The AMD Mantle Thread

[The Alias]

No, why make a fallacy?
Fact is OpenGL can do 100.000 drawcalls...today.
(Actually it can do millions if you just focus on drawcalls)
Sorry if that burst Mantle's PR campaign...but facts are facts.

fact also is the 100,00 draw calls opengl can do would never be in an actual gameplay situation; whereas mantle's would be .

 

[PPB]

From the moment he thinks that every single object and texture represents a draw call in a contemporary game, you know he's not even familiar with basic concepts such as proxies when working on 3D modeling/rendering, let alone grasp how those concepts are applied in video game engines to reduce total draw calls.

 

[Paul98]

Well, despite its higher overhead, Dx has a proven track record making gaming possible across multiple architectures. Obviously one is always looking for improvement. But there is rarely a "free lunch" so to speak. Improvement in one area comes at a sacrifice in others. For mantle, obviously the improvement in efficiency comes at the cost of cross platform compatibility. IMO, it will take considerable time to see how much improvement mantle gives in a wide variety of games and how (or if) it can be adopted to a wide variety of situations.

People keep on talking about compatibility, yet forget that DX is not only windows only but depends on the version of windows.

OpenGL is cross platform compatible, yet most games in windows still use DX.

AMD also wants mantle to be cross platform, OSX, mobile, linux, whatever else.

Mantle is what people have been wanting for years, and you don't sacrifice performance as anything that DX or OpenGL can do you will be able to do on Mantle. But what you get to do with mantle is do what you want to do, not what DX thinks you should do. You aren't constrained by the API. Lets say you are drawing a simple picture onto the screen. If you try to do it using one API's function you might be able to draw the picture at 10fps, another you might get 100, and another you might get 1000. Depending on the API and the function you can get a HUGE difference in speed.

 

[Paul98]

From the moment he thinks that every single object and texture represents a draw call in a contemporary game, you know he's not even familiar with basic concepts such as proxies when working on 3D modeling/rendering, let alone grasp how those concepts are applied in video game engines to reduce total draw calls.

But he isn't the only one, otherwise he would be called out on it as soon as he posted it. Most people have no clue what any of this stuff means, you can read all about it but unless you actually have spent time with DX and OpenGL, have worked on a game you really won't have much of an understanding of what these things mean.

 

[Noctifer616]

EA spokespeople told SemiAccurate that the Mantle patch was still on and still full steam ahead, although sadly not on Steam, and will be out in January. The full statement from them is as follows: “After much consideration, the decision was made to delay the Mantle patch for Battlefield 4. AMD continues to support DICE on the public introduction of Mantle, and we are tremendously excited about the coming release for Battlefield 4! We are now targeting a January release and will have more information to share in the New Year.”

Looks like it's going to be the next year.

 

[DiogoDX]

Today we got this official statement from AMD

After much consideration, the decision was made to delay the Mantle patch for Battlefield 4. AMD continues to support DICE on the public introduction of Mantle, and we are tremendously excited about the coming release for Battlefield 4! We are now targeting a January release and will have more information to share in the New Year.

http://www.pcper.com/news/Graphics-Cards/AMD-Mantle-Battlefield-4-Delayed-New-Year

 

[frozentundra123456]

People keep on talking about compatibility, yet forget that DX is not only windows only but depends on the version of windows.

OpenGL is cross platform compatible, yet most games in windows still use DX.

AMD also wants mantle to be cross platform, OSX, mobile, linux, whatever else.

Mantle is what people have been wanting for years, and you don't sacrifice performance as anything that DX or OpenGL can do you will be able to do on Mantle. But what you get to do with mantle is do what you want to do, not what DX thinks you should do. You aren't constrained by the API. Lets say you are drawing a simple picture onto the screen. If you try to do it using one API's function you might be able to draw the picture at 10fps, another you might get 100, and another you might get 1000. Depending on the API and the function you can get a HUGE difference in speed.

If I recall correctly, it was the developers, not AMD that were proposing mantle to be cross platform. If you have a link where AMD says mantle is compatible with other hardware, feel free to post it. In any case, I very seriously doubt intel and nVidia will ever adopt it, both for "political" reasons, and the likely hood that it would be much less efficient on anything but gcn hardware. The mantle proponents can restate the supposed advantages of mantle as many times as they like, but that is just a deflection. It doesnt remove the compatibility and fragmentation issues. As far as different versions of DX, I do not like the way microsoft is limiting the newest version of Dx to windows 8, but cross platform compatibility is much less an issue than with mantle. No matter what hardware you have, there is a version of Dx that will work, unlike mantle, which will require specific hardware, and hardware which is a very small portion of the overall graphics spectrum.

 

[SiliconWars]

Why wouldn't AMD want Mantle to be cross platform? If you mean multi-vendor (ie works for Nvidia and Intel), did you ever stop to consider that maybe they just don't care? AMD is a business, not a charity handing out the fruits of their labour.

 

[3DVagabond]

http://www.pcper.com/news/Graphics-Cards/AMD-Mantle-Battlefield-4-Delayed-New-Year

Nice of them to tell us before the end of the year. I hope they weren't too put out having to take the time to batch email a one liner. /sarc

 

[bystander36]

Why wouldn't AMD want Mantle to be cross platform? If you mean multi-vendor (ie works for Nvidia and Intel), did you ever stop to consider that maybe they just don't care? AMD is a business, not a charity handing out the fruits of their labour.

The industry will care when we start having to support many different platforms using different API's. It seems exciting to some dev's now, but I doubt they'll be all that excited when they are supporting DirectX, Mantle, NvidiaX and IntelZ.

 

[dacostafilipe]

Fragmentating API is not an improvement.

And doing nothing does not improve it either ...

 

[Paul98]

If I recall correctly, it was the developers, not AMD that were proposing mantle to be cross platform. If you have a link where AMD says mantle is compatible with other hardware, feel free to post it. In any case, I very seriously doubt intel and nVidia will ever adopt it, both for "political" reasons, and the likely hood that it would be much less efficient on anything but gcn hardware. The mantle proponents can restate the supposed advantages of mantle as many times as they like, but that is just a deflection. It doesnt remove the compatibility and fragmentation issues. As far as different versions of DX, I do not like the way microsoft is limiting the newest version of Dx to windows 8, but cross platform compatibility is much less an issue than with mantle. No matter what hardware you have, there is a version of Dx that will work, unlike mantle, which will require specific hardware, and hardware which is a very small portion of the overall graphics spectrum.

AMD has already said that they want it to support multiple platforms.

But you missed the point, you keep on talking about compatibility with DX. DX is only compatible with windows and the version that you can use depends on the version of windows. This isn't for any other reason than they want people to upgrade their systems for new version of DX.

Mantle on the other hand works only with GCN( maybe forever, or at least quite a while). But mantle will work on multiple version of windows, and they want to have cross platform compatibility with OSX, mobile, linux and whatever else.

The point I am making is that DX is windows only any hardware, Mantle should eventually be any platform only GCN hardware. They both have something limiting their compatibility.

 

[SiliconWars]

The industry will care when we start having to support many different platforms using different API's. It seems exciting to some dev's now, but I doubt they'll be all that excited when they are supporting DirectX, Mantle, NvidiaX and IntelZ.

They've all made it clear that they want shot of DX anyway so given the chance they'll take it. DX is so unwieldy that it would probably be easier to support Mantle and Nvidia's version than it is to support DX alone.

Anyway, what is AMD supposed to do - wait until Nvidia decided to make their own API instead? This day was always coming, it was only ever a case of who jumped first.

 

[bystander36]

AMD has already said that they want it to support multiple platforms.

But you missed the point, you keep on talking about compatibility with DX. DX is only compatible with windows and the version that you can use depends on the version of windows. This isn't for any other reason than they want people to upgrade their systems for new version of DX.

Mantle on the other hand works only with GCN( maybe forever, or at least quite a while). But mantle will work on multiple version of windows, and they want to have cross platform compatibility with OSX, mobile, linux and whatever else.

The point I am making is that DX is windows only any hardware, Mantle should eventually be any platform only GCN hardware. They both have something limiting their compatibility.

Do you mind posting a quote from AMD, not a dev, that says this? If it was going to be supported on all brands, you'd think it would support all AMD systems too, but it is only their GCN cards. This is the reality of a close to the metal type of API. Close to the metal means it is tied closely to the hardware it is written for.

I wouldn't mind a new API. I just think it should be one that supports all of the PC industry, and not such a small segment.

DX9 supports all gaming hardware and Windows versions that are capable of playing a modern game. DX11 is close to claiming the same. I don't think it is much of an issue.

 

[Paul98]

Do you mind posting a quote from AMD, not a dev, that says this? If it was going to be supported on all brands, you'd think it would support all AMD systems too, but it is only their GCN cards. This is the reality of a close to the metal type of API. Close to the metal means it is tied closely to the hardware it is written for.

I wouldn't mind a new API. I just think it should be one that supports all of the PC industry, and not such a small segment.

DX9 supports all gaming hardware and Windows versions that are capable of playing a modern game. DX11 is close to claiming the same. I don't think it is much of an issue.

How about you read what I wrote...

If you want to talk compatibility look at OpenGL

 

[bystander36]

They've all made it clear that they want shot of DX anyway so given the chance they'll take it. DX is so unwieldy that it would probably be easier to support Mantle and Nvidia's version than it is to support DX alone.

Anyway, what is AMD supposed to do - wait until Nvidia decided to make their own API instead? This day was always coming, it was only ever a case of who jumped first.

Dev's are tired of high level API's on Windows, but most recognize the reason for it. Sure it would be nice to code close to the metal, but close to the metal means it only gets used on a few systems.

The alternative to high level API's, is an API for every platform. That means a lot more work, and likely a lot more bugs and patches. And they will be much slower at fixing them, as they have to fix different problems on every version.

It might be possible to improve DX. I'm sure it can be, but improving DX is probably a much better solution than a new API for every platform. It might be possible to fix OpenGL. Maybe there could be a new API, but not one built by AMD or Nvidia, one that they work together on. It however will not excite dev's, because it will have to be higher level to be compatible with what ever is thrown at it.

 

[bystander36]

How about you read what I wrote...

If you want to talk compatibility look at OpenGL

Both are close to as compatible for actual gaming. DX isn't quite as compatible, but it is close enough for them to use the advantages it brings.

There are obviously trade offs, but losing 1% (made up number, but not far off) of their player base to use an API that is easier to develop, is worth it to them. How is 4 API's better? (Mantle, Nvidia current arch, Intel's current arch and Dx for what doesn't work on those).

 

[frozentundra123456]

AMD has already said that they want it to support multiple platforms.

But you missed the point, you keep on talking about compatibility with DX. DX is only compatible with windows and the version that you can use depends on the version of windows. This isn't for any other reason than they want people to upgrade their systems for new version of DX.

Mantle on the other hand works only with GCN( maybe forever, or at least quite a while). But mantle will work on multiple version of windows, and they want to have cross platform compatibility with OSX, mobile, linux and whatever else.

The point I am making is that DX is windows only any hardware, Mantle should eventually be any platform only GCN hardware. They both have something limiting their compatibility.

You can't seriously believe you own reasoning in that last paragraph can you? Mantle will be limited to perhaps 10 to 15 percent of the PC market, while DX is applicable to all PCs not running IoS or Linux, which must be well above 90% of the market. To say they have similar compatibility issues is simply absurd.

 

[SiliconWars]

Dev's are tired of high level API's on Windows, but most recognize the reason for it. Sure it would be nice to code close to the metal, but close to the metal means it only gets used on a few systems.

The alternative to high level API's, is several API's for every platform. That means a lot more work, and likely a lot more bugs and patches. And they will be much slower at fixing them, as they have to fix different problems on every version.

That's the thing though - it probably won't. The effort required in working around DX in order to get the kind of games they want is at crisis level.

Had there been 10 different vendors like in the 90's then yes, possibly. As it is you're only really talking about AMD and Nvidia. Mantle + Nvidia's API combined will almost certainly be easier and faster to develop on than DX is alone, assuming Nvidia does theirs right.

It might be possible to improve DX. I'm sure it can be, but improving DX is probably a much better solution than a new API for every platform. It might be possible to fix OpenGL. Maybe there could be a new API, but not one built by AMD or Nvidia, one that they work together on. It however will not excite dev's, because it will have to be higher level to be compatible with what ever is thrown at it.

You're just not going to get a high level multi-vendor API that gets near Mantle's performance. Who is going to create it anyway?

The cat is out of the bag, Nvidia has no real alternative but to make their own and they are almost certainly doing so right now. I just don't buy the idea that this is bad for the industry - it's bad for Nvidia right now because they are so far behind in development - but when they catch up it'll be good for everyone.

 

[SiliconWars]

You can't seriously believe you own reasoning in that last paragraph can you? Mantle will be limited to perhaps 10 to 15 percent of the PC market, while DX is applicable to all PCs not running IoS or Linux, which must be well above 90% of the market. To say they have similar compatibility issues is simply absurd.

The problem with DX is it's a barrier to low cost gaming.

First you have the M$ Windows tax just to get DX. Then you have the issue with lower end hardware performing much worse than it should.

In reality the DX gaming market is nowhere near 90%, and is likely less than what AMD can reasonably get with GCN (all GCN APU's will be some 40 million units this year with another 40 million consoles). That's a very big potential market with half of it being "new" because of Mantle empowering APU's.

 

[bystander36]

You're just not going to get a high level multi-vendor API that gets near Mantle's performance. Who is going to create it anyway?

The cat is out of the bag, Nvidia has no real alternative but to make their own and they are almost certainly doing so right now. I just don't buy the idea that this is bad for the industry - it's bad for Nvidia right now because they are so far behind in development - but when they catch up it'll be good for everyone.

You do realize that they will need different API's for different architectures within the same brand, if they plan to make changes.

What cat is out of the bag? We still don't have anything to even see the value of Mantle yet. Just a bunch of talk and PR. And yes, Dev demos are PR. PR to sell their engine to other dev's, and PR to get people to by AMD GCN cards. We have to wait to see the actual value of it.

And how much improvement is needed to be worth the effort of many different API's? If we only see 10%-20% improvements, is it worth the added costs of having to code for 2 additional API's?

 

[Lonbjerg]

The problem with DX is it's a barrier to low cost gaming.

First you have the M$ Windows tax just to get DX. Then you have the issue with lower end hardware performing much worse than it should.

In reality the DX gaming market is nowhere near 90%, and is likely less than what AMD can reasonably get with GCN (all GCN APU's will be some 40 million units this year with another 40 million consoles). That's a very big potential market with half of it being "new" because of Mantle empowering APU's.

Are you messing Mantle and consoles together now again?

/topic
Funny how this "easy to implement with low man hours"...seems to drag out another month...color me surprised

 

[3DVagabond]

Dev's are tired of high level API's on Windows, but most recognize the reason for it. Sure it would be nice to code close to the metal, but close to the metal means it only gets used on a few systems.

The alternative to high level API's, is an API for every platform. That means a lot more work, and likely a lot more bugs and patches. And they will be much slower at fixing them, as they have to fix different problems on every version.

It might be possible to improve DX. I'm sure it can be, but improving DX is probably a much better solution than a new API for every platform. It might be possible to fix OpenGL. Maybe there could be a new API, but not one built by AMD or Nvidia, one that they work together on. It however will not excite dev's, because it will have to be higher level to be compatible with what ever is thrown at it.

Do you think they haven't considered everything you've stated here?

 

[SiliconWars]

You do realize that they will need different API's for different architectures within the same brand, if they plan to make changes.

That's something that the vendors will have to decide is worth it or not. It's one of the reasons why Mantle is possible now - AMD made the decision that GCN would be a long lasting arch.

What cat is out of the bag? We still don't have anything to even see the value of Mantle yet. Just a bunch of talk and PR. And yes, Dev demos are PR. PR to sell their engine to other dev's, and PR to get people to by AMD GCN cards. We have to wait to see the actual value of it.

In which case why bother discussing it? If Mantle doesn't give compelling cost, time to market or performance benefits then it'll die all by itself.

And how much improvement is needed to be worth the effort of many different API's? If we only see 10%-20% improvements, is it worth the added costs of having to code for 2 additional API's?

Like I said, it will probably cost less than DX - a lot less. Consoles to Mantle is very easy and there is no reason why Nvidia couldn't have something very similar, but they have to accept that is how it is done. They were the ones who didn't want the console wins because of low margins.

Console to Mantle and Console to Nvidia's Mantle can easily be faster, easier and cheaper than Console to DX. The performance improvements are just a benefit to the vendors.

 

[3DVagabond]

Are you messing Mantle and consoles together now again?

/topic
Funny how this "easy to implement with low man hours"...seems to drag out another month...color me surprised

Well, it seems to be issues with the game not the API.