AMD @ GDC: Partnership with MS next-generation graphics.

[Gloomy]

No bugs here. A little disappointed in the way reviews for Mantle were handled.

If you subscribe to the narrative that an API change doesn't significantly change graphics card performance, then the "at best" 10% increase in average FPS is likely caused by a proportionate increase in minimum fps. A quick thought experiment: if you have a high of 90 and a low of 30, for an average of 60. An increase by 10% to 66 where the high doesn't change means you went from 30 to 42 on your low.

Hope this puts the subjective "smoothness" people are experiencing into perspective. Also should put general reviewer incompetence into perspective too, lol. :thumbsup:

Then consider the cases where Mantle increased performance by 20%. 30%. 40%. The "low end". You would look pretty stupid, I think, if you were to imply these kinds of gains are low, small, unnoticeable, unwelcome, etc. Actually I'd go so far as to say that these gains are huge, from just a software change, it's shameful so much was left on the table to begin with.

If you think they're insignificant, then imo you have two options open to you if you want to life a happy, fulfilled life; you can join the Amish and move to Pennsylvania to churn butter, as technology isn't something you're mentally equipped to handle... or buy a console and play your video games exclusively on it, with the rest of the Luddites who don't care about progress and actively resist it.

 

[krumme]

I wonder if FrostBite will be easy/easier to port to DX12 now that they have already done the groundwork to port to mantle.

Well Johan more or less says its the same. "Where have i read this before..."
I guess its obvious the same work that needs to be done.

"Baker: APIs come and go. Once you support more than one, it's pretty easy to support a dozen–assuming there is parity in the hardware features, and assuming you don't have to rewrite your shaders in an entirely different language. If you release a title right now, you would end up with likely six paths. An Xbox360, a PS3, a PS4, a Xbox One, a DX9, and a DX11. For us, the graphics system is just a module that talks to the API. All we did for Mantle was replace the D3D module with a Mantle one. It's about 3,000 to 4,000 lines of code for the Mantle version, which took me personally about two months to write. In terms of support, at least for us, it wasn't terribly difficult."

http://www.maximumpc.com/AMD_Mantle_Interview_2014

Now DX12 is perhaps not quite as low level, and adds some complexity, but i guess its easy to go from Mantle to a higher level or what?

"Baker: Relative to Microsoft's Direct3D (D3D), Mantle is indeed more low-level. But it's not low-level in the sense that we are exposed to individual architectural decisions. For example, Mantle still abstracts the details of the shader cores themselves, so that we don't even know if we are running on a vector machine or a scalar machine. What isn't abstracted is the basic way a GPU operates. The GPU is another processor, just like any other, that reads and writes memory. One thing that has happened is that GPUs are now pretty general in terms of functionality. They can read memory anywhere. They can write memory anywhere. A lot of the things an API has traditionally managed aren't really necessary any more. Mantle puts the responsibility onto the developer. Some feel that is too much, but this really isn't any different than managing multiple CPUs on a system, which we have gotten pretty good at. We don't program multiple CPUs with an API, we just handle it ourselves. Mantle gives us a similar capability for the GPU."

 

[positivedoppler]

I think most people underestimate how important Mantle is to AMD's overall strategy. The strategy of Mantle me think is not really to make AMD more competitive against Nvidia, but to make the GPU more important than the CPU taking away Intel's advantage. Mantle I think is a bigger threat to Intel's home PC profit margin than anything else.

So regardless if DirectX12 is a copy of Mantle or something entirely different, AMD is the big winner even if Nvidia ends up benifiting more from Mantle! If DirectX12 ends up being Mantle, AMD will get a slight lead on Nvidia but Nvidia will eventually catch up. This will hardly be worth all the effort AMD's poured into developing Mantle. By gaining widespread adoption of Mantle, the will who will really hurt is Intel as more computer buyers will allocate their budget to the GPU as oppose to the CPU. I think AMD much rather fight against Nvidia for a buyer's budget than go head to head against Intel.

To sum up my theory, suppose every year laptop and PC buyer spend 2 billion dollars. Suppose right now 1 billion is allocated to the cpu while 1 billion is allocated to the GPU. Of that, AMD owns 30% of the gpu market and 10% of the cpu market so they will take home 300 million from gpu and 100 million from cpu. Now suppose DirectX12 takes off and buyers no longer need a top of the line Intel, now buyers will allocated 1.2 billion to the Gpu budget and 800 million to the cpu budget. If AMD maintains the same market share, 30% of 1.2 billion is 360 million and another 80 million from cpu. They'll make 40 million more

 

[Gloomy]

I think most people underestimate how important Mantle is to AMD's overall strategy. The strategy of Mantle me think is not really to make AMD more competitive against Nvidia, but to make the GPU more important than the CPU taking away Intel's advantage. Mantle I think is a bigger threat to Intel's home PC profit margin than anything else.

So regardless if DirectX12 is a copy of Mantle or something entirely different, AMD is the big winner even if Nvidia ends up benifiting more from Mantle! If DirectX12 ends up being Mantle, AMD will get a slight lead on Nvidia but Nvidia will eventually catch up. This will hardly be worth all the effort AMD's poured into developing Mantle. By gaining widespread adoption of Mantle, the will who will really hurt is Intel as more computer buyers will allocate their budget to the GPU as oppose to the CPU. I think AMD much rather fight against Nvidia for a buyer's budget than go head to head against Intel.

To sum up my theory, suppose every year laptop and PC buyer spend 2 billion dollars. Suppose right now 1 billion is allocated to the cpu while 1 billion is allocated to the GPU. Of that, AMD owns 30% of the gpu market and 10% of the cpu market so they will take home 300 million from gpu and 100 million from cpu. Now suppose DirectX12 takes off and buyers no longer need a top of the line Intel, now buyers will allocated 1.2 billion to the Gpu budget and 800 million to the cpu budget. If AMD maintains the same market share, 30% of 1.2 billion is 360 million and another 80 million from cpu. They'll make 40 million more

This is flawed. If CPU overhead is reduced (and Mantle reduces that A LOT, DX12 should be similar) then developers will find some other way to use the extra cycles. This could be better AI, bigger environments, better physics, sound, you name it. There is quite a lot they could do better.

 

[Erenhardt]

The angst towards Mantle was driven by it's exclusive nature,

And you are disgusted by exclusive features, aren't you?
SIG: "EVGA GTX 650 Ti SSC 1 GB PhysX".

At least be consistent...

I believe that when dx12 comes out, mantle will not be the same mantle we have now. Not only will it do a better job in utilizing CPU, but will make a better use of unused parts of GPUs.

Mantle does two important things.
Gives big performance boost.
Is great developer tool.

Even if dx12 manages to get the overhead sorted, will it be any different as an dev tool?

 

[positivedoppler]

This is flawed. If CPU overhead is reduced (and Mantle reduces that A LOT, DX12 should be similar) then developers will find some other way to use the extra cycles. This could be better AI, bigger environments, better physics, sound, you name it. There is quite a lot they could do better.

Perhaps developers will find better ways to utilize the cpu, but it's hard to deny that almost every single website that has reviewed Mantle crowned low end cpu owners as the biggest winner of Mantle. I mean is it logical for more people follow the direction of the winner?

 

[Gloomy]

Perhaps developers will find better ways to utilize the cpu, but it's hard to deny that almost every single website that has reviewed Mantle crowned low end cpu owners as the biggest winner of Mantle. I mean is it logical for more people follow the direction of the winner?

Intel CPUs are still faster even with Mantle. AMD CPUs win nothing but last place.

 

[R0H1T]

And you are disgusted by exclusive features, aren't you?
SIG: "EVGA GTX 650 Ti SSC 1 GB PhysX".

At least be consistent...

I believe that when dx12 comes out, mantle will not be the same mantle we have now. Not only will it do a better job in utilizing CPU, but will make a better use of unused parts of GPUs.

Mantle does two important things.
Gives big performance boost.
Is great developer tool.

Even if dx12 manages to get the overhead sorted, will it be any different as an dev tool?

Perhaps we'll even have games utilizing HSA properly, now wouldn't that be awesome

 

[krumme]

This is flawed. If CPU overhead is reduced (and Mantle reduces that A LOT, DX12 should be similar) then developers will find some other way to use the extra cycles. This could be better AI, bigger environments, better physics, sound, you name it. There is quite a lot they could do better.

Well it could easily be both. The consoles play bf4 using 6 times 3.5mm2 cpu cores for a total of 21mm2 sans l2. Its miniscule. A low level api and more cores easily and for very low cost releases plenty of cpu power for games.

We can have entire new types of rts games. Far better ai. Better psysics. And all at lower cost. We can have our cake and eat it.

 

[yulgrhet]

Am to be thinking Mantle PS4 focus API, DX 12 to including XB1 focus Mantle variant.

 

[Paul98]

This is great, it's nice to see things finally moving forward again. Should allow much easier porting between console and PC. Allow developers to build something around Mantle starting now and port over to DX12 when it comes out.

 

[cytg111]

This is great, it's nice to see things finally moving forward again. Should allow much easier porting between console and PC. Allow developers to build something around Mantle starting now and port over to DX12 when it comes out.

Biting the bullet(as in costs) now with mantle, getting dev support behind it, AMD may have a bigger say about the lowlevel internals in dx12. - makes me wonder if that was not the objective from the get-go. Speaking of MS, and this is just in my mind, if we're going to see dx12 outside of windows9 then its a direct consequence to a competitive envoriment. So to put it in other words, if mantle does not gain momentum, be ready to invest in windows9.

Also great news for XB1 owners .. thats like a getting a gift.. for free

 

[3DVagabond]

They had "discussions" four years ago, the same way that DICE had "discussions" with Microsoft- and DICE had no success convincing Microsoft to improve their API, and had to go to AMD. Then the actual active development starts last year, oddly enough when Mantle appears on the scene.

Yes, amazing how the actual quote gets turned into something completely different. It goes from a conversation to being deeply involved.

DX11 MT was broken. Several devs have said it. And thats why it didnt have any support. Secondly it was not worth it, as it wasnt nessesary for the consoles.

This is also forgotten. It was just marketing BS, hype, and spin by AMD and their paid shills. Now MS is promising to fix it with, funny enough, low level hardware access, and it's "All hail MS, oh and nVidia too". Flip flopping and flag waving like this is embarrassing.

No bugs here. A little disappointed in the way reviews for Mantle were handled.

If you subscribe to the narrative that an API change doesn't significantly change graphics card performance, then the "at best" 10% increase in average FPS is likely caused by a proportionate increase in minimum fps. A quick thought experiment: if you have a high of 90 and a low of 30, for an average of 60. An increase by 10% to 66 where the high doesn't change means you went from 30 to 42 on your low.

Hope this puts the subjective "smoothness" people are experiencing into perspective. Also should put general reviewer incompetence into perspective too, lol. :thumbsup:

Then consider the cases where Mantle increased performance by 20%. 30%. 40%. The "low end". You would look pretty stupid, I think, if you were to imply these kinds of gains are low, small, unnoticeable, unwelcome, etc. Actually I'd go so far as to say that these gains are huge, from just a software change, it's shameful so much was left on the table to begin with.

If you think they're insignificant, then imo you have two options open to you if you want to life a happy, fulfilled life; you can join the Amish and move to Pennsylvania to churn butter, as technology isn't something you're mentally equipped to handle... or buy a console and play your video games exclusively on it, with the rest of the Luddites who don't care about progress and actively resist it.

I question the sincerity or the abilities of any reviewer who played Mantle down as no big deal. It's obvious that it's a game changer.

 

[Shivansps]

Funny enoght, even his puppet "Star Swarm" supports command lists that, according to AMD, are broken.

 

[sontin]

Biting the bullet(as in costs) now with mantle, getting dev support behind it, AMD may have a bigger say about the lowlevel internals in dx12.

They have nothing to say. Microsoft writes the drivers for the Xbox One.

BTW: nVidia is the only partner which has been released a DX12 driver:

NVIDIA is currently the only GPU vendor to have a DX12 capable driver in the hands of developers and the demo that Microsoft showed at GDC was running on a GeForce GTX TITAN BLACK card.

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Talks-DX12-DX11-Efficiency-Improvements

 

[PPB]

Funny enoght, even his puppet "Star Swarm" supports command lists that, according to AMD, are broken.

It does when running the DX 11 path. They were pretty clear they spent more time optimizing for that path, and that includes using every stunt you can to have better performance, even if such stunts involve using features that dont really work as well as should (in this case: command lists).

I would rather use a bent screwdriver to remove a screw that use no screwdrive at all. Doesnt deny the fact that the screwdriver is bent and prolly will be a less apt tool than a straight screwdriver

PS: Johan Andersson said that command lists are fundamentally broken, getting your facts straight is key in making at least less silly assumptions.

 

[sontin]

Johan Andersson needs to advertise his API. So he has no problem to lie about other things.

I mean this guy said it is "crucial" that Windows 7 must get DX12 but yet he has no problem to release a new API on a much smaller hardware configuration than Windows 8...

 

[Carfax83]

Saying "work began" 4 years ago on DX12 is not a compliment, or an endorsement or a positive. It's an extremely puzzling statement at best because that would mean it will take 6+ years for DX12 to see be released.

Are you an armchair software engineer now that knows what it takes to design an API? You seem as though you believe things like this just spontaneously occur after a group of guys get together for coffee and donuts or something

AMD had been working on Mantle for 3 years before it was announced, and Mantle only involved on architecture, and not several. Also, as ams23 pointed out repeatedly, Mantle was spearheaded by AMD without the involvement of other IHVs, which would speed up the process.

Source please.

I should have said AMD mislead people into believing that they wanted Mantle to be an open industry standard with adoption by other IHVs....not as a replacement for DirectX..

Source

It was never intended for that though. You don't design an open standard without the cooperation of other IHVs from inception.

Warning issued for inflammatory language.
-- stahlhart

 

[Carfax83]

Only AMD haters would give credence to such a comment, I've been beta testing win8 (& later versions) for nearly 3yrs now & I can safely say that there was no DX12 4yrs back & there wouldn't have been one in 2015 if it weren't for the utter failure of win8, on all platforms !

Talk about self aggrandizement. If DX12 is just now in alpha stage, why on Earth would Microsoft tell you, a mere beta tester, anything about it?

 

[Carfax83]

How about Dan Baker, the creator of Star Swarm demo himself?

"The main difference between OpenGL and D3D is that D3D made an attempt to be threaded, and failed, whereas OpenGL has not yet attempted it. "

http://www.maximumpc.com/AMD_Mantle_Interview_2014

DX11 MT is so broken, that Dan Baker himself implemented it in TWO of his game engines; the Lore engine and the Nitrous Oxide engine.

The first implementation allowed NVidia to thoroughly trounce AMD in CPU limited scenarios in Civilization 5, and as sontin said, deferred contexts doubles, or more than doubles in some cases the performance in the Star Swarm benchmark..

DX11 MT may not be nearly as good as Mantle, but to call it broken is silly.

 

[Erenhardt]

^its not broken. Bent, but not broken.

 

[monstercameron]

DX11 MT is so broken, that Dan Baker himself implemented it in TWO of his game engines; the Lore engine and the Nitrous Oxide engine.

The first implementation allowed NVidia to thoroughly trounce AMD in CPU limited scenarios in Civilization 5, and as sontin said, deferred contexts doubles, or more than doubles in some cases the performance in the Star Swarm benchmark..

DX11 MT may not be nearly as good as Mantle, but to call it broken is silly.

Broken and non-functional are different things...its not very good according to Dan baker himself, I don't understand what you are getting at?

 

[Gloomy]

I've already addressed deferred context's performance increase before. Using the final result as proof that Mantle is unnecessary is extremely deceptive as Nvidia performance in it with deferred context enabled see-saws like crazy.

Anyway it looks like their new driver will likely fix these issues-- 15 FPS increase in average is no joke, that can be +30 to their minimums like I mentioned above.

Using the performance you can get now as proof is absolutely insane though. Why would you do try to deceive people like that?

 

[sontin]

No, "broken and non-functional" have the same meaning.
I mean it would be the same to call DX11.1 and DX11.2 "broken" because nVidia only supports FeatureLevel 11_0.

Deferred Context works just fine. However the implementation is not as smooth as with Mantle or DX12. But that it's normal for a lot of techniques - see for example Tessellation.

 

[Carfax83]

And you are disgusted by exclusive features, aren't you?
SIG: "EVGA GTX 650 Ti SSC 1 GB PhysX".

At least be consistent...

At least educate yourself before making unfounded statements. PhysX isn't platform exclusive. It runs on damn near every platform you can imagine, from PC to console to mobile..

And NVidia is porting several aspects of PhysX to DirectCompute with the Flex initiative:

PhysXInfo.com:Is FLEX purely GPU accelerated library, or will it support CPU execution? Is it plausible to see FLEX ported to OpenCL or DirectCompute?

Miles Macklin: Right now we have a CUDA implementation and a DirectCompute implementation is planned. We are considering a CPU implementation.
I have also built FLEX for Linux (Ubuntu 12.04 64bit) and it works great, in some cases it is faster than Windows.

Source

Even if dx12 manages to get the overhead sorted, will it be any different as an dev tool?

DirectX has always been a great dev tool, with great support and documentation. That's mostly why it's exceeded OpenGL over the years in developer uptake.

I don't see DirectX 12 being any different in that regard.