The AMD Mantle Thread

[Stuka87]

What I would like to know is what problems are there in DX11 where major gains could be had and how they would translate into real world performance.

Are you familiar at all with software development? If not, you most likely wont fully understand the answer to that.

But to be extremely general, anytime you have an API that is at a higher level, you have more overhead, as more work has to be done before the call that you make actually makes it to the lower level.

DirectX is a fairly high level API, as it is not hardware specific. Mantle is a very low level API that bypasses a lot of the translation that has to go on in order for DirectX to work. Mantle also supports calls that do not exist in DirectX, as they are GCN only.

 

[flopper]

What I would like to know is what problems are there in DX11 where major gains could be had and how they would translate into real world performance.

Its guesses as far.
can be a lot since you can optimize the gameengine for the GCN.
I assume somewhere between 20-60%

 

[SiliconWars]

http://fudzilla.com/home/item/32730-battlefield-4-on-mantle-should-end-up-much-faster

We have spent some time talking to people close to the matter, both from EA/ DICE and AMD and we have learned that both parties anticipate a significant performance boost with the Mantle version of BF 4.

 

[Creig]

Its guesses as far.
can be a lot since you can optimize the gameengine for the GCN.
I assume somewhere between 20-60%

It's probably no different than previous game development. A studio could choose to employ extensive use of low-level programming, in which case they can see a significant increase in performance. They could also decide instead to use just a few low-level calls to alleviate the worst of their performance bottlenecks and use a high-level API for everything else.

The amount of performance increase is going to be a complete variable from game to game.

 

[Kenmitch]

It's probably no different than previous game development. A studio could choose to employ extensive use of low-level programming, in which case they can see a significant increase in performance. They could also decide instead to use just a few low-level calls to alleviate the worst of their performance bottlenecks and use a high-level API for everything else.

The amount of performance increase is going to be a complete variable from game to game.

Maybe a hacker/developer could just make a app that takes advantage of the low-level api's ability to add the benefits to games that don't have it.

Maybe call it Mantle-Izer.

 

[smackababy]

Maybe a hacker/developer could just make a app that takes advantage of the low-level api's ability to add the benefits to games that don't have it.

Maybe call it Mantle-Izer.

That wouldn't really be possible without having access to the source code.

 

[VulgarDisplay]

Maybe a hacker/developer could just make a app that takes advantage of the low-level api's ability to add the benefits to games that don't have it.

Maybe call it Mantle-Izer.

It's called a game engine. For example, Unreal Engine, Cryengine, Frostbite, etc.

 

[Stuka87]

Maybe a hacker/developer could just make a app that takes advantage of the low-level api's ability to add the benefits to games that don't have it.

Maybe call it Mantle-Izer.

That would defeat the purpose. You would have to add an intermediate API on top of Mantle, which then adds more overhead, and you may as well just stick with DirectX.

 

[Skurge]

Anyone make an mention of weather Mantle will also increase IQ. Better lighting or AA, etc?

 

[smackababy]

Anyone make an mention of weather Mantle will also increase IQ. Better lighting or AA, etc?

I think the idea is that Mantle will make rendering calls much faster, allowing better IQ at equal performance or the same IQ at improved performance.

However, that is all up to the developers and how fleshed out the Mantle API actually is. And then, how much time a developer is going to put into Mantle development for a small segment of the market. They will likely have to make a DX version anyway, so after optimizing that to their content, how much extra effort would they put into optimizing a Mantle version, if any? It is probably better to just believe they will recompile with the changes of DX to Mantle (if Mantle provides a high level API like DX) and the performance will scale as such.

 

[blackened23]

No offense, but your arguments sound no different than the hardcore nvidia fans who will hear nothing of Mantle. Experienced developers have already commented on the state of the PC API being sad, and if I had to guess, you have zero knowledge of programming. I'm quite sure of this. So with that being the case, trying to argue technicalities is pretty hilarious.

I agree with the sentiment that we need to see real benchmarks, but i'm keeping an open mind. Nobody is saying you should buy into the hype. You, on the other hand, seem to be arguing from the perspective of someone who isn't objective and doesn't have programming knowledge on top of this.

Let's just wait and see. In the meantime, the 290X should be evaluated under current metrics while ignoring Mantle. When Mantle comes to use, then we can see if it's a good value add. Discussing it prior to then, especially without technical knowledge, is worthless. Like I said - 290X will be evaluated by the current metrics when it is released because Mantle is a non issue until December. In other words, it will be measured by DX11 benchmarks. So the 290X will just be viewed and judged completely by it's DX11 performance, as it should be - If Mantle changes things, we can judge it then.

To be clear: 290X won't be judged by Mantle at launch. It will be judged by DirectX.

 

[smackababy]

No offense, but your arguments sound no different than the hardcore nvidia fans who will hear nothing of Mantle. Experienced developers have already commented on the state of the PC API being sad, and if I had to guess, you have zero knowledge of programming. I'm quite sure of this.

I agree with the sentiment that we need to see real benchmarks, but i'm keeping an open mind. You, on the other hand, seem to be arguing from the perspective of someone who isn't objective and doesn't have programming knowledge on top of this.

Let's just wait and see. In the meantime, the 290X should be evaluated under current metrics while ignoring Mantle. When Mantle comes to use, then we can see if it's a good value add. Discussing it prior to then, especially without technical knowledge, is worthless. Like I said - 290X will be evaluated by the current metrics when it is released because Mantle is a non issue until December. In other words, it will be measured by DX11 benchmarks. So the 290X will just be viewed and judged completely by it's DX11 performance, as it should be - If Mantle changes things, we can judge it then.

I find this post pretty laughable. 10/10. Would read again.

 

[gorobei]

http://techreport.com/news/25461/report-amd-to-introduce-arm-based-tablet-chip-this-year

looks like this might be part of the multiplatform tactic for mantle. ARM core with GCN soc much like tegra. i guess casual market can get some benefit from direct to metal. probably not in games but maybe in augmented reality stuff.

 

[Paul98]

Normally I don't think much of these sort of things, but considering the consoles video cards, and this being a part of frostbite. I could see this getting major adoption if it's worth the extra effort.

 

[Skurge]

From what I understand about draw calls. They should be able to have a lot more stuff on screen while reducing CPU load. I'm not sure how that would affect a game like bf4. I guess a ton more particle when walls are blown up? Or a much denser forest in dragon or a lot more npcs in mass effect? Maybe every character having tressfx instead of just the main character? I'm just wondering what kind of benefits we can see outside of performance.

 

[ShintaiDK]

From what I understand about draw calls. They should be able to have a lot more stuff on screen while reducing CPU load. I'm not sure how that would affect a game like bf4. I guess a ton more particle when walls are blown up? Or a much denser forest in dragon or a lot more npcs in mass effect? Maybe every character having tressfx instead of just the main character? I'm just wondering what kind of benefits we can see outside of performance.

Even assuming you reduce CPU load significantly. There is also the GPU load that doesnt change.

 

[Skurge]

Even assuming you reduce CPU load significantly. There is also the GPU load that doesnt change.

Yes, but it should make better use of the GPU. No?

 

[krumme]

The consumers and the shareholders will share the benefits.

The shareholders in the gaming industry taking 80% 5% for amd 5% for us consumers because of increased competition.

Dont sniff of 5%. There is bigger money in this world that we can dream of. So 5% can really account for something.

 

[ShintaiDK]

Yes, but it should make better use of the GPU. No?

No.

 

[Deasnutz]

No.

Of course it should. The question I have is how crossfire scaling might improve.

 

[Deasnutz]

I think it might be pretty fair to infer the performance we've seen so far of Battlefield 4 on the two next-generation consoles to mantles potential. If we do we should all temper our expectations of what it will do.

 

[Imouto]

Yes, but it should make better use of the GPU. No?

I just hope this makes MMORPGs a bit less crappy.

 

[Jacky60]

A week late to the party but clearly Mantle is going to make some new games far better performing and hopefully all of them. Now I can believe the 290x could be twice as fast as Titan in BF4. This surely has the potential to completely transform the PC games market and very seriously hurt Nvidia in the gaming space.
Will existing cards with GCN architecture benefit from new games coded with Mantle e.g. BF4?
I'm pretty sure I know the answer is yes but correct me if wrong please.

 

[Jacky60]

I think the idea is that Mantle will make rendering calls much faster, allowing better IQ at equal performance or the same IQ at improved performance.

However, that is all up to the developers and how fleshed out the Mantle API actually is. And then, how much time a developer is going to put into Mantle development for a small segment of the market. They will likely have to make a DX version anyway, so after optimizing that to their content, how much extra effort would they put into optimizing a Mantle version, if any? It is probably better to just believe they will recompile with the changes of DX to Mantle (if Mantle provides a high level API like DX) and the performance will scale as such.

Given the lack of raw power of both next gen consoles it seems to me likely that to make games look pretty/have better effects a developer would use Mantle within certain constraints (financial). If you add the likely combined adoption numbers of both PS4 and Xbox One and all AMD discrete GPU owners they're substantially more than Nvidia's. It's a superb way to leverage loads more performance from the relatively inferior CPU's AMD has and slower GPU performance today. We saw how PS3 games (notoriously difficult to code for initially) developed over the life of that console (The Last of us) and it would be hard to believe they ran on the same machine, same with Crysis3 on xbox 360. Developers, as the fat sweaty man said, really work hard and this will significantly lower the effort/performance (pretty stuff on screen) ratio for them on both next gen consoles and all ported games for PC. Now I understand why the specs for the PS4 and Xbox1 are so meagre..because with Mantle they can be. I'm going to have to start rooting for Nvidia now.

 

[Erenhardt]

Yes, but it should make better use of the GPU. No?

Form what I understand we can expect heat problems on GCN cards with second class coolers while under 125% :ninja: Mantle load. That depends if they are going to release the beast within.