AMD launching open-source GameWorks alternative; GPUOpen

[Pieter3dnow]

Again, that notion is based on Microsoft doing more with DirectX 12 in months than AMD did with Mantle in years.

Your "math" doesn't add up.
(Nothing that big (new API) moves that fast, I kid you not).

And your appeal to "warm feelings" for AMD does as little as PR slides to me...

The only developer that has a DX12 game out is one that is alpha version of Ashes of the singularity and the only reason for it is that they ported the Mantle version

https://youtu.be/eXCXJoRsgJc?t=163

With Frostbite 3 you can play multiple games using Mantle , then again you get your facts wrong.

https://en.wikipedia.org/wiki/Mantle_(API)#Support

Main page: Video games that support Mantle
Battlefield 4[29]
Battlefield Hardline
Thief
Star Citizen[30][31]
Plants vs. Zombies: Garden Warfare
Civilization: Beyond Earth[32]
Dragon Age: Inquisition
Mirror's Edge Catalyst
Sniper Elite III[33]
Star Wars Battlefront[34]

 

[razor1]

Honestly it really seems like Microsoft intentionally only designed DX12 for Windows 10 to promote a higher adoption rate. Windows 10 released with DX12 support, but DX12 potentially could have been available sooner if they designed it with Windows 7/8 compatibility.

Well it probably could have but the driver model has changed, so Win 7/8 would have needed some major changes to support that.

Windows 10 also is a universal OS for all their hardware from mobile to desktop, which DirectX 12 is better suited for.

I like that MS stopped backwards compatibility to older OS's with Direct X 12 it makes it much easier to create newer software when doing this. They took a page out of Apple's book.

Costs less to create do to less testing and of course less to create for older hardware.

BTW Vulkan is going to be the same, it will not support anything older than OGL 4.1 cards I think, pretty much the same limitation as Direct X 12 features wise on hardware.

 

[razor1]

The only developer that has a DX12 game out is one that is alpha version of Ashes of the singularity and the only reason for it is that they ported the Mantle version

https://youtu.be/eXCXJoRsgJc?t=163

With Frostbite 3 you can play multiple games using Mantle , then again you get your facts wrong.

https://en.wikipedia.org/wiki/Mantle_(API)#Support

Yeah but most of those games listed don't play as well as they should on AMD's newest GPU Fiji, when running Mantle.

 

[razor1]

From what I have read and seen, the push for low level apis was instigated by certain game devs in the industry that wanted more low level access to the hardware. That was people and groups like Johan Anderson from DICE.

I imagine they shopped the idea to microsoft and nvidia in addition to amd. It would make more sense to get microsoft to implement such a change as it would be vendor neutral, what they likely got was a more lukewarm response from the non amd parties, or a timeline that was much slower and on the back burner than AMD offered.

You all can make claims about dx12 being fully baked and planned in its current form from 4-5 years ago, but guess what. dx12 barely shipped in the tail end of 2015, and we STILL have no released games that run dx12.

Mantle shipped late 2013, bf4 was the first title that I recall using it and the mantle version was pushed back to early 2014.

That is at LEAST a 2.5 year head start to actual released low level api versions of a shipped game, not just talk.

People on these boards keep bitching about things actually being SHIPPED having more weight than talk and slides, well guess what? THAT IS WHAT MANTLE WAS. That fact that actual GAMES running on the api shipped 2.5 years earlier than the soonest estimated dx12 title suggests to me that AMD pushed harder on the lower level api front than either nvidia or microsoft.

Microsoft might have been working on similar things in the back burner, but they sure as hell were not as aggressive getting it to market compared to amd, mantle showed the proof of concept and actual GAMES released and then and only then did microsoft seem to get off its ass and push faster and harder.


These are suppositions of course, I was not in the back rooms and privy to the internal discussions, and neither were any of you. But we have examples of a YEARS long lead time to actual RELEASED games that can be used as evidence by objective non biased people to support the notion that AMD was the company that pushed harder on this.



What is gained by being so incredulous on this point? Because it might give a positive point in AMDs favor? Even though I prefer amds approach to variable refresh rate monitors and support, I am not going to deny that it was nvidia that pushed that tech forward first. Of all the hills to fight and die on, just to avoid having to say the most infinitesimally positive thing about AMD... against all reasonable evidence. How god damn petty.

Direct X next in Xbox one was already done by the time Mantle was even looked into by AMD, now getting that to a desktop PC's, Microsoft was against that when Steven Balmer was at the helm of MS. But MS has changed its tune with there approach to software development with the new CEO. Direct X next is essentially Direct X 12, well with tweaks and enhancements. This was done with nV's and AMD's and others involvement. This is the reason why both nV's and AMD's latest hardware and even last generation hardware in some cases are fully compatible with DX 12. If they weren't involved in development of DX 12, we would have run across the issue of what happened with the FX series, where MS had to make a cap version for one of the IHV's. And Pieter this answers the question I asked you, whey cap versions are done.

So MS did have a very mature base to start from to create Direct X 12, it wasn't a reactionary event because of Mantle. MS is not a company that reacts quickly to market changes, we have seen that in the past for many years. Actually any company that is that big, its hard to stay nimble and fast reacting because there are many politics and hurdles to get around to create something new.

 

[KazeoHin]

Direct X next in Xbox one was already done by the time Mantle was even looked into by AMD, now getting that to a desktop PC's, Microsoft was against that when Steven Balmer was at the helm of MS. But MS has changed its tune with there approach to software development with the new CEO. Direct X next is essentially Direct X 12, well with tweaks and enhancements. This was done with nV's and AMD's and others involvement. This is the reason why both nV's and AMD's latest hardware and even last generation hardware in some cases are fully compatible with DX 12. If they weren't involved in development of DX 12, we would have run across the issue of what happened with the FX series, where MS had to make a cap version for one of the IHV's. And Pieter this answers the question I asked you, whey cap versions are done.

So MS did have a very mature base to start from to create Direct X 12, it wasn't a reactionary event because of Mantle. MS is not a company that reacts quickly to market changes, we have seen that in the past for many years. Actually any company that is that big, its hard to stay nimble and fast reacting because there are many politics and hurdles to get around to create something new.

Not to be 'That guy' But the Direct X version on XBone was initially Direct X 11, And EVEN IF Microsoft planned secretly to launch DX 12 on the platform, that platform is a highly parallel, AMD-based environment... which is exactly what Mantle is designed around... sooo who developed what tech first?

 

[tybert7]

Microsoft may have had a mature base to spring off of, and low level apis were certainly not a new thing as they already existed in the consoles, but who moved to put that on the pc first?

That seems to have been amd. The idea that a much smaller company like amd beating microsoft to the punch in bringing a low level api to windows had little to no bearing on microsofts own timetable seems implausible to me. I bet nvidia was lobbying microsoft hard behind the scenes for them to get behind their own iteration and release it to market. But the delay and gap of market release and game releases is the most telling to me. If microsoft was really pushing hard for their own low level api like some people here claim they were on the pc, why has/is it taking so much longer to get both dx12 lauched to consumers and actual working games released?

Some people are here claiming that amd saw what microsoft was working on ahead of time, then raced ahead of them to beat them to market. More, that devs like the dice team and others that were asking for more low level apis chose to go with a willing vendor specific partner rather than a cross vendor partner like microsoft.

Really guys? Why would dice bother with amd alone when they did unless amd was the entity that was pushing to actually get this to market earlier? It seems like microsofts time table for release, or lack thereof, was designed to be much later. If it was ever meant to be released at all. And that is exactly what we've seen.

First released mantle game = early 2014
First released dx12 game = mid to late 2016

But sure, microsoft was always pushing hard to bring low level apis to the pc before amd even dreamed of it. It stands to reason the entire reason amd bothered with mantle, along with ea and dice, was precisely because microsoft was either less interested at the time or dragging their feet.

 

[razor1]

Not to be 'That guy' But the Direct X version on XBone was initially Direct X 11, And EVEN IF Microsoft planned secretly to launch DX 12 on the platform, that platform is a highly parallel, AMD-based environment... which is exactly what Mantle is designed around... sooo who developed what tech first?

No it wasn't. It was an extension of Dx 11 with many low level improvements, the precursor to Dx12, if you don't believe me, DL the leaked SDK and look for the Dx12 features, most of them are all in there.

https://en.wikipedia.org/wiki/DirectX

DirectX 11.X is a superset of DirectX 11.2 running on the Xbox One.[34] It actually includes some features, such as draw bundles, that were later announced as part of DirectX 12

There are other features they didn't mention in there like multiple command streams etc. So yeah its got many of the "new" features of Dx12 and Mantle and Vulkan.

 

[razor1]

Microsoft may have had a mature base to spring off of, and low level apis were certainly not a new thing as they already existed in the consoles, but who moved to put that on the pc first?

That seems to have been amd. The idea that a much smaller company like amd beating microsoft to the punch in bringing a low level api to windows had little to no bearing on microsofts own timetable seems implausible to me. I bet nvidia was lobbying microsoft hard behind the scenes for them to get behind their own iteration and release it to market. But the delay and gap of market release and game releases is the most telling to me. If microsoft was really pushing hard for their own low level api like some people here claim they were on the pc, why has/is it taking so much longer to get both dx12 lauched to consumers and actual working games released?

Some people are here claiming that amd saw what microsoft was working on ahead of time, then raced ahead of them to beat them to market. More, that devs like the dice team and others that were asking for more low level apis chose to go with a willing vendor specific partner rather than a cross vendor partner like microsoft.

Really guys? Why would dice bother with amd alone when they did unless amd was the entity that was pushing to actually get this to market earlier? It seems like microsofts time table for release, or lack thereof, was designed to be much later. If it was ever meant to be released at all. And that is exactly what we've seen.

First released mantle game = early 2014
First released dx12 game = mid to late 2016

But sure, microsoft was always pushing hard to bring low level apis to the pc before amd even dreamed of it. It stands to reason the entire reason amd bothered with mantle, along with ea and dice, was precisely because microsoft was either less interested at the time or dragging their feet.

First DX12 game was shown off with Forza, came out well before any mantle game was available for download and it was a full game. don't forget that.

http://www.geforce.com/hardware/technology/dx12/windows-10

March of 2014

That was the first game that used Dx12.

First game to show that used mantle was Battlefield, which was Jan of 2015 I think which was shown off in a video although they did show some basic simulations with Mantle first in Oct. 2013

 

[Anarchist4000]

SPIR-V isn't a shader language, it has the ability to precompile shader code from different sources, that is the only difference it will have, but in the for now it will only use GLSL later on other languages will be added to it. Direct X already has this feature, as with OGL, ES and i think other versions of Ogl as well *which already uses Spir-V.

I never called it a shader language. I said it was an intermediate language. My understanding of mantle was that it took HLSL(Microsoft shader language) and compiled into an ELF format to be consumed by AMD GPUs. Mantle didn't really have an intermediate language to the best of my knowledge. That is probably the biggest iteration from Mantle to Vulkan. What I've seen so far, it allows almost any language to target it. HLSL, GLSL, C, C++, Cuda, etc.

Porting software over to PC, developers have to be aware low level languages don't port over well to different hardware, so there has to be individual paths for different ISV's which just doesn't bode well for AMD hardware... Fiji is a perfect example of how code that works well on older AMD hardware tends to bottleneck on Fiji. Maxwell too has this problem but with nV having 80% of the market developers are forced to look into nV's code path more.

Be aware yes, but it should be trivial to port in many circumstances. At the very least a direct port should at least run without rewriting all the shader code. A Windows to Linux port for example should need minimal tweaks to the graphics API. The shaders being platform agnostic. The biggest current hurdle porting console to PC/Nvidia would probably be dealing with any async behavior.

No they are different. OGL ES integration into Vulkan or using them together like you have stated isn't going to happen. Vulkan's directive is to seamlessly integrate tile based rendering into a full 3d API. There has to be some major changes to accomplish that. Most mobile GPU's use tile based rendering vs fully 3d rendering like PC or console GPU's.

Your comparison was "Vulkan vs Mantle" with Vulkan being tied to OpenGL for mobiles. I never said they would be integrated, but that Vulkan and OpenGL ES would both be available on at least Android platforms for developers. That said I have read somewhere that OpenGL and Vulkan should be able to pass data and work together. I believe it was Vulkan filling a texture and then reusing it through a pointer passed to OpenGL.

As for tile based rendering, I'm not sure if that is abstracted through Vulkan's API, hidden by the actual drivers, or if Vulkan's explicit nature would let you tile however you felt was necessary. It may just be a matter of creating as many tiles as you feel are necessary given device capabilities. If you grab a device that reports a really low amount of available memory or specifies a maximum render target size, it may be up to you to make it work.

From what I have read and seen, the push for low level apis was instigated by certain game devs in the industry that wanted more low level access to the hardware. That was people and groups like Johan Anderson from DICE.

I imagine they shopped the idea to microsoft and nvidia in addition to amd. It would make more sense to get microsoft to implement such a change as it would be vendor neutral, what they likely got was a more lukewarm response from the non amd parties, or a timeline that was much slower and on the back burner than AMD offered.

The low level APIs already existed on consoles to a significant degree. DICE would have had access to these through various SDKs. Getting a low level API for PCs in general I'd imagine was the more likely goal to facilitate ports. It's more a question of if DICE wanted memory management or a different threading model. Explicit memory management should make a devs life easier porting from console to PC as they were already doing that work. Going to a high level API like DX11 or OpenGL would introduce uncertainty into their rendering paths. The async behavior exposed through Mantle would be an entirely different beast as it didn't exist anywhere but AMD's engineering department.

As for which came first, DX12 or Mantle, they probably share that honor. Odds are the idea came out of a joint working group developing specs for XBox. MS had the feature on paper as a possibility for XBox and AMD had implemented it in hardware, but they shelved the feature for any number of reasons: performance, political, simplicity. DICE go to AMD who provide access to an internal testing tool and things develop from there. This is just speculation, but it seems likely.

 

[razor1]

I never called it a shader language. I said it was an intermediate language. My understanding of mantle was that it took HLSL(Microsoft shader language) and compiled into an ELF format to be consumed by AMD GPUs. Mantle didn't really have an intermediate language to the best of my knowledge. That is probably the biggest iteration from Mantle to Vulkan. What I've seen so far, it allows almost any language to target it. HLSL, GLSL, C, C++, Cuda, etc.

Sorry miss read what you posted, Yeah pretty much that's how it works.

Yes that is a change from Mantle to Vulkan,

It will allow it but not initially for Vulkan,

Be aware yes, but it should be trivial to port in many circumstances. At the very least a direct port should at least run without rewriting all the shader code. A Windows to Linux port for example should need minimal tweaks to the graphics API. The shaders being platform agnostic. The biggest current hurdle porting console to PC/Nvidia would probably be dealing with any async behavior.

It is not a trivial change, Async aside which you are correct will create major changes, number of CPU threads, how the data is stored and shifted, resource allocation optimizations, also GPU/CPU horse power available to the consoles vs. the PC. All of these things will affect low level optimizations.

This is why consoles have been able to keep up with PC's in consoles for longer period of time.

Your comparison was "Vulkan vs Mantle" with Vulkan being tied to OpenGL for mobiles. I never said they would be integrated, but that Vulkan and OpenGL ES would both be available on at least Android platforms for developers. That said I have read somewhere that OpenGL and Vulkan should be able to pass data and work together. I believe it was Vulkan filling a texture and then reusing it through a pointer passed to OpenGL.

I stated they will be integrated, that is their goal and they have stated as such.

As for tile based rendering, I'm not sure if that is abstracted through Vulkan's API, hidden by the actual drivers, or if Vulkan's explicit nature would let you tile however you felt was necessary. It may just be a matter of creating as many tiles as you feel are necessary given device capabilities. If you grab a device that reports a really low amount of available memory or specifies a maximum render target size, it may be up to you to make it work.

All I know is what they stated they want to integrate them together, not sure exactly how they are doing it, we will have to wait and see.

The low level APIs already existed on consoles to a significant degree. DICE would have had access to these through various SDKs. Getting a low level API for PCs in general I'd imagine was the more likely goal to facilitate ports. It's more a question of if DICE wanted memory management or a different threading model. Explicit memory management should make a devs life easier porting from console to PC as they were already doing that work. Going to a high level API like DX11 or OpenGL would introduce uncertainty into their rendering paths. The async behavior exposed through Mantle would be an entirely different beast as it didn't exist anywhere but AMD's engineering department.

That is true it would be easier than going to a high level API, and this is why dev's have to pay attention to this things can go wrong, look at Batman AK, that shows an inexperienced dev team having many issues with a port. And to everyone else that pointed Batman AK's performance as crap game works. I have been playing it without game works at mid 40;s fps with everything maxed out at 1440p, but with game works on its performance went down to high 30's, Titan X, and this was before the patch release, the patch solved many issues but still has issues, its just a bad port by an inexperienced team and its not gameworks that caused all of its problems.

As for which came first, DX12 or Mantle, they probably share that honor. Odds are the idea came out of a joint working group developing specs for XBox. MS had the feature on paper as a possibility for XBox and AMD had implemented it in hardware, but they shelved the feature for any number of reasons: performance, political, simplicity. DICE go to AMD who provide access to an internal testing tool and things develop from there. This is just speculation, but it seems likely.

I think that is exactly how it happened. MS started it, AMD pushed it because it was advantageous to them and of course dev's love it too because its extra performance they can use.

 

[tybert7]

First DX12 game was shown off with Forza, came out well before any mantle game was available for download and it was a full game. don't forget that.

http://www.geforce.com/hardware/technology/dx12/windows-10

March of 2014

That was the first game that used Dx12.

First game to show that used mantle was Battlefield, which was Jan of 2015 I think which was shown off in a video although they did show some basic simulations with Mantle first in Oct. 2013

That must have been a typo, the mantle patch was late Jan / Early February 2014 I believe for bf4.

http://www.anandtech.com/show/7728/battlefield-4-mantle-preview


All I see listed for forza was a demo of dx12, where is the released game for dx12? Where is ANY released game (not a demo, a released game???) Those are coming in 2016.

That's two years later to get from demo status to fully released games. If microsoft was really in the catbird seat as far as being the progenitor of releasing a low level api for the pc, they sure are taking a LONG ass time to get it done. Just under 2 years later for the api support to hit, and over 2 years for actual working games.

Clear as day for all to see.

EDIT:

Also, people keep forgetting that ps4 exists too from sony. AMD probably thought (and there was talk of this at the time) that having GCN architechture in the two main consoles would be enough to spur developers to write mantle versions of the games independently of whether or not microsoft ever got on the dx12 path. And any discussions of a lower level api related to the current consoles would HAVE to involve amd at some level by default because it's their hardware in both.

So as I said, it seems likely that the devs wanted to capitalize on the low level access both the xbox AND ps4 had access to on the pc, microsoft would seem a more natural candidate to go to as dx XX is the standard on windows and is cross vendor, but for whatever reason, the dice guys chosen partner was AMD, maybe that was because they got to use gcn on both the consoles and the pc, maybe it was because amd was more aggressive on the time table, but there was SOME reason why they went there first instead of a cross vendor angle, and the late to the party dx12 push suggests to me microsoft was slow rolling the push.

 

[razor1]

That must have been a typo, the mantle patch was late Jan / Early February 2014 I believe for bf4.

http://www.anandtech.com/show/7728/battlefield-4-mantle-preview


All I see listed for forza was a demo of dx12, where is the released game for dx12? Where is ANY released game (not a demo, a released game???) Those are coming in 2016.

That's two years later to get from demo status to fully released games. If microsoft was really in the catbird seat as far as being the progenitor of releasing a low level api for the pc, they sure are taking a LONG ass time to get it done. Just under 2 years later for the api support to hit, and over 2 years for actual working games.

Clear as day for all to see.

EDIT:

Also, people keep forgetting that ps4 exists too from sony. AMD probably thought (and there was talk of this at the time) that having GCN architechture in the two main consoles would be enough to spur developers to write mantle versions of the games independently of whether or not microsoft ever got on the dx12 path. And any discussions of a lower level api related to the current consoles would HAVE to involve amd at some level by default because it's their hardware in both.

So as I said, it seems likely that the devs wanted to capitalize on the low level access both the xbox AND ps4 had access to on the pc, microsoft would seem a more natural candidate to go to as dx XX is the standard on windows and is cross vendor, but for whatever reason, the dice guys chosen partner was AMD, maybe that was because they got to use gcn on both the consoles and the pc, maybe it was because amd was more aggressive on the time table, but there was SOME reason why they went there first instead of a cross vendor angle, and the late to the party dx12 push suggests to me microsoft was slow rolling the push.

Yeah it was Feb of 2014 you are right. API development does take time because there are many hardware IHV's in the mix, 3 major ones that have have different goals, then you add MS on top of that, thats 4 major corporations that have to some what agree on direction, and MS isn't building the hardware they have to take input from all three of them.

I don't think Mantle had much to do with it maybe a little bit but its hard to say. Balmer was against anything with Xbox one to PC development systems being similar, that is specific to protect Xbox one game sales. This was the reason why Dx12 development was slowed down for the PC, its pretty easy to see how MS's policies have changed towards this and may other aspects of Windows, with the new CEO. Windows 10 free upgrades, cross platform use for Windows 10 with phones, tablets and PC's. MS could have done this may years ago starting with Windows 7 but never did till now, the reasons for this is the believe that if phones and mobile devices were to have this they would cannibalize PC sales, which we can see now that wouldn't have happened.

 

[ChosenUno]

The only developer that has a DX12 game out is one that is alpha version of Ashes of the singularity and the only reason for it is that they ported the Mantle version

https://youtu.be/eXCXJoRsgJc?t=163

With Frostbite 3 you can play multiple games using Mantle , then again you get your facts wrong.

https://en.wikipedia.org/wiki/Mantle_(API)#Support

BF Hardline
http://www.hardocp.com/article/2015/04/01/battlefield_hardline_performance_video_card_review/4#.VnKS5JN96Rs

DA: I
http://www.hardwarepal.com/dragon-age-inquisition-benchmark-mantle-vs-directx/
http://www.guru3d.com/articles-pages/dragon-age-inquisition-vga-graphics-performance-benchmark-review,9.html

So 2/6 released games, when turning Mantle on, decreases performance. Such a great technology, isn't it?

 

[TheBlueChanell]

They killed Mantle while it was still in beta. Are you expecting a similar fate for this concept? Probably.

They killed of Mantle because of Vulkan and for the betterment of the industry. Vulkan is also very much tied to this new initiative so all that mantle work will pay off here.

I don't see why people bring this up as a bad thing or because they think it failed.

 

[Factum]

They killed of Mantle because of Vulkan and for the betterment of the industry. Vulkan is also very much tied to this new initiative so all that mantle work will pay off here.

I don't see why people bring this up as a bad thing or because they think it failed.

Try and go back and see what AMD stated when they launched Mantle.
Then compare it to the actual timeline of Mantle.

Just try.

 

[Pieter3dnow]

BF Hardline
http://www.hardocp.com/article/2015/04/01/battlefield_hardline_performance_video_card_review/4#.VnKS5JN96Rs

DA: I
http://www.hardwarepal.com/dragon-age-inquisition-benchmark-mantle-vs-directx/
http://www.guru3d.com/articles-pages/dragon-age-inquisition-vga-graphics-performance-benchmark-review,9.html

So 2/6 released games, when turning Mantle on, decreases performance. Such a great technology, isn't it?

Yeah i have seen these things it is simple if you do not make use of enough batches the cpu can cope with it. But run the same games on a slower cpu and it will pick up the pace.

If games would use 60K+ batches then you would see a really good jump that is where 4 cores+ becomes something special.

To prove the point people laughed at 8 jaguar core chips the consoles use they were saying how can such a shit cpu drive a games console it will be slower then the previous generation they were all touting that the cpu performance would not be enough. Battlefield 4 on PS4 runs higher batch count then the PC version ....

The idea of Mantle(DX12/Vulkan) is still valid just that the bulk of the programming done still very conservative regarding batches.

 

[Pieter3dnow]

First DX12 game was shown off with Forza, came out well before any mantle game was available for download and it was a full game. don't forget that.

http://www.geforce.com/hardware/technology/dx12/windows-10

March of 2014

That was the first game that used Dx12.

First game to show that used mantle was Battlefield, which was Jan of 2015 I think which was shown off in a video although they did show some basic simulations with Mantle first in Oct. 2013

Event Showcase Demo GPU Used
GDC 2014 1st DX12 Game Demo Forza DX12 GeForce GTX TITAN Black
GDC 2015 1st Confirmed DX12 Game Demo Fable Legends DX12 GeForce GTX 980
BUILD 2015 1st 4-way SLI DX12 Demo WITCH CHAPTER 0 [cry] DX12 4 x GeForce GTX TITAN X
BUILD 2015 1st CryEngine DX12 Demo King of Wushu DX12 GeForce GTX 980
BUILD 2015 1st Explicit Multi-adapter DX12 Demo Unreal Engine Race DX12 GeForce + Integrated

Why does it list it like this explain please since you know everything ...

So when GDC14 was around it was DX11 with some Nvidia special nonsence touted as DX12 when next game was up it was confirmed , why would it list such things differently.

Forza never made it anywhere but a game demo we all know how Nvidia does a demonstration it can have wooden screws as long as there is something to admire.

 

[razor1]

Yeah i have seen these things it is simple if you do not make use of enough batches the cpu can cope with it. But run the same games on a slower cpu and it will pick up the pace.

If games would use 60K+ batches then you would see a really good jump that is where 4 cores+ becomes something special.

To prove the point people laughed at 8 jaguar core chips the consoles use they were saying how can such a shit cpu drive a games console it will be slower then the previous generation they were all touting that the cpu performance would not be enough. Battlefield 4 on PS4 runs higher batch count then the PC version ....

The idea of Mantle(DX12/Vulkan) is still valid just that the bulk of the programming done still very conservative regarding batches.

LOL is that the only thing you can think of is batch counts? I think you should go and look into it a bit more

 

[razor1]

Why does it list it like this explain please since you know everything ...

So when GDC14 was around it was DX11 with some Nvidia special nonse touted as DX12 when next game was up it was confirmed , why would it list such things differently.

Forza never made it anywhere but a game demo we all know how Nvidia does a demonstration it can have wooden screws as long as there is something to admire.

I never stated Forza was made to release did I? I stated it was demoed reading comprehension on the fritz again?

It wasn't nV that showed it off, it was Microsoft using nV hardware

Again skewing something to your own benefit doesn't do you anything but make you look like what you truly are. And yes I know alot more then what I post most of the time, because its not worth going to into depths with people of your mental stature. If I wanted to talk to a kid I can talk to my niece, she knows more about computers and programming than you and she is only 9.

 

[Pieter3dnow]

LOL is that the only thing you can think of is batch counts? I think you should go and look into it a bit more

Actually no I have played BF4 on a slower cpu and it worked very well with Mantle. Got decent enough framerates on Ultra where DX11 could not get past a few settings on high and most medium ...

 

[TheBlueChanell]

Try and go back and see what AMD stated when they launched Mantle.
Then compare it to the actual timeline of Mantle.

Just try.

Plans change and the Khronos Group and it's partners showing interest in Mantle/Vulkan most likely changed AMD's plans.

It's a win for the industry and if this is adopted across Android, SteamOS, Linux and the Console's then everybody wins including us, the consumers.

 

[razor1]

Actually no I have played BF4 on a slower cpu and it worked very well with Mantle. Got decent enough framerates on Ultra where DX11 could not get past a few settings on high and most medium ...

So what settings where those and how do they affect batch numbers, because telling ya you read into it wrong, batch amounts only affect the CPU, so do those settings that you had to put down affect the CPU or batch amounts?

 

[JustReason]

So what settings where those and how do they affect batch numbers, because telling ya you read into it wrong, batch amounts only affect the CPU, so do those settings that you had to put down affect the CPU or batch amounts?

I would think distance would and that is usually a setting in most games. Mantle was a boon to minimums and consistency. Frame rate CF in Mantle looked like a straight line. Never before had I seen that with CF/SLI.

 

[ChosenUno]

I would think distance would and that is usually a setting in most games. Mantle was a boon to minimums and consistency. Frame rate CF in Mantle looked like a straight line. Never before had I seen that with CF/SLI.

Not in those 2 games I mentioned

 

[JustReason]

Try and go back and see what AMD stated when they launched Mantle.
Then compare it to the actual timeline of Mantle.

Just try.

I am not sure what you are alluding to here but ...

Mantle was announced and released at given times easily posted so no need to argue that part. Now as to the part about failure ( a completely ignorant and intentional misrepresentation):

http://www.anandtech.com/show/9036/...changing-direction-in-face-of-dx12-and-glnext

As far as “Mantle 1.0” is concerned, AMD is acknowledging at this point that Mantle’s greatest benefits – reduced CPU usage due to low-level command buffer submission – is something that DX12 and glNext can do just as well, negating the need for Mantle in this context. For AMD this is still something of a win because it has led to Microsoft and Khronos implementing the core ideas of Mantle in the first place, but it also means that Mantle would be relegated to a third wheel. As a result AMD is shifting focus, and advising developers looking to tap Mantle for its draw call benefits (and other features also found in DX12/glNext) to just use those forthcoming APIs instead.

Now given the change in environment, it would be a great failure on AMDs part to continue pushing Mantle when in its entirety DX12 does the same thing. But this change, the release of DX12, forced a change in plans from Mantles inception. Its how business works. Its like a basketball defense on the court. It may have an initial plan but must adapt to the offences attack or risk failure.

 

[razor1]

I would think distance would and that is usually a setting in most games. Mantle was a boon to minimums and consistency. Frame rate CF in Mantle looked like a straight line. Never before had I seen that with CF/SLI.

Well he is saying that using a more power CPU it doesn't show any benefits of using lower batch amounts, which is absolutely wrong. His original quote was this

Yeah i have seen these things it is simple if you do not make use of enough batches the cpu can cope with it. But run the same games on a slower cpu and it will pick up the pace.

If games would use 60K+ batches then you would see a really good jump that is where 4 cores+ becomes something special.

To prove the point people laughed at 8 jaguar core chips the consoles use they were saying how can such a shit cpu drive a games console it will be slower then the previous generation they were all touting that the cpu performance would not be enough. Battlefield 4 on PS4 runs higher batch count then the PC version ....

The idea of Mantle(DX12/Vulkan) is still valid just that the bulk of the programming done still very conservative regarding batches.

 

[JustReason]

Not in those 2 games I mentioned

You mentioned instances current after AMDs pull of Mantle. It does not speak to Mantles performance whilst it was being supported. I don't have those games so I can speak to continued support but I can say and it is well documented that Mantle works very well in CIV BE to this day.

Seriously I recommend to most of you if not in school now, please go take a class in debate and rational thought. Too many of the arguments here on this site rarely stay in context of the debate and tend greatly to the side of intentional misrepresentation of the facts just to fit their skewed argument.

 

[JustReason]

Well he is saying that using a more power CPU it doesn't show any benefits of using lower batch amounts, which is absolutely wrong. His original quote was this

I think what he meant was that higher power CPUs get less benefit as they weren't bottle-necked as hard. One would see greater increases with lower CPU SKUs. We saw this with early benchmarks where the 6 core E Intel gained little but the AMD APU gained huge.

 

[razor1]

I think what he meant was that higher power CPUs get less benefit as they weren't bottle-necked as hard. One would see greater increases with lower CPU SKUs. We saw this with early benchmarks where the 6 core E Intel gained little but the AMD APU gained huge.

That would make sense since Intel's CPU's have better IPC and DX11 won't be able to use all the cores efficiently. Changing the settings though won't and shouldn't be necessary in this case.

And his qoute about developers are conservative about using batch counts its just BS, the reason why DX11 games use less batches is because there was a limitation when making a game. If that limtation was not there then I'm sure they will go higher. But when developing a game, as a 3d artist, I myself have done the modelling for levels, I always take that into consideration when creating level assets and placing them into a level editor, optimization isn't only about programming but how levels are built too.

Using texture plates instead of using multiple textures for multiple objects, combining objects together in one mesh file, manual telling the engine through the editor these two object or more should be rendered together, etc. There are many ways to do this, but its always in the back of my mind when doing this type of work. And I still do with Dx12, because even though it might have that limitation removed, I would rather not tax the CPU with this if there are better ways to allocate the CPU for other things.

Talking about Batch amounts is pointless, because the renderer is still rendering what the game developer intended to render. If the map was just one huge 3d file vs a million different objects, its all the same yeah that would be poor level design with either of those two cases, but end results are the same, with huge penalties of course.

Now if you want to have more batches, physics CPU usage increases, memory address space will not be used as efficiently after a certain point of increased batches, bandwidth won't be used as efficiently, etc. There are many other sub systems that will be affected, not just the CPU. All this and the GPU will be tasked more, because now it has to keep track of more objects, along with the CPU. there are so many things that change with batch amounts, its not as simple as saying lets use more batches because the CPU can handle it because of the API.

BTW there is not distance setting in BF4

here is a screenshot of the video settings options.

http://images.bit-tech.net/content_images/2013/11/amd-radeon-r9-290-review/bf4-b.jpg

 

[{NG}Fidel]

I agree with razer on a lot of the points here but I still think this was a good move. The question is will AMD support it? To me it seems clear management wants the GPU team with more autonomy.