The AMD Mantle Thread

[Skurge]

Explain me how you can add multi vendor support to Mantle whiout making it another DX? there is no way to do that whiout abstractions.

Mantle does have abstraction layer. Mantle is designed for Modern GPUs. It tosses backwards compatibility to make better use of the hardware and to give the Devs more control. It shouldn't be hard to make it work with modern GPUs.

 

[AnandThenMan]

So both Dx, and supposedly mantle when we actually see the numbers, have advantages and disadvantages. It still remains to be seen how it all shakes out, but to try to paint Dx as holding back gaming is only looking at part of the equation. It may be holding back image quality, but it does offer forwards, backwards, and across hardware compatibility, which has made PC gaming the platform that it is.

This very argument has been used to talk about why PhysX is bad for the industry. The difference is, Mantle does not arbitrarily exclude visuals, why is it so hard for some to understand this. Mantle doesn't make DX cease to exist.

On abstraction layers, just about everything out there has code layers. It's becoming more and rare for any piece of hardware to have software that talks directly to it. At one time it was a necessity because processing power was extremely limited, now coding inefficiencies are nicely cloaked because of massive processing power. But the pendulum has swung too far, way too many cycles are wasted.

 

[bystander36]

You seem unable to imagine a world where DX and Mantle can coexist. Now if Mantle becomes so successful that most game devs won't even want to bother with DX, then either MS will have to pull up their socks, or the likes of Nvidia will have to get on board with Mantle. If Mantle does little in the gaming world, then nothing will have changed and we'll continue to be stuck in DirectX mud.

But most likely, Mantle is going to push gaming forward it is the inevitable progression of things. And more importantly (also very concerning for MS) Mantle is not specifically tied to Windows.

I'm quite capable of imagine that. And if Mantle is everything it is supposed to be, that means more hardware specific API's, more fragmented software, and DX will still have to be supported.

So DX is not a fault if you believe it still has to be supported. It does exactly what it is supposed to. It allows all hardware to work on it, within reason. Mantle is aimed at a small minority that they hope to grow. The problem is that it is a start down a vary dark path, one of which DX has solved in the past.

You are right, DX will be supported, which means what you call a fault, is a strength. It will continue to be supported because DX does what it has to. Mantle is trying to show that multiple API's is ok. Let's hope it isn't overly successful, or we may all end up suffering in the future.

 

[Cloudfire777]

That's a whole lot of assumptions you're making. But you do agree that Mantle is going to at the very least show 45% uplift in some scenarios.

On CPU bound games I think you will see the most gains for i7 systems. But I don`t think they will be nearly as close to 45%. But thats where the reduced overhead will do its magic. On games that doesn`t scale with CPU, like BF4, the improvements will be very little on i7 systems.
RTS games like Starcraft and such will show most gains.

There is a reason why they chose to showcase that Oxide game which is the typical CPU bound game. They did show BF4 during CES, but it was kinda like "See here is BF4 running on Mantle" and that was pretty much it.

That is my forecast of Mantle. Then we will see how Microsoft will react with DirectX12 (or 11.x) and how it will react to Mantle. They have probably already done similar improvements like Mantle on the Xbox One version of DX.

Competition drives things forward. Its a good thing

 

[bystander36]

This very argument has been used to talk about why PhysX is bad for the industry. The difference is, Mantle does not arbitrarily exclude visuals, why is it so hard for some to understand this. Mantle doesn't make DX cease to exist.

For starters, GPU PhysX is the only part that gets excluded to one piece of hardware, and it is typically a bit of added fluff, not a rewrite of the whole API. Because it is limited to Nvidia, you do not see it used often. Though general PhysX runs well on all CPU's.

 

[3DVagabond]

As others have said, Dx is actually what allowed the great advantages of PC gaming. The most appealing part of PC gaming to me is the ability to play current games, future games, and past games and not worry whether my hardware and OS will be compatible, as well as to play very different types of games. There is no current game that I have played, and I have a relatively low end system, in which I have said "this game sucks because it does not have enough units on the screen, or the image quality is too low." Would more be nice, of course, but there is a price to pay for that in compatibility. If I could not play it because of the lack of an appropriate API for my hardware, *that* would suck.

So both Dx, and supposedly mantle when we actually see the numbers, have advantages and disadvantages. It still remains to be seen how it all shakes out, but to try to paint Dx as holding back gaming is only looking at part of the equation. It may be holding back image quality, but it does offer forwards, backwards, and across hardware compatibility, which has made PC gaming the platform that it is.

From what the devs are saying not only is DX holding them back artistically (which is important to the people making the games) but all of this compatibility is not automatic by any means. It's actually really really difficult, and the thicker and more widely compatible it gets the more effort it's taking to get anything to run on it. Sometimes with technology things reach the end of their usefulness and it's better to start with a clean sheet of paper than to continue to patch the old one.

 

[krumme]

While is not gona be any different, 3 API gona be a mess, thats is even more incentive to program on DX and forget about dealing with 3 brands stuff.

It also gona be a dissaster for us too, i hope that never happen, thats also why i hope Mantle to fail badly, i dont want Intel and Nvidia to respond with their own APIs.

How is it a mess? When you start the game you will hardly notice it.

Programming for multiple cores in dx is a mess
Having many drawcalls in dx is impossible or at least a major mess
Debugging in dx is a mess because the programmer dont know if the error is in the huge driver
Memory control in dx is a mess
Updating 250 megs dx driver to get 3% perf in a game is a mess

All this repeating1995 fragmenting talk is unsubstatiated.

 

[Gloomy]

For starters, GPU PhysX is the only part that gets excluded to one piece of hardware, and it is typically a bit of added fluff, not a rewrite of the whole API. Because it is limited to Nvidia, you do not see it used often. Though general PhysX runs well on all CPU's.

For starters, Mantle is the only part that gets excluded to one piece of hardware, and it is typically a bit of added fluff, not a rewrite of the whole game. Because it is limited to AMD, you do not see it used often. Though general Direct3D runs well on all GPUs.

 

[bystander36]

For starters, Mantle is the only part that gets excluded to one piece of hardware, and it is typically a bit of added fluff, not a rewrite of the whole game. Because it is limited to AMD, you do not see it used often. Though general Direct3D runs well on all GPUs.

You do realize Mantle is the whole graphics API, and not a little bit of extra physics, right? I guess not. With PhysX, every system will use PhysX as their physics engine, then the dev's add a few extra features that requires a GPU from Nvidia for added effect.

That's quite different than a different graphics API for different brands.

 

[3DVagabond]

1. No matter how good amd technology and products is. Without cash you can not use it. Production scalability and production is hindered at gf. There is no money for oem developmemt. B2c and b2b marketing. brand development. Logistics. Sales. Everything is kept down. Withoit cash yoir business is severely constrained. You have next to no options for new initatives and moves. Therefore is amd weak.

Mantle is special because its a low cost developer driven product. Its eg in EA interest to get a competitive advantage.

2.
I personally think some of nv business practice is very unethical. But they know how to run a business and they know what they sell. And lets not forget how efficient kepler is. Nv is in a very dificult position because of amd and especially intel igpu getting comparable better to the dgpu each year. They are going to leave at some time and they are betting at arm and mobile big time. Untill now unsuccessfully but its always difficult to start and lets give them some time. One year with denver we will know.

I already said the definition of AMD being the smallest player wasn't important. I disagree with you, but that has nothing to do with nVidia being in the position to somehow take on M$.

So you are saying that nVidia is just going to get out of the Windows business?

 

[AnandThenMan]

You do realize Mantle is the whole graphics engine...

Direct3D and OpenGL are a graphics engine?

 

[krumme]

As others have said, Dx is actually what allowed the great advantages of PC gaming. The most appealing part of PC gaming to me is the ability to play current games, future games, and past games and not worry whether my hardware and OS will be compatible, as well as to play very different types of games. There is no current game that I have played, and I have a relatively low end system, in which I have said "this game sucks because it does not have enough units on the screen, or the image quality is too low." Would more be nice, of course, but there is a price to pay for that in compatibility. If I could not play it because of the lack of an appropriate API for my hardware, *that* would suck.

So both Dx, and supposedly mantle when we actually see the numbers, have advantages and disadvantages. It still remains to be seen how it all shakes out, but to try to paint Dx as holding back gaming is only looking at part of the equation. It may be holding back image quality, but it does offer forwards, backwards, and across hardware compatibility, which has made PC gaming the platform that it is.

I agree. But how is that going to get different when mantle is here?

(Programmability might actually be a big factor also but lets put that aside)

 

[bystander36]

Direct3D and OpenGL are a graphics engine?

Sorry for the typo.

 

[Gloomy]

You do realize Mantle is the whole graphics engine, and not a little bit of extra physics, right? I guess not.

Nice, you moved the goal post.

You defend PhysX because the hardware locked part is just extra fluff. How is that any different than defending Mantle when the hardware locked part is just extra fluff (in the form of performance)?

Serious question. :|

 

[blackened23]

I think the point was, you're comparing a full graphics API to something that isn't a graphics API? I could be wrong though.

 

[Gloomy]

Stop shifting the goalposts.

Both AMD and Nvidia users get the same game when a game supports Physx. Nvidia users get extra sparkles from Physx though.

Both AMD and Nvidia users get the same game when a game supports Mantle. AMD users get extra performance from Mantle though.

How is that any different? Explain it to me. Because I see absolutely no difference.

 

[bystander36]

Nice, you moved the goal post.

You defend PhysX because the hardware locked part is just extra fluff. How is that any different than defending Mantle when the hardware locked part is just a extra fluff (in the form of performance)?

Serious question. :|

Assuming you are serious here.

With PhysX, you are talking about a few settings. A few extra features that are even available to run on any system, they just run poorly on ones without an Nvidia GPU. When bugs are fixed, when the game is fine tuned, the majority of the fixes work for everyone. The dev's only have to focus on a single API.

With Mantle, you'll have two separate code paths. Two different API's. Their efforts have to be split. And if Mantle proves to be very successful, that will likely lead to more API's, and more split focus.

And even with PhysX's one API that works both and GPU's and CPU's, you still find the hard hitting PhysX that requires GPU PhysX isn't used often, because it is for one brand.

 

[AnandThenMan]

Assuming you are serious here.

With PhysX, you are talking about a few settings. A few extra features that are even available to run on any system, they just run poorly on ones without an Nvidia GPU. When bugs are fixed, when the game is fine tuned, the majority of the fixes work for everyone. The dev's only have to focus on a single API.

With Mantle, you'll have two separate code paths. Two different API's. Their efforts have to be split. And if Mantle proves to be very successful, that will likely lead to more API's, and more split focus.

And even with PhysX's one API that works both and GPU's and CPU's, you still find the hard hitting PhysX that requires GPU PhysX isn't used often, because it is for one brand.

Almost all of your concerns are actually unfounded according to all the devs I've listened to talking about Mantle.

 

[bystander36]

Almost all of your concerns are actually unfounded according to all the devs I've listened to talking about Mantle.

Yet ones you aren't listening to, are concerned. Funny how that works.

 

[Gloomy]

Alright, I get your point. I still think Mantle and PhysX are more alike than different though.

 

[bystander36]

If Mantle works well, but doesn't blow Nvidia out of the water, it can coexist without messing up the industry. But if Nvidia has to come up with its own API, all these same dev's agree, it would be a terrible thing.

In every scenario, DX or something similar, needs to exist.

 

[krumme]

Yet ones you aren't listening to, are concerned. Funny how that works.

Mantle obviously gets tons of traction from the devs. Otherwise you wouldnt be discussing here. I dont think you do it to protect your grand children from the evil mantle giving them the cute physx instead

 

[moonbogg]

So, after the BF4 Mantle patch comes and the results are in, will people still argue about Mantle? If its sucks, will the issue be settled? If it kicks epic ass, will the issue be settled?

 

[AnandThenMan]

Yet ones you aren't listening to, are concerned. Funny how that works.

I'm a bit confused by all of what you're saying. Are you concerned Mantle will be so successful that this will fragment gaming and ultimately hurt gamers?

And yes the devs I've listed to that are actually using Mantle have been very positive, it is the game developers that asked for a Mantle like API in the first place, don't forget that. What devs have voiced the extreme concern regarding Mantle that you are?

 

[blackened23]

What devs have voiced the extreme concern regarding Mantle that you are?

Former developer John Carmack and Tim Sweeney of Epic Games. Unreal Engine (Epic Games) is also the most often used multi platform engine, I should add - UE3 was used as the engine from around 70% of multi platform 360, PS3 and PC games if I remember correctly. UE4 will be the same. Sweeney actually seemed pretty hostile to the idea of Mantle. So I guess that fits under extreme concern.

The only devs i've seen praising Mantle were those paid to be on board by AMD. I mean, I don't have an issue with AMD Paying DICE. There's nothing wrong with that. But you have to look at it for what it is - let's not kid ourselves here. When AMD is paying you to use Mantle, you don't bite the hand that feeds.

If AMD is paying you 8 million bucks, you're not going to go badmouth Mantle in the press. It doesn't work that way. Again, I don't have an issue with AMD subsidizing costs. It's whatever. But it's also not like the devs don't have an interest in it - they're being paid by AMD. I thought this was more or less obvious, but maybe not.