary rendering -- every Mantle game will also have directX. So there is absolutely zero fragmentation. Everybody, nvidia - intel and AMD all get 3d accelerated rendering. This is as opposed to 1996 where you MIGHT get 3d rendering, but if you chose the wrong card you had CPU rendering which looked horrible. Such as the rendition verite: The verite had an excellent version of vQuake yet was SOL when Tomb Raider for PC was released which supported the powerVR. So you had a situation in 1996-1998 where some games were forced to use software rendering, but if you were lucky and had the RIGHT card you could get 3d accelerated rendering. (Luckily, I made the right choice with my CL voodoo and voodoo II, good times!) Anyway, that is not the case here therefore there isn't a situation where some get 3d and some do not. Everyone gets 3d acceleration. Intel and NV get DX and they will do just fine. Times are different. There is zero fragmentation, everything will still have DirectX, period.

 

[Zanovar]

Glide is not a valid comparison to Mantle. Anyone that had a 3dfx card back in those days, around 1996 or so, remember that there were 6-8 different brands competing with each other and 3D rendering APIs did not exist whatsoever. There was no D3D. There was no DirectX - so everything *had* to be a native API. This led to a mess with the PowerVR using a specific rendering API, the rendition Verite using another one, and 3dfx using glide, of course. And there was the S3 Virge which also had it's own thing going on. The point here is that was no single standard, there 6 disparate chips all competing with each other and they all had different rending APIs. This led to a MESS with any games that had built in 3d rendering - one example coming to mind is quake. There was the 3dfx mini GL wrapper. Then there was vQuake for rendition verite cards. Then PowerVR? Sorry, no go, you're using software quake. And then the initial version of Tomb Raider supported the PowerVR chip. But it did not support Glide or Rendition. It was a mess. Basically some games used proprietary APIs, and others were forced to use software rendering (which looked like garbage).

A common unified APi made sense then because there were 6 competing standards and none of them were obviously compatible with each other. Now? Mantle will work on GCN architecture, and you will still get DirectX. Developers will still make DX versions of their games - let's not kid ourselves - some games will use mantle and GCN will benefit, but this will in no way prevent nvidia from playing those same games with DirectX. So I find the hostility and ruffled feathers to be absolutely hilarious. What do YOU guys care if Mantle isn't adopted? Taking this awfully personally eh? I use a 780 and I don't really care if it does or doesn't take off. If it does take off, my hats off to AMD. They apparently aren't being idiots for once in the past ten years and are using their hardware as leverage over the PC gaming ecosystem JUST as nvidia has done for years.

Now with regard to Mantle being adopted on a wide scale, that's a tough question. Probably not. But it is safe to say that AMD has SOME leverage on various vendors who can integrate the API into their engines (Frostbite 3 is due to be used in roughly 10 EA games) and then there is the rumor that Mantle is similar the XB1 API. Nobody knows the specifics of the next-generation console APIs, not a single person here, so no one can confirm or deny this. But it could be possible. If it is intertwined with any next-generation console API, that would be a big win for AMD for sure.

But in the end, I don't understand the ruffled feathers. It's pretty hilarious. Even more hilarious are the folks trying to downplay BF4 being a win for AMD. Come on, who are you trying to kid here? BF3 far outsold any PC game by a huge amount in the past 3 years, and it is still selling. The same will happen with the PC version of BF4. Anyway, Nvidia users will still get their directx. If mantle doesn't take off, that's fine, nothing will change. But Mantle will be available in a wide variety of EA games since many of their games will be using Frostbite 3.

nvm